Research and analysis

Supporting case studies: What works for AI upskilling in the UK

Updated 27 July 2026

Applies to England

Acknowledgements

This report received support from The British Academy Policy-led Innovation Fellowship award IF24RBDS\240034, in partnership with Skills England. Input from different areas and sectors helped shape the workshop designs and findings. The opinions expressed in this report are those of the author and do not necessarily reflect the views of the department.

 1. How to read these case studies

This document forms part of the SKAI programme, Skills for AI: What Works for AI Upskilling in the UK. The programme translates national evidence on AI skills into 4 connected outputs: employer guide, supporting case studies, an insight briefing based on research, and a research evidence, analysis and methodology report. This case study report examines how organisations across the UK are developing AI capability in practice, spanning a range of sectors, organisational sizes, and delivery models, including large employers, SMEs, training providers, and public services.

The purpose of these case studies is illustrative rather than representative. Each example reflects a specific organisational context, set of constraints, and stage of AI adoption. Taken together, they provide insight into how different approaches to AI upskilling operate in real settings, the challenges organisations encounter, and the conditions that support sustained capability‑building. These case studies should be used to understand how different approaches work in practice, rather than as a complete or representative picture of all organisations.

The case studies were developed through qualitative interviews and validated with participating organisations. They form a core part of the evidence base underpinning the wider SKAI programme. Cross‑cutting patterns from these cases inform the principles set out in the employers’ guide and the analysis presented in the research evidence, analysis and methodology report, while this document preserves the detail needed to understand how those principles are enacted in practice.

2. Case studies grouped by model

Figure 1 summarises the case studies in this section, showing the type of organisation, the challenges they faced, the approaches they took, and the impact observed in practice.

By looking at both what organisations did and what changed as a result, this section highlights practical approaches to AI upskilling that could be used more widely. These examples show how different models can work in different settings, and what it takes to build AI capability at scale.

Note on reported outcomes

The outcomes described in these case studies are based on evidence reported by participating organisations during interviews and validation discussions. They have not been independently audited or verified as performance data. They should therefore be read as self-reported, practice-based examples of how AI upskilling is being implemented and experienced in different organisational contexts. Where productivity gains, time savings or other outcomes are mentioned, these reflect the organisation’s own reported evidence.

Figure 1. Summary of AI upskilling case studies

Note that ‘observed impact’ reflects the specific evidence reported in the workshops and interviews, and ‘key outcomes/value’ summarises the broader organisational value drawn from evidence.

Airbus: Building AI capability through long-term workforce transformation

Organisation type/scale: Large multinational aerospace and defence organisation

Main challenge: Developing AI capability across a complex, global workforce beyond specialist roles

Main intervention: Structured learning pathways, competency frameworks, communities of practice, and governance-linked (partly mandatory) training (for example for GenAI access)

Nature of the journey: From data foundations to organisation-wide AI capability embedded in workforce planning and organisational learning systems

Observed impact (evidence from practice): Employees transitioned into data and AI roles, AI applied across functions, sustained engagement with learning, in a pilot involving approximately 2,000 employees (primarily office-based) users reported saving up to 4 hours per week

Key outcomes/value: Built internal capability at scale, improved workforce confidence and mobility, and demonstrated productivity gains in some functions

KPMG: Embedding AI upskilling as an organisational capability

Organisation type/scale: Large professional services firm

Main challenge: Enabling responsible, scalable AI adoption in a regulated environment

Main intervention: Transformation-led learning model, leadership sponsorship, Spark (embedded in Microsoft Teams), apprenticeships, and employee-led communities

Nature of the journey: From AI literacy and trust-building to role-specific application and continuous learning embedded in workflows

Observed impact (evidence from practice): AI applied in day-to-day professional tasks, evidence of time savings and efficiency gains in specific use cases (for example substantial reductions in analysis time in apprenticeship projects), increased employee confidence and engagement

Key outcomes/value: Strong organisation-wide engagement, evidence of productivity gains in applied contexts, and enhanced capability to support clients

Roche: Building enterprise-wide AI capability through responsible upskilling

Organisation type/scale: Large multinational pharmaceutical organisation

Main challenge: Rapidly upskilling a large global workforce under strict regulatory constraints

Main intervention: Mandatory global training, hybrid delivery, integration with approved AI tools, and local champion-led learning

Nature of the journey: From specialist-led AI use to enterprise-wide baseline capability with ongoing reinforcement

Observed impact (evidence from practice): Cultural shift towards greater confidence in AI use, measurable productivity gains in specific functions (for example in one function, around 50% of employees reported saving 4 or more hours per week), strong engagement across workforce

Key outcomes/value: Improved responsible AI use, measurable time savings in some functions, and strengthened organisational capability

Vertis Media: Building in-house AI capability in a creative SME

Organisation type/scale: Small UK-based marketing and advertising SME

Main challenge: Adapting to AI disruption with limited internal resources

Main intervention: In-house training, shared dashboards, workflow integration, and continuous learning

Nature of the journey: From experimentation to structured adoption and strategic repositioning

Observed impact (evidence from practice): Reduced time spent on reporting and research, AI integrated into workflows, new services developed, contributed to an estimated 20% increase in turnover

Key outcomes/value: Reported increased productivity, AI-enabled service expansion, and improved strategic positioning

Congregation: Building AI capability through applied workforce learning

Organisation type/scale: Learning and transformation consultancy (creative industries focus)

Main challenge: Fragmented AI use and limited behaviour change in client organisations

Main intervention: Four-phase applied learning model (diagnosis, leadership alignment, role-specific learning, embedding)

Nature of the journey: From scattered experimentation to structured organisational practice

Observed impact (evidence from practice): Client organisations reported improvements in workflow efficiency, reduced friction, more consistent AI use, and increased confidence among staff and leaders

Key outcomes/value: Stronger alignment between AI use and business objectives, and more structured and sustainable learning practices

Cast Consultancy: Building AI capability through use-case-led upskilling in construction consultancy

Organisation type/scale: SME UK construction consultancy

Main challenge: Adopting AI in a low digital-maturity sector while maintaining quality and accountability

Main intervention: Use-case-led training focused on Gateway 2 submissions, combining tool-based learning with provider-led sessions and ongoing practice

Nature of the journey: Incremental shift from awareness to applied use within a specific workflow, with early-stage progression towards broader capability

Observed impact (evidence from practice): Increased confidence in using and validating AI outputs, improved efficiency in complex documentation workflows, ongoing feedback loop with provider

Key outcomes/value: Strengthened responsible AI use and workflow efficiency, with a clear foundation for wider adoption

LinkedIn Learning: Supporting organisational AI upskilling through structured learning frameworks

Organisation type/scale: Learning platform/workforce learning provider

Main challenge: Moving organisations from informal experimentation to structured AI capability development

Main intervention: AI capability framework with defined levels, curated pathways, and personalised recommendations linked to roles and career signals

Nature of the journey: From awareness to applied learning and longer-term capability development aligned with workforce planning

Observed impact (evidence from practice): Platform insights and organisational examples suggest increased participation in AI learning, application of AI in routine tasks, and improved ability to track skills progression and gaps

Key outcomes/value: Improved AI fluency among non-technical staff and supports structured workforce capability development

100 School: Building AI capability through challenge-based learning

Organisation type/scale: AI education provider/enterprise training partner

Main challenge: Lack of sustained AI adoption despite access to tools and training

Main intervention: Challenge-based programme with daily tasks, peer learning, and leadership participation

Nature of the journey: From awareness to habit formation and capability activation

Observed impact (evidence from practice): In one engagement 88% activation and over 1,400 real work outputs were generated, significant increases in confidence and identification of AI champions, similar patterns observed in other client cases

Key outcomes/value: Strong behaviour change, high engagement, and scalable capability-building model

NHS Primary Care: AI upskilling through practice-level adoption

Organisation type/scale: Public healthcare (GP practice context)

Main challenge: High administrative burden and limited structured AI training under regulatory constraints

Main intervention: AI tools (for example clinical scribe) supported by provider-led training and ongoing updates

Nature of the journey: Incremental, practice-led adoption through local experimentation and peer learning

Observed impact (evidence from practice): Evidence from one practice and local networks suggests reduced documentation time, improved patient interaction, and more consistent clinical records

Key outcomes/value: Improved efficiency and quality of patient interaction, with safe and responsible AI use emerging locally

Good Things Foundation: Addressing the AI literacy gap through inclusive digital learning

Organisation type/scale: Digital inclusion charity/community learning provider

Main challenge: Addressing AI literacy gaps among digitally excluded individuals

Main intervention: Open-access AI Gateway using plain English, bite-sized modules, and supported community delivery

Nature of the journey: Entry-level journey focused on confidence-building and safe experimentation

Observed impact (evidence from practice): Early qualitative evidence shows increased confidence among learners and staff (with some learners beginning to use AI independently), supports progression towards further digital and AI learning

Key outcomes/value: Supports inclusion, confidence-building, and foundational engagement with AI

2.1 Large enterprise, regulated environments

2.1.1 Airbus: Building AI capability through long-term workforce transformation

Key lessons

1: Communities of practice are essential for scaling AI capability

Social learning, peer exchange, and visibility of real use cases helped make AI accessible and reduced barriers to engagement across the organisation.

2: Start with business problems, not technology

Framing AI initiatives around concrete operational challenges ensured relevance and avoided unfocused experimentation.

3: Strong data governance underpins responsible and sustainable AI use

Clear competencies, mandatory training, and shared responsibility supported trust, value creation, and long-term adoption.

Challenge

Airbus faced a long-term and evolving challenge in developing AI capability across a large, complex, and safety-critical organisation operating across aeronautics, defence, space, and helicopters. While AI had been embedded in Airbus products for over a decade, its use was historically concentrated within a small number of highly specialised technical roles, primarily focused on aircraft systems and performance.

As digital technologies matured and data volumes increased, Airbus recognised that AI could create value far beyond the physical product. Large and complex datasets existed across supply chains, manufacturing, procurement, delivery scheduling, and operational planning. However, these datasets were often embedded within legacy systems and could only be meaningfully interpreted by people with deep organisational and domain knowledge. This made external sourcing of AI expertise insufficient and highlighted the need to build capability internally.

Scale and workforce diversity added further complexity. Airbus operates across multiple divisions, countries and functions, with employees working in highly technical, operational, managerial, and support roles. The challenge was not only to develop advanced AI specialists, but also to ensure that non-technical employees and leaders understood the value of AI, how it could be applied responsibly, and how it related to their day-to-day work.

Airbus therefore faced a dual challenge:

• building advanced AI and data science capability at scale
• ensuring AI skills were accessible, inclusive, and relevant across a diverse global workforce

Intervention

Airbus adopted a long-term, capability-led approach to AI upskilling, rooted in earlier digital transformation efforts and progressively expanded over more than a decade. Rather than introducing AI training as a standalone initiative, AI capability development was embedded within workforce planning, competency frameworks, and organisational learning systems.

Key interventions included:

1: Early investment in data analytics as a foundation for AI

Airbus began by strengthening data analytics and data governance capabilities across the organisation. This created a structured progression from data analytics to data science and, ultimately, AI. Once supported, employees transitioned into new roles over time, rather than relying primarily on external recruitment.

2: Structured learning pathways for AI and data roles

Formal learning pathways were introduced for data analysts and data scientists, enabling progression from beginner to advanced levels. Around 2,000 employees, approximately 3% of the total Airbus workforce, completed these pathways, forming the initial core of Airbus’s internal AI and analytics capability. These cohorts later became anchors, as mentors for wider learning and knowledge-sharing activities.

3: Communities of practice and social learning

Airbus established internal AI and analytics communities to enable practitioners to share projects, troubleshoot challenges, and exchange best practice. Regular showcases, mentoring relationships, and informal discussions helped surface real use cases across the organisation. An internal podcast series, AI for Newbies, helped demystify by explaining AI concepts in short, accessible formats.

These early cohorts became visible internal experts who could demonstrate real business value, share knowledge, and legitimise AI as a practical organisational capability rather than a niche technical activity.

4: Formal recognition through competencies

AI-related competencies were embedded within Airbus’s job and competency framework. New competencies, including Generative AI Essentials, could be added to any role, regardless of function. Competency assessments helped identify skills gaps and link them directly to learning and development needs.

This learning architecture was embedded within Airbus’s annual workforce and competency planning cycle. Each year, the Airbus Digital Academy produced an updated digital and data competence strategy identifying priority capabilities for the business. These were used by HR business partners and line managers to inform recruitment, redeployment, and development decisions, ensuring that AI and data skills were aligned with evolving business needs rather than treated as optional training.

5: Multimodal learning design

Learning combined formal courses, informal content, community engagement, and real-world application. Employees could choose learning formats that suited their role, confidence level, and learning preferences. Data governance, ethical use, and sustainability considerations were integrated throughout all pathways.

Journey

Airbus’s AI upskilling journey was incremental and adaptive, evolving alongside technological developments and organisational maturity.

The journey unfolded in stages:

1: From digital foundations to AI capability

Early efforts focused on automation, data analytics, and digital transformation. As data maturity increased, learning pathways expanded into data science and AI.

2: From specialist expertise to wider workforce engagement

Initially, AI capability was concentrated among technical specialists. Communities of practice and internal showcases made AI work visible and accessible to non-technical staff. Over time, AI became a shared organisational capability rather than a niche technical domain.

3: Personalised pathways based on role and need

Two core pathways were established: essentials pathways for employees and managers requiring baseline AI understanding, and practitioner and specialist pathways for advanced technical and analytical roles.

Employees could add AI competencies alongside their existing roles without changing job titles. Employees progressed through these pathways in different ways. Some applied for new roles as data analysts or data scientists after completing training. Others gradually transformed their existing roles to become more data-driven without formally changing job titles. In many cases, individuals began applying new skills mid-pathway, reshaping how their current jobs were performed rather than waiting for a formal transition.

4: Embedding responsibility and governance

Mandatory training was introduced before employees could access generative AI tools. Strong data governance ensured employees understood what data could and could not be shared with AI systems. Responsible and sustainable AI use was treated as a foundational capability.

Mandatory training ensured that experimentation was grounded in data protection, ethical use, and organisational accountability rather than unmanaged adoption.

5: Leadership engagement and cultural reinforcement

Annual AI and analytics events celebrated achievements and recognised innovation. Senior leaders participated in these events, reinforcing the strategic importance of AI capability. Over time, AI engagement became self-sustaining, requiring minimal central coordination.

Outcome

Airbus’s long-term approach to AI upskilling resulted in strong and sustained engagement across the organisation. AI capability became embedded not only within technical teams, but also among managers and non-technical employees, supporting more informed decision-making and greater confidence in the use of AI tools.

Uptake of learning pathways, community participation, and mandatory training remained high, with no evidence of drop-off following initial deployment. The visibility of AI use cases and continued peer exchange helped maintain momentum and relevance.

Importantly, Airbus was able to move from isolated experimentation to structured, organisation-wide capability development, supporting both operational efficiency and strategic transformation.

AI upskilling impact

Airbus’s approach to AI upskilling aligns with the AI skills framework set out in AI Skills for the UK Workforce report, spanning technical, non-technical, and responsible and ethical AI skills. Specialist pathways supported advanced technical capability, while organisation-wide training focused on non-technical skills. Responsible AI skills were embedded too.

Taken together, these features position Airbus at the strategy stage (stage 8) of the AI skills adoption pathway, where AI use is aligned with organisational goals. With emerging evidence of progression towards the scaling stage, through organisation-wide adoption and measurable productivity impact.

Airbus’s experience shows that effective AI upskilling operates across multiple levels:

• individual - increased confidence, capability, and understanding of how AI can support everyday work
• team - improved collaboration between technical specialists and business functions
• organisation - integration of AI into operational planning, workforce development, and strategic decision-making

A further important impact was increased internal labour mobility. As employees acquired AI and data skills, many were able to move between roles and functions within Airbus, creating an internal job market that retained talent while spreading digital capability across the organisation.

A recent pilot of generative AI tools provides a clear illustration of measurable impact. During a trial involving approximately 2,000 employees, Airbus found that the average user saved up to 4 hours per week when using an AI assistant effectively. These time savings were observed primarily in office-based, non-production roles and reflected improvements in routine tasks, analysis, and administrative activities.

At scale, this reduction in time spent on lower-value tasks frees capacity for higher-value and more creative work, while also reinforcing the importance of appropriate training and responsible use. Crucially, these gains were realised alongside strong data governance and mandatory learning requirements, supporting trust and sustainability rather than unmanaged adoption.

Beyond productivity, Airbus also reported wider benefits, including:

• improved confidence among non-technical employees
• stronger engagement with AI across career stages
• better problem framing and decision quality
• increased awareness of ethical, data, and sustainability considerations

Airbus also emphasised that while productivity gains are measurable, the full value of AI upskilling lies in building long-term, transferable workforce capability that is difficult to capture in simple financial metrics, a challenge that is particularly significant for smaller firms.

2.1.2 KPMG: Embedding AI upskilling as an organisational capability

Key lessons

1: AI upskilling must be treated as an organisational capability, not a course

Embedding learning within strategy, governance, and business change is essential for sustained impact.

2: Application drives adoption and value

AI training works when it is tightly linked to real tasks, use cases, and productivity outcomes, not abstract knowledge alone. Champions are essential to providing bottom-up specific use cases and support in the flow of work.

3: Leadership sponsorship and democratised access matter

Clear permission to learn, easy access to tools, clear guidelines and inclusive design enable engagement across roles, grades, and career stages (reassuring staff who may fear using AI).

Challenge

KPMG UK faced a multifaceted challenge in building AI capability across a large professional services organisation. The rapid emergence of generative AI tools from late 2022 onwards meant that employees were encountering generative AI externally before clear organisational positions, governance arrangements, or training pathways were fully in place.

This created immediate pressures around safe use, regulatory compliance, and professional judgement, particularly given KPMG’s obligations to regulators and clients. Any AI adoption needed to maintain trust, credibility, and ethical standards, rather than prioritising speed alone.

Scale added further complexity. With approximately 17,000 employees in the UK, operating across audit, tax, advisory, and corporate services, KPMG needed to support AI capability development across very different roles, starting points, and career stages. The challenge was not limited to developing technical specialists, but to ensuring broad AI literacy, responsible use and confidence while enabling deeper expertise where required.

A further challenge related to careers and progression. AI skills were rapidly shifting from ‘nice to have’ to essential, particularly in advisory roles. Employees increasingly saw AI capability as shaping future career pathways, yet there was limited clarity in the wider market around qualifications, progression routes, or how to sustain learning in such a fast-moving domain.

KPMG therefore faced a dual challenge:

• enabling responsible, scalable AI adoption internally
• developing an approach that could credibly inform and support client organisations facing similar AI skills pressures

Intervention

KPMG deliberately avoided treating AI upskilling as a single course or isolated learning initiative. Instead, it embedded AI learning within a broader organisational transformation programme, underpinned by a ‘client zero’ philosophy to use the AI services and solutions provided to clients.

Key interventions included:

1: Acting as client zero for AI adoption

KPMG tested AI tools, training approaches, and governance internally before taking them to market. This allowed the firm to learn from real implementation challenges, adoption barriers, and workforce responses before advising clients. This approach strengthened credibility and ensured that training was grounded in real operational use rather than abstract theory.

2: Embedding AI learning within a global AI transformation programme

AI upskilling sat alongside service redesign, technology deployment, and governance, rather than being owned solely by L&D. Learning was explicitly linked to business change, adoption, and value creation. The AI learning offering evolved to support innovations in the field and to enable greater personalisation of learning by level of literacy, context, role and task.

3: The ‘Summer of AI’ series (from 2023 onwards)

A leadership-sponsored, firm-wide programme that created permission to learn, experiment, and ask questions about AI. Combined internal expertise, external partners (such as Microsoft), and industry speakers. Set a consistent tone each year around emerging AI developments and future priorities. Signalled from the top that AI capability was a strategic investment in people, not an optional add-on.

4: AI learning embedded in everyday workflows through Spark

Spark, a KPMG-developed AI-assisted learning coach, was embedded directly within Microsoft Teams. It enabled employees to ask questions about learning needs and receive contextual, conversational guidance and tailored recommendations from multiple learning libraries. It also lowered barriers to access by removing the need to search and navigate separate learning platforms, and supported ‘learning in the flow of work’, rather than separating training from daily practice.

5: Employee-led AI communities

Grassroots AI communities emerged organically and were supported rather than controlled. Regular ‘tea and talk’ sessions enabled peer learning, demonstrations, and discussion of emerging trends. These communities helped normalise AI use and reinforced curiosity and experimentation.

6: AI for business value apprenticeship

Over 130 employees across grades, including senior leaders, enrolled in an AI-focused apprenticeship. The programme explicitly linked AI learning to applied use cases, productivity improvements, and service innovation. Impact was assessed through real business outcomes rather than course completion alone. For instance, one learner developed a tool for a project that reduced analysis effort by 24 days (from 200 to 4 hours).

Supporting future generations

In addition to developing its own people, through its ‘Opening Doors to Opportunities’ social mobility programme, KPMG UK is proactively supporting future generations to develop essential AI skills and understanding well before they enter the workforce.

KPMG’s strategic goal is to develop the skills of one million young people by 2030, and a key part of this involves providing early opportunities for students to engage with AI and understand its implications in the professional world.

KPMG stated that since 2023, it has hosted over 16,000 Year 5 to Year 11 students for a half-day ‘world of work experiences’ in its offices across the UK. Students develop their AI skills and awareness through real-life KPMG consultancy client challenges affecting sectors ranging from healthcare to construction, technology, and sustainability.

This early exposure to AI helps to build confidence in young people and fosters a lifelong learning mindset, preparing them for the evolving demands of the future workforce. After taking part, 83% of students were more aware of how to apply their skills and talents to develop their ambitions, and 81% felt confident about finding a good job.

Journey

KPMG’s AI upskilling journey was deliberately iterative and adaptive, reflecting the pace of AI development and the diversity of the workforce.

The journey unfolded in stages:

1: Building foundations and trust

Early focus on AI literacy, ethical use, governance, and responsible decision-making. Addressed uncertainty and risk concerns in a regulated professional services environment. Helped employees understand not only how AI works, but when and when not to use it, including clear direction on the context in which it should not be used.

2: Shifting from awareness to application

Training moved beyond generic AI introductions to role-specific and task-based use cases. Employees explored how AI could augment everyday activities such as document review, analysis, drafting, and insight generation. Central teams supported this shift through tools, learning infrastructure and leadership-sponsored programmes, while local teams and employee communities used workshops, hackathons and applied use cases to test AI in everyday work.

3: Differentiated pathways rather than one-size-fits-all training

All employees had access to baseline AI tools and learning. Deeper pathways emerged for those whose roles or interests required advanced capability. This supported engagement across career stages without forcing uniform depth.

4: Embedding AI learning across career stages

Early-career employees begin to develop the behaviours and skills needed to maximise AI use, such as critical thinking, while also being given access to AI tools from week one, alongside guidance on responsible and ethical use. More experienced staff were supported to reskill and adapt as AI reshaped professional work. Continuous access to tools like Spark enabled ongoing, self-directed progression.

5: Continuous feedback and refinement

Learning coaches, apprenticeship providers, and internal teams captured emerging use cases and productivity gains. These insights fed back into programme design, ensuring learning remained relevant and up to date.

Outcome

KPMG’s approach resulted in strong engagement with AI learning and tools across the organisation. Uptake was supported by visible leadership sponsorship, ease of access, and clear relevance to everyday work.

Crucially, AI upskilling translated into tangible business outcomes. Early evidence from the AI for business value apprenticeship and KPMG’s internal initiatives demonstrated significant productivity improvements, including dramatic reductions in time required for complex analytical tasks in certain use cases. This demonstrates how AI is steadily augmenting every aspect of how KPMG delivers work, moving from isolated pilots to deeply embedding intelligent tooling into core processes.

Through live client projects, KPMG has shown how AI can radically accelerate analysis, enhance judgement, and reshape workflows, signalling a shift toward an AI‑enabled professional environment that is trusted by design. It reflects KPMG’s broader ambition to normalise AI across both personal and professional contexts, equipping people with the confidence and capability to work in fundamentally new ways.

Beyond productivity, the organisation identified additional benefits, including:

• development of new services and client offerings
• improved employee confidence and sense of control
• reduced cognitive overload
• inclusion benefits, particularly for neurodivergent colleagues and those working in a second language

KPMG was also able to translate internal learning directly into client support, strengthening its ability to guide organisations through AI adoption, skills development, and change management. KPMG has also supported clients across a range of sectors to embed and adopt Spark to offer hyper-personalised learning for their workforce to build capability across their organisations. KPMG has used its own launch of Copilot to deliver upskilling on proof of concepts in insurance, banking, higher education, FMCGs and infrastructure.

AI upskilling impact

KPMG’s approach to AI upskilling aligns with the AI skills framework set out in the AI Skills for the UK Workforce report, spanning technical, non-technical, and responsible and ethical AI skills. Technical AI skills were developed for a smaller group of specialists and advanced users, while organisation-wide training prioritised non-technical capabilities. Responsible and ethical AI skills were embedded from the outset, reflecting KPMG’s regulatory context and professional obligations. Taken together, these features position KPMG at the integration stage (stage 7) of the AI skills adoption pathway, where AI tools and learning are embedded into everyday workflows, with clear momentum towards more strategic alignment for the strategy stage.

KPMG’s experience shows that effective AI upskilling operates at multiple levels:

• individual - increased confidence, curiosity, and capability to use AI responsibly
• team - shared experimentation, peer learning, and use-case development
• organisation - integration of AI into processes, services, and strategic decision-making

Importantly, the most critical skills extended beyond technical expertise. Non-technical capabilities such as critical thinking, adaptability, leadership, and ethical judgement were central to successful AI adoption. Responsible AI skills were treated as a core professional competence rather than a specialist concern.

2.1.3 Roche: Building enterprise-wide AI capability through responsible upskilling

Key lessons

1: Make AI training mandatory to establish shared responsibility

Mandatory training created a common baseline, reduced uneven adoption, and signalled that responsible AI use is a core organisational expectation.

2: Focus on applicability, not just knowledge

Training was most effective when employees could immediately apply learning using approved tools in their own workflows.

3: Design for accessibility and diverse learning styles

A mix of delivery modes, languages, and peer-led learning ensured engagement across a large and diverse workforce.

4: Treat AI upskilling as a continuous journey

Communities of practice and champions were essential for sustaining momentum beyond the initial rollout and adapting to rapid technological change.

Challenge

Roche is a multinational pharmaceutical company headquartered in Basel, Switzerland, operating across the end-to-end drug development lifecycle (from pre-clinical discovery through to early and late development, manufacturing, supply, and patient access). Roche reported operating in around 120 countries and employing over 100,000 people globally. In the UK, Roche’s workforce was estimated close to 3,000 employees in 2026.

Roche works across 5 therapeutic areas (including oncology, immunology, ophthalmology, cardiovascular, and neurosciences). Given the growing strategic importance of AI in the UK, Roche identified key AI capability areas at the corporate level to be applied through the drug development lifecycle, incorporating potential use cases as outlined below.

Roche identified several core business objectives where AI could create value when used at scale. Roche described 4 key targets where AI could accelerate impact:

• speed (delivering medicines faster to patients)
• reducing the cost of drug development
• increasing probability of technical success (PTS) in the development pipeline, by leveraging AI and machine learning to enhance data analysis and predictions
• increasing the number of patients who could benefit from a medicine by identifying new therapeutic or regulatory pathways

Realising these benefits required widespread AI literacy, not just isolated technical excellence.

Several interrelated challenges prompted the need for organisation-wide AI upskilling:

1: AI was often perceived as the domain of data scientists or technical specialists rather than a capability relevant to the wider workforce

Many non-technical employees viewed AI as abstract, opaque, or detached from day-to-day roles, limiting adoption outside specialist teams.

2: The pace of AI development far exceeded the speed at which formal education and professional training systems had adapted

Most employees had not encountered AI during education for different career pathways, creating a widespread skills gap across job levels and functions.

The integration of AI also presents a dynamic challenge. As employees increasingly engage with widely accessible large language models, comprehensive policies and training are needed to ensure appropriate use and mitigate risks related to data security and compliance. This applies across many different types of industry. Taken together, Roche faced a clear challenge: how to rapidly upskill a large, diverse workforce (across 120 countries and over 100,000 employees) in AI in a way that was accessible, responsible, and aligned with both regulatory constraints and strategic business goals.

Intervention

To address these challenges, Roche designed and implemented a corporate-level AI training programme delivered to the entire global workforce, including UK-based employees. The programme was positioned as a foundational capability intervention rather than a niche or optional learning initiative.

The training programme was launched and completed in the second half of 2025. It was mandatory for all employees, reaching over 100,000 people within a short period of time.

1: Organisation-wide, mandatory training

A central design decision was to make AI training mandatory rather than voluntary. This signalled that AI capability was a core organisational requirement, comparable to compliance or safety training, rather than an optional enhancement for interested individuals. Different learning styles were deployed to meet different needs. These ranged from reading materials and virtual classroom training to demonstrations, chats and other formats. Roche reported positive feedback and strong engagement. The AI champion also came directly from the function and therefore understood the department’s work and culture.

Training was rolled out in phases across departments rather than by job level. Employees in different divisions and functions began the training at different points, but all were required to complete the full programme within the rollout window. This approach enabled large-scale delivery while accommodating operational constraints.

Each employee had access to a personalised digital dashboard showing completed and outstanding modules, providing transparency and accountability at individual level.

2: Broad and balanced curriculum

The training curriculum combined technical, non-technical, and responsible AI components. Topics included:

• practical use of AI tools, such as prompting techniques and interaction with large language models
• guidance on using organisation-approved tools, including internal Notebook LM environments and enterprise AI-enabled products
• ethical considerations, including bias, limitations of AI-generated outputs, and the importance of critical evaluation
• clear articulation of a human-in-the-loop model, emphasising that humans remain accountable for decisions and outputs

The curriculum was designed to be accessible to non-technical employees while still providing enough substance to support confident and informed use of AI in daily work.

3: Hybrid and flexible delivery

Training was delivered using a hybrid model. Some modules were delivered through live virtual sessions, allowing employees to engage synchronously with instructors and peers. Others were asynchronous, including short readings, recorded materials, and practical exercises.

Sessions typically lasted around 30 to 45 minutes, enabling employees to integrate learning alongside existing workloads.

Importantly, the programme allowed flexibility in sequencing. Employees could complete modules in different orders depending on availability, rather than following a rigid linear pathway. This flexibility proved critical for a global workforce operating across time zones, travel schedules, and variable operational pressures.

4: Integrated access to AI tools

Alongside training, Roche aligned upskilling with access to approved AI tools. Roche described Gemini as being integrated into its existing Google productivity environment (for example, email and document workflows), with enterprise AI access available to employees at scale.

This alignment between training and tool availability supported immediate application, reinforcing learning and reducing the risk that training would remain abstract or disconnected from practice.

Journey

Roche’s AI upskilling journey reflects a shift from specialist-led experimentation to enterprise-wide capability building.

1: From specialist perception to shared capability

Before the intervention, AI was often viewed as a specialised or technical function. Following the training, employees across functions began to see AI as a practical tool that could support everyday tasks, from document summarisation and content creation to planning and analysis.

This shift was particularly significant for non-technical staff, who reported increased confidence in experimenting with AI tools without needing advanced technical expertise.

2: Embedding social learning through champions

While the corporate training programme established a common baseline, Roche recognised that continuous learning would be essential given the pace of AI development.

Within individual functions, AI champions were appointed to sustain momentum beyond the initial rollout. These champions organised regular, voluntary knowledge-sharing sessions, branded as part of an ‘Everyday AI’ initiative.

One function (Regulatory) organised regular sessions once every 2 weeks and scheduled at different times (for example, one session timed for Europe and one for North America). Participation was voluntary, yet attendance was consistently high, with around 150 employees joining each session. Peers showcased real examples of how colleagues were using AI in their work, enabling peer-to-peer learning grounded in lived experience rather than abstract instruction.

This community-based approach helped normalise experimentation, surface practical use cases, and sustain engagement without requiring constant central coordination.

3: Accessibility and inclusion

Accessibility was a central design principle throughout the journey. Training content was delivered by country outside of headquarters and so could utilise local languages.

Different learning styles were supported through a mix of reading, live interaction, practical exercises, and informal discussion.

Small-group sessions enabled employees to ask questions in psychologically safe settings. Roche described small cohorts of around 10 people engaging with champions in chat-based formats to ask questions in a safe environment, with groups randomised to reduce hierarchy effects.

Employees could also revisit completed modules at any time, supporting reinforcement and self-paced learning as confidence and needs evolved.

4: Open questions and future development

Impact will be assessed function-by-function in 2026. In 2025, the organisation’s initial goal was for employees to save one hour per person per week through the use of AI. In 2026, each function will define its own impact measures, reflecting the specific ways AI is being used in its work. Employees can also repeat the online training at any time as a refresher. Plans for the next phase of AI capability development are currently being discussed.

While the initial rollout was completed successfully, Roche recognises that AI upskilling is an ongoing process rather than a one-off intervention. Open questions remain around onboarding new joiners, updating content in response to regulatory change, and deepening training around advanced topics such as fine-tuning, testing, and data governance.

These questions are now part of the organisation’s next phase of AI capability development.

Outcome

The AI training programme delivered several clear outcomes at individual, team, and organisational levels.

1: Cultural shift and confidence

One of the most significant outcomes was a cultural shift in how AI was perceived. Employees moved from seeing AI as distant or threatening to viewing it as a supportive tool embedded within their work. Fear of job displacement reduced as employees gained hands-on experience and a more realistic understanding of AI’s capabilities and limitations.

Employees increasingly spoke openly about their use of AI, sharing examples of impact and learning from one another.

2: Measurable productivity gains

At a corporate level, Roche set a clear productivity ambition for the AI training programme: each employee should save at least one hour per week through effective use of AI.

For example, in one function, Regulatory, this target was exceeded. Around 50% of employees saved 4 hours or more per week through the use of AI tools, around 25% saved approximately one hour per week, and the remaining around 25% saved less than one hour per week.

Prior to the training, productivity savings were negligible. Gains were primarily associated with routine and knowledge-intensive tasks, including document summarisation, drafting text, preparing presentations, and synthesising information.

3: Responsible and governed use

The programme reinforced clear guardrails around data use, reflecting the sensitivity of patient and clinical information in the pharmaceutical context.

Roche described strict controls to prevent sensitive data being processed outside the corporate firewall. Roche reported that while some internal information categories could be processed (referred to as C1 to C4), GDPR-related personal data could not. Roche also noted that data is not always neatly organised into GDPR and non-GDPR categories, creating an ongoing operational challenge that requires continued reinforcement through training and governance.

Roche also highlighted the need to strengthen workforce understanding of more advanced concepts such as fine-tuning and testing, and why these processes cannot be conducted using certain patient-level datasets.

Overall, the training reduced the risk of inappropriate or unsafe AI use by raising awareness and shared responsibility, while also surfacing areas for deeper capability building.

4: Position in the AI adoption journey

In terms of the AI adoption pathway, Roche has progressed strongly through strategy and integration stages and is actively working on scaling (stage 9). While AI solutions are increasingly embedded in specific programmes and functions, enterprise-wide scaling remains complex due to organisational size, regulatory constraints, and data sensitivity.

Nevertheless, the breadth and speed of the training intervention represent a substantial step towards sustainable scaling grounded in capability rather than isolated tools.

2.2 SMEs, creative industries, construction and consultancies

2.2.1 Vertis Media: Building in-house AI capability in a creative SME

Key lessons

1: Start early and build capability internally

Early engagement with AI enabled Vertis Media to adapt proactively rather than reactively to industry disruption.

2: Create shared learning structures

Central dashboards, regular sharing, and collective reflection prevented fragmentation and supported continuous learning.

3: Protect time for learning and experimentation

Paying employees for learning time and embedding AI discussion into normal work routines were critical to sustained engagement.

Together, these lessons demonstrate that even small creative enterprises can build effective, responsible AI capability when learning is treated as an ongoing, organisation-wide journey.

Background

Vertis Media is a UK-based marketing and advertising agency founded in 2015. The company delivers digital marketing strategy, advertising campaigns, and training services for a diverse client base, ranging from global brands such as IKEA and Vodafone to start-ups and scale-ups. While historically focused on consumer-facing sectors, Vertis Media has increasingly pivoted towards business to business (B2B) marketing, working with highly specialised industries including legal, finance, and oil and gas.

Vertis Media operates as a small and flexible organisation, with a core team of around 6 employees supported by a network of freelancers as required. This structure enables rapid adaptation to market change, but also places pressure on internal capability, as individuals are required to perform multiple roles. AI skills development at Vertis Media has therefore been designed and delivered primarily as an in-house initiative, tightly aligned with the company’s evolving business model and service offer.

Vertis Media began its AI upskilling journey in 2023, recognising early that generative AI would significantly disrupt marketing, advertising, and campaign management activities, and that proactive capability-building was essential for business sustainability.

Challenge

Vertis Media identified several interrelated challenges that prompted the introduction of structured AI upskilling within the organisation.

First, the company recognised that AI was fundamentally reshaping the marketing and advertising industry. Activities that previously required multiple specialists, such as campaign management, reporting, and analysis, could increasingly be completed by fewer people using AI-enabled tools. Without timely upskilling, this shift risked eroding competitiveness and client value.

In practice, Vertis Media reported that AI adoption did not lead to staff reductions. Instead, it changed the type of roles required within the organisation. As AI reduced time spent on execution-focused tasks, the company began shifting towards higher-value roles centred on business development, client relationships, and strategic services. This evolution reflects a broader lesson for SMEs: the risk is not AI replacing jobs, but failing to adapt business models and skill requirements as technology evolves.

Second, the rapid proliferation of AI tools created both opportunity and complexity. While AI opened up new possibilities for faster research, deeper analysis, and more creative output, it also introduced uncertainty around tool selection, appropriate use, and skill requirements. For a small organisation, experimenting without structure risked wasted effort and overload.

Third, Vertis Media faced a strategic challenge linked to business growth and diversification. AI-enabled capability made it possible to apply existing marketing expertise to new, highly specialised B2B sectors that had previously been inaccessible. However, doing so required new workflows, enhanced data handling, and confidence in applying AI responsibly and effectively.

The core challenge was therefore to integrate AI into everyday work in a way that improved productivity, enabled new services, and supported staff capability, without overwhelming a small team or relying on external hiring.

Intervention

Vertis Media adopted a phased, predominantly in-house approach to AI upskilling, evolving over time as organisational learning matured.

1: Early self-learning and external exposure

The initial phase focused on self-directed learning by the founders, who explored emerging AI tools and platforms independently. This was complemented by engagement with an external provider offering an early collaborative AI platform and training resources, which helped establish a foundational understanding of language models and team-based AI use.

2: Development of internal training and resources

Building on this foundation, Vertis Media created its own internal AI training materials tailored to the specific needs of the business. These included short videos, guidance documents, and reference materials hosted on an internal platform, enabling staff to access relevant learning on demand.

3: Centralised learning infrastructure

A key intervention was the creation of a central internal dashboard that brought together AI tools, learning resources, updates, and experiments. This acted as a shared knowledge hub, reducing duplication and ensuring that learning was collective rather than siloed.

4: Ongoing sharing and operational integration

Regular weekly operational meetings were used to share new learning, tools, and applications, embedding AI discussions into normal business rhythms. Automation systems were also introduced to filter and summarise relevant external developments, helping the team stay informed without information overload.

Over time, Vertis Media shifted from broad experimentation to consolidation, focusing on refining internal processes, tools, and workflows that could be used consistently across the organisation and eventually offered to clients as part of new service propositions.

Journey

Vertis Media’s AI upskilling journey has been continuous since 2023 and reflects the realities of learning within a small creative enterprise.

1: From exploration to focus

Initial experimentation generated enthusiasm but also highlighted the risk of fragmentation. A deliberate pause enabled the team to assess what was working and to prioritise tools and processes with clear business value.

2: Role-sensitive learning pathways

While all staff developed foundational AI awareness, learning pathways diverged based on role. Creative specialists undertook deeper training in AI-enabled image, video, and content creation tools, while other team members focused on workflow optimisation, analysis, and client-facing applications.

3: Embedding AI into everyday systems

AI was progressively integrated into project management, reporting, and operational dashboards. This enabled rapid application of learning and allowed impact to be realised within months rather than years.

4: Strategic reflection and repositioning

As capability matured, Vertis Media entered a strategic reflection phase, using AI-enabled insight to support a broader repositioning towards niche business to business markets. At the time of writing, the organisation identifies itself as operating at the strategy stage of AI skills adoption, with implementation underway and scaling ambitions emerging.

Outcome

Vertis Media did not undertake a formal cost–benefit analysis of AI adoption. However, the financial impact became visible through operational efficiencies and service expansion. Time savings from faster research, reporting, and content production improved productivity, while new AI-enabled services contributed to an estimated 20% increase in turnover.

Vertis Media began to observe tangible outcomes within a few months of introducing structured AI upskilling.

Key outcomes included:

• significant reductions in time spent on repetitive tasks such as campaign reporting and analysis
• faster and more cost-effective research and insight generation
• the ability to develop and offer new services without proportional increases in headcount
• greater flexibility in responding to new client opportunities and sectors

While productivity gains were substantial, the organisation also recognised the importance of managing cognitive load. As AI increased the volume of information processed, conscious decisions were required to prioritise projects and avoid over-extension.

Responsible AI use is managed through ongoing internal discussion, shared team norms, and direct oversight from the founders. The organisation is deliberate about where AI tools are appropriate and regularly revisits these boundaries as technologies evolve. At the same time, Vertis Media recognises that AI now performs some tasks previously used for early career learning. To address this, staff are trained to review, critique, and refine AI outputs, ensuring they understand the reasoning behind results and can confidently challenge or improve them.

Overall, AI upskilling supported Vertis Media’s transition towards higher-value work, service innovation, and strategic resilience within a highly competitive creative market.

AI upskilling impact

Vertis Media’s in-house approach aligns with the AI skills framework outlined in AI Skills for the UK Workforce, particularly for SMEs:

• non-technical AI skills were developed through workflow redesign, problem framing, and decision-making
• technical literacy was built around practical use of AI tools relevant to marketing, analysis, and creative production
• responsible AI skills were addressed through internal discussion, shared norms, and careful consideration of where and how AI should be used

Impact is visible across levels:

• individual - increased confidence, faster task completion, and enhanced creative capability
• team - shared learning, consistent processes, and improved collaboration
• organisation - strategic repositioning, service diversification, and improved scalability

Vertis Media’s experience illustrates how small firms can move beyond awareness towards strategy-led AI adoption without extensive external investment.

2.2.2 Congregation: Building AI capability through applied workforce learning

Key lessons

1: AI training must be rooted in organisational context

Generic or tool‑led training fails to create impact unless it reflects how work is actually done within a specific organisation.

2: Behaviour change matters more than tool adoption

Sustained value comes from changing workflows, habits, and collaboration patterns, not from exposure to the latest technologies.

3: Leadership engagement and visible wins are critical

Clear sponsorship, protected learning time, and early ‘firework moments’ help build momentum, trust, and legitimacy for AI adoption.

Together, these lessons demonstrate that effective AI upskilling is less about technology itself and more about how organisations prepare their people to learn, adapt, and work differently in an AI‑enabled environment. Upskilling is not possible unless individuals understand why it is needed, how it will benefit them, how the skills will be used and how they will be supported.

Congregation is a UK-based learning and transformation consultancy specialising in applied skills development for organisations operating in the creative industries, including marketing, media, fast-moving consumer goods (FMCG), consumer packaged goods (CPG), and creative and in-house agencies. With over 2 decades of experience supporting organisations through digital and technological change, Congregation focuses on building practical capability within existing workforces rather than relying on external recruitment or short-term training interventions.

In recent years, Congregation has expanded its work to support organisations in developing AI skills and capabilities, delivering tailored AI training programmes to clients across the creative industries. These programmes are designed for employees, managers, and senior leaders, and focus on enabling effective, responsible, and role-relevant use of AI to enhance creative output, workflow efficiency, and decision-making. Rather than positioning AI as a purely technical capability, Congregation frames AI upskilling as a workforce and culture transformation challenge, grounded in the realities of creative and commercial work.

Challenge

Across the sectors Congregation serves, organisations were facing a rapidly intensifying challenge: while awareness of generative AI was high and initial experimentation had often taken place, tangible business impact remained limited.

Most client organisations had already invested in introductory or ‘AI 101’ training. However, this activity rarely translated into sustained behaviour change, consistent workflows, or measurable value. AI use tended to be fragmented, with isolated pockets of experimentation driven by a small number of enthusiastic individuals, while managers and leaders often felt under‑confident or disengaged. This created capability gaps between teams, uneven adoption, and uncertainty about how AI should be used responsibly in day‑to‑day work.

Fear and uncertainty further compounded these challenges. Employees expressed concern about job displacement, quality degradation, and reputational risk, while leaders struggled to distinguish between hype‑driven narratives and realistic opportunities for productivity, creativity, and innovation. In many organisations, the absence of clear principles, governance, and shared language resulted in reactive decision‑making rather than strategic capability development.

Congregation’s clients therefore faced a common set of challenges:

• limited translation of AI awareness into practical, role‑relevant capability
• fragmented experimentation without shared workflows or standards
• low confidence and high anxiety among both employees and managers
• lack of a coherent vision for where AI should, and should not, be used

The challenge was not simply to introduce new tools, but to build sustainable, organisation‑specific AI capability that aligned with business objectives, culture, and workforce realities.

Intervention

Congregation developed an applied, behaviour‑led approach to AI upskilling that treats AI as a catalyst for organisational learning and workflow transformation, rather than a standalone technical intervention. The model is deliberately flexible and sector‑agnostic, enabling it to be adapted to different organisational contexts while retaining a consistent underlying structure.

The intervention is typically delivered through 4 interlinked phases:

1: Mapping and diagnosis

Congregation begins by working closely with clients to map where AI fits within their value chain, workflows, and risk landscape. This includes assessing existing capability, leadership readiness, governance structures, and access to tools. Rather than assuming technological readiness, this phase surfaces organisational constraints, regulatory considerations, and cultural barriers that shape what is realistically achievable.

2: Leadership priming and direction‑setting

A small group of senior leaders and internal influencers participate in an intensive primer session. This establishes a shared understanding of the AI landscape, clarifies opportunities and risks, and supports the development of an initial AI ‘north star’ or manifesto. These principles articulate where AI will add value, where it will not be used, and how human judgement, creativity, and responsibility will be preserved.

3: Applied learning and role‑specific education

AI learning is then extended across the organisation through highly practical, hands‑on workshops. Employees engage directly with tools relevant to their roles, focusing on personal productivity, workflow optimisation, and quality improvement. Sessions are tailored to specific teams (for example creative, commercial, strategy), ensuring relevance and immediate applicability rather than generic instruction.

4: Optimisation and embedding

The final phase focuses on consolidating learning into repeatable workflows, playbooks, prompts, and shared practices. Congregation supports organisations to identify low‑value tasks for automation, protect high‑value human work, and establish routines for sharing learning, experimentation, and improvement. This phase emphasises cultural change, peer learning, and sustained momentum.

Across all phases, Congregation prioritises common language, internal champions, and alignment with organisational metrics, ensuring that AI learning is rooted in how the business actually operates.

Journey

Congregation’s AI capability‑building journey with clients is explicitly designed as an ongoing process rather than a one‑off intervention.

1: From experimentation to coherence

Early engagements often reveal scattered AI use and uneven confidence. Through mapping and leadership alignment, organisations move towards a clearer, shared direction that reduces duplication and uncertainty.

2: From individual curiosity to collective capability

Initial learning focuses on personal activation and confidence‑building. Over time, role‑based workshops and shared artefacts (such as prompts and workflows) transform individual experimentation into collective organisational practice.

3: From awareness to behaviour change

Rather than tracking success through tool adoption alone, Congregation focuses on observable changes in how work is done. Employees begin to remove unnecessary steps, improve briefing quality, shorten approval cycles, and collaborate more effectively across teams.

4: Embedding learning cultures

Sustained impact is supported through rituals such as regular showcases, peer‑led learning sessions, and leadership storytelling. These mechanisms help normalise AI use, reduce stigma or fear, and ensure learning evolves alongside technological change.

Congregation’s model recognises that not all employees need or want deep technical expertise. Learning pathways accommodate different levels of ambition, from baseline AI literacy to champion roles, without forcing uniform progression.

Outcome

Across client organisations, Congregation’s approach has led to consistent improvements in confidence, capability, and workflow effectiveness. Employees report reduced anxiety around AI and greater clarity about how it can support, rather than threaten, their roles.

Leadership at Congregation explained that organisations had observed tangible benefits including:

• shorter cycle times and reduced friction in content and asset production
• improved quality of briefs, decision‑making, and creative output
• increased consistency in AI use across teams
• stronger collaboration between technical and non‑technical roles

Congregation’s leadership team explained that leaders report greater visibility of AI activity and a clearer basis for strategic decision‑making. Rather than unmanaged experimentation, AI use becomes intentional, responsible, and aligned with business priorities.

While precise financial impacts vary by organisation and role, Congregation emphasises that the most significant outcomes lie in sustained behaviour change, improved learning capacity, and readiness to adapt as AI technologies continue to evolve.

AI upskilling impact

Congregation’s approach aligns closely with the AI skills framework presented in AI Skills for the UK Workforce, addressing non‑technical AI skills, technical literacy, and responsible AI skills.

Rather than positioning organisations at a fixed maturity level, Congregation supports progression from early experimentation towards structured, organisation‑wide adoption. Many client organisations move from ad‑hoc use towards the integration stage of AI skills adoption, with emerging evidence of progression towards scaling as practices become embedded.

Impact is evident across multiple levels:

• individual - increased confidence, reduced fear, and improved productivity
• team - shared workflows, better collaboration, and improved output quality
• organisation - clearer strategy, stronger governance, and sustainable learning cultures

2.2.3 Cast Consultancy: Building AI capability through tool-based upskilling in construction consultancy

Key lessons

1: Start with a real business use case

AI upskilling is most effective when linked to a clearly defined workflow that delivers tangible value. This ensures relevance and supports engagement.

2: Build training around human judgement

In professional services, AI outputs must be interpreted and validated by experts. Training should therefore emphasise judgement, accountability, and quality assurance.

3: Create a feedback loop with the AI provider

Establishing a continuous feedback loop between users and the AI provider enables ongoing improvement of both the tool and user capability. This transforms training into a dynamic, co-development process rather than a one-off activity.

4: Address behavioural and cultural barriers

Successful AI adoption requires addressing employee concerns and building trust. Framing AI as a tool that enhances professional work, rather than replacing it, is critical for engagement.

Challenge

Cast Consultancy faced a growing strategic challenge in preparing its workforce for AI-enabled change within the construction sector. As a professional services business, operating in the built environment, the firm’s work depends heavily on professional judgement, project expertise, commercial oversight, and client-facing advisory capability. Its workforce includes project managers, quantity surveyors, cost managers, and advisory specialists, most of whom operate in pre-construction and delivery management rather than on-site activities. The firm employs around 90 people, with offices in London and Edinburgh.

Although Cast Consultancy had begun to consider the implications of AI for its business model around 18 months to 2 years earlier, formal AI-related training only started within the last 12 months. Prior to this, AI was mainly encountered indirectly through enhanced reporting and management information systems rather than through deliberate workforce development. Leadership recognised that this would not be sufficient. The consultancy model in construction is expected to change significantly, and the firm sought to proactively identify where AI could create value while maintaining service quality and professional accountability.

The challenge was therefore not whether to adopt AI, but how to do so in a way that is practical, responsible, and aligned with business needs. In professional services, AI-generated outputs cannot be passed directly to clients without validation. Human judgement remains central, and the firm retains full responsibility for outcomes. This created a dual challenge:

• identifying a high-value use case for AI
• ensuring that staff are trained to use it effectively without compromising trust, quality, or accountability

A further challenge relates to the wider sector context. Construction has often lagged behind other industries in digital transformation and AI adoption. Cast Consultancy therefore needed to move forward in an environment with limited sector-wide examples while ensuring that any training delivered was grounded in real workflows rather than generic AI awareness.

Intervention

Cast Consultancy adopted a targeted, use-case-led approach to AI upskilling. Rather than beginning with general AI awareness training, the firm identified a specific, high-value workflow: supporting Gateway 2 submissions under the Building Safety Act.

Gateway 2 submissions are a specific regulatory requirement under the Building Safety Act 2022, mainly applying to higher-risk buildings (for example high-rise residential buildings). These submissions are complex, document-intensive, and time-sensitive, requiring the integration of drawings, specifications, and technical documentation.

The firm partnered with an external AI provider offering a proprietary tool designed to support this process. The tool enables faster identification of inconsistencies, missing information, and potential compliance issues before submission to the regulator. This created a clear and immediate use case for AI adoption.

Training was built directly around this tool. Approximately half of the business, particularly the project management function, was involved, with around 10 to 15 project managers actively engaging with both the tool and the associated training.

The training model was collaborative and practice-based. The AI provider delivered in-house sessions, allowing employees to test the tool, raise questions, and work through real examples. Rather than a one-off training session, this evolved into an ongoing learning process. Staff could experiment with the tool in live contexts and receive tailored guidance based on actual project needs.

The content focused primarily on non-technical and responsible AI skills. It covered how the tool processes data, how outputs should be interpreted, and where human validation is required. Particular emphasis was placed on responsible AI use, including confidentiality, GDPR compliance, and the handling of sensitive client data.

Journey

Cast Consultancy’s AI adoption journey has been incremental, pragmatic, and use-case driven. Rather than implementing a broad AI strategy from the outset, the firm began with a specific application and built capability around it.

In terms of the AI adoption pathway, Cast Consultancy sits between the reflection and upskilling stages. At the organisational level, the firm is still refining its broader AI strategy. However, for the Gateway 2 use case, it has clearly reached the upskilling stage, with structured training, active tool use, and integration into live workflows.

The journey can be understood across 4 stages.

1: From strategic awareness to applied experimentation

The firm first recognised that AI could affect its business model and began exploring practical applications. Rather than adopting AI symbolically, it focused on identifying a use case with clear operational value.

2: From tool selection to targeted upskilling

Following the identification of the Gateway 2 process, the firm introduced a specific AI tool and built training directly around it. Learning was embedded within practice, ensuring immediate relevance.

3: From one-way training to collaborative learning

Training evolved into a two-way process. Staff using the tool provided feedback based on real project experience, which informed improvements to the system. This created a continuous learning loop between the organisation and the provider.

4: From isolated use case to broader capability development

The Gateway 2 tool acted as an entry point for wider AI capability-building. While the firm is still developing a more comprehensive AI literacy strategy, this use case has established a foundation for broader adoption.

Outcome

The training resulted in increased confidence and capability among staff using the AI tool. Employees developed a clearer understanding of how the system processes information and how its outputs should be interpreted within professional workflows.

Staff also gained a broader understanding of AI, particularly how large language model-based systems analyse complex inputs such as drawings and technical documents. Importantly, they developed the ability to critically assess outputs and identify where human intervention is required.

A key outcome was the establishment of a strong feedback relationship with the AI provider. Staff insights, based on real-world use, contributed to ongoing improvements in the tool. This shifted the organisation’s role from passive user to active contributor in shaping AI functionality.

The training also addressed behavioural challenges. Some resistance emerged among employees concerned about the implications of AI for professional roles. Training therefore emphasised that AI is a support tool rather than a replacement for expertise, helping to reframe adoption as an opportunity to enhance rather than diminish professional value.

AI upskilling impact

Cast Consultancy’s approach aligns primarily with non-technical AI skills and responsible AI skills within the AI skills framework. The focus was on practical use, interpretation of outputs, and safe and compliant application of AI in professional contexts, rather than on technical development.

The case demonstrates that AI upskilling is most effective when directly linked to a specific business challenge. In this instance, the regulatory and operational demands of Gateway 2 submissions provided a clear and immediate driver for capability-building.

Positioning the organisation within the AI adoption pathway provides further insight. Cast Consultancy is at a transitional stage, moving from reflection to structured upskilling. The Gateway 2 use case represents a more advanced stage of adoption, serving as a foundation for broader organisational transformation.

The impact can be observed at multiple levels:

• individual - employees developed confidence in using AI tools and improved their ability to evaluate outputs critically
• team - shared understanding of AI-enabled workflows improved collaboration and consistency in tool use
• organisational - the firm has begun embedding AI into its operating model, recognising the need for ongoing capability development and structured adoption

2.3 Platforms, training providers, and activation models

2.3.1 LinkedIn Learning: Supporting organisational AI upskilling through structured learning frameworks

Key lessons

From LinkedIn’s experience designing and supporting AI learning at scale, 3 lessons stand out that may be relevant to other organisations undertaking similar work.

1: AI capability development requires structured learning pathways

Organisations benefit from frameworks that define clear levels of AI capability and provide structured learning progression.

2: AI training should be linked to real workplace applications

Learning initiatives are more effective when aligned with specific technology deployments or business use cases.

3: Workforce AI skills extend beyond technical expertise

Successful AI adoption requires a combination of technical, non-technical, and responsible AI capabilities across the workforce.

This example outlines how LinkedIn has approached workforce AI capability development within its learning ecosystem. It is intended to share practical design choices and lessons from implementation, rather than to promote specific products or services.

Challenge

Organisations across sectors are experiencing rapid changes in workforce skill requirements as artificial intelligence technologies become integrated into everyday work practices. Analysis from LinkedIn’s Economic Graph suggests that workforce skill requirements are changing rapidly: 38% of job skills changed globally between 2016 and 2023, and looking ahead to 2030 this figure could rise to as high as 70%, with generative AI a key driver of this shift.

For many organisations, AI adoption initially emerged through informal experimentation by individual employees or small teams. While this experimentation helped employees explore new tools, it often resulted in uneven capability development across organisations. Some teams quickly developed practical AI skills, while others lacked confidence or understanding of how AI could support their work.

Learning and development leaders therefore face several challenges when supporting workforce AI capability building:

• defining what AI readiness means across different job roles
• moving from informal experimentation toward structured skills development
• supporting non-technical staff in understanding and applying AI tools
• integrating responsible AI practices into everyday work
• demonstrating that learning initiatives contribute to organisational capability rather than isolated training activity

These challenges highlight the need for structured learning approaches that help organisations move from ad hoc experimentation toward coordinated workforce development.

Intervention

As part of its response to these challenges, LinkedIn developed an AI upskilling framework to structure its approach to workforce capability development across different levels of expertise.

The framework organises AI capability development into 5 broad levels:

• understanding
• applying
• building
• training and maintaining
• deep specialisation

Rather than focusing on a single training programme, the framework is intended to provide organisations with a shared reference model that can guide learning design, workforce planning, and capability assessment across different roles.

Within the LinkedIn Learning ecosystem, this framework is supported by a large catalogue of AI-related courses and curated learning pathways. The scale of available content makes it possible to support both rapid familiarisation and deeper, role-specific capability development, rather than relying on a single standardised programmed.

Training approaches typically combine short practical learning modules with longer structured learning pathways. For example, short learning activities are designed to help employees quickly understand how AI tools can support everyday tasks, while longer learning pathways provide more advanced skills development for specialised roles.

To address these challenges of making AI learning relevant across diverse roles, LinkedIn incorporates personalised learning recommendations based on role, skills and career signals. This helps employees engage with AI learning that reflects their context, particularly where confidence or exposure varies.

This approach supports organisations to introduce AI learning at different levels simultaneously, supporting both general AI awareness among non-technical staff and deeper technical skills development where required.

Journey

Based on LinkedIn’s experience working with organisations adopting structured AI learning approaches, several common stages of capability development can be observed.

Organisations adopting AI learning frameworks often progress through several stages of capability development. While organisational approaches vary, the evidence suggests a broadly consistent progression in how AI capability development is introduced and scaled.

The first stage typically focuses on awareness and basic AI literacy, where employees explore introductory learning materials to understand how AI technologies work and how they might affect their roles.

Once employees become familiar with AI concepts, organisations begin introducing applied learning pathways aligned with specific technologies or business initiatives. For example, learning programmes may be linked to the rollout of new AI-enabled workplace tools, helping employees understand how to integrate these tools into daily workflows.

As organisations gain experience with AI technologies, learning initiatives may evolve into longer-term capability development programmes, where AI skills are incorporated into broader talent development strategies and workforce planning.

In some cases, organisations use structured learning plans to support the rollout of specific AI tools. For example, one learning initiative linked to the deployment of Microsoft Copilot involved a 6-week learning programme designed to help employees understand the capabilities of the tool and apply it in everyday work tasks.

Outcome

Evidence from LinkedIn’s engagement with organisations using structured AI learning frameworks suggests several indicative outcomes.

First, structured learning pathways help organisations improve AI fluency among non-technical staff, enabling employees to better understand how AI tools can support routine work tasks such as document preparation, information retrieval, and meeting preparation.

Second, learning initiatives linked to AI tool rollouts can help increase employee engagement with new technologies. When training is introduced alongside technology deployment, employees are more likely to experiment with AI tools and integrate them into their workflows.

Third, structured frameworks allow organisations to track progression in AI capability across the workforce. Learning dashboards and skills tracking tools enable organisations to monitor participation, identify capability gaps, and align training with broader workforce development strategies.

However, evidence also suggests that training programmes are most effective when they are closely linked to practical workplace applications, rather than delivered as generic AI training.

AI upskilling impact

The LinkedIn Learning framework highlights several ways in which structured learning environments can support workforce AI capability development.

At the individual level, learning pathways help employees develop foundational AI literacy and build confidence in using AI tools within their roles.

At the team level, shared learning frameworks can help establish a common understanding of AI maturity across different departments, improving collaboration between technical and non-technical teams.

At the organisational level, structured learning frameworks enable organisations to monitor workforce AI capability and align training initiatives with broader technology adoption strategies.

Importantly, the framework emphasises the development of multiple categories of AI skills, including:

• technical skills, such as understanding how AI systems operate
• non-technical skills, such as applying AI tools in everyday work tasks
• responsible AI skills, including awareness of ethical considerations, governance, and safe AI use

This integrated approach allows organisations to move beyond isolated training courses toward broader AI capability development.

2.3.2 100 School: Building AI capability through challenge-based learning

There is a growing range of organisations offering AI training and upskilling services to employers. The following case study presents one example of how such services are being delivered in practice. It is intended to illustrate features of contemporary AI training approaches rather than to promote a specific provider.

Key lessons

Based on 100 School’s experience working with organisations across sectors and sizes, the following lessons have emerged. While drawn from specific client engagements, the findings are likely to be relevant for other organisations navigating similar AI adoption challenges.

1: Infrastructure matters more than content

AI adoption depends on learning infrastructure, social accountability, and daily practice, not just high-quality training materials. Organisations that have invested in content libraries or video-based platforms consistently report that access alone does not drive behaviour change. The structure around learning (daily rituals, peer visibility, leadership modelling) is what converts awareness into sustained practice.

2: Start with mindsets and frameworks, not tools

Focusing on how employees think about AI is more effective than tool-led training alone. Tool-specific training often fails because it teaches features rather than workflows and becomes outdated as tools evolve. Principle-first, tool-agnostic content gives professionals transferable mental models they can apply across any AI platform, today and in the future.

3: Measure learning by creation, not completion

The most meaningful indicator of AI learning is whether participants apply AI to real work tasks, rather than whether they complete quizzes or watch videos. When participants create tangible outputs, such as prompts, assistants, dashboards and workflow automations, they build both competence and confidence in a way that passive consumption cannot replicate.

4: Champions emerge from practice, not appointment

Organisations that try to appoint AI champions before building a shared foundation often struggle. The most effective champions are those who emerge naturally through consistent daily practice and peer engagement. A structured programme surfaces these individuals organically, providing organisations with data on who is ready for advanced training and peer coaching roles.

5: The activation layer is the missing piece

Many organisations have already invested in enterprise AI tools and licences. The gap is not tooling or awareness, it is the activation layer that makes those investments pay off. Structured, behaviour-focused training that sits between tool deployment and workforce adoption is where the majority of enterprise AI value remains untapped.

100 School is a UK-based AI education company that works with organisations across sectors to support practical, responsible, and sustained adoption of AI in the workplace. It specialises in AI upskilling programmes designed for professionals and teams who are not necessarily technical specialists, but who need to understand and use AI confidently and effectively as part of their everyday roles.

100 School delivers its training primarily through fully online, challenge-based learning models that emphasise real-world application, habit formation, and peer learning. Its approach is grounded in the recognition that AI adoption is as much a behavioural and cultural challenge as it is a technical one. As a result, its programmes focus on helping organisations move beyond awareness of AI towards meaningful and sustained use, aligned with business needs and workforce capability.

Since 2020, 100 School has trained over 60,000 professionals globally. Its enterprise clients range from smaller, high-growth companies to large multinational organisations. Its programme, 15 Days of AI, has been delivered to organisations ranging from global professional services firms to high-growth technology companies.

Challenge

Organisations approaching 100 School typically face 3 interrelated challenges:

1: The imagination gap

Employees may be curious or concerned about AI, but they cannot envision what it would look like integrated into their specific daily workflow. They lack the mental models, the vocabulary, and the practical reference points to bridge the gap between awareness and action. AI is often perceived as abstract, technical, or disconnected from real job tasks. This imagination gap is the foundational barrier that sits beneath all other adoption challenges.

2: The adoption plateau

Many organisations have already invested in enterprise AI licences (such as ChatGPT Enterprise, Microsoft Copilot, Claude or Google Gemini), but usage data tells a consistent story: a small core of enthusiastic users and a long tail of infrequent or non-users. The tools are deployed, but the behaviour change has not occurred. Significant potential value is left unrealised because employees are unsure how to apply AI effectively within the context of their specific roles, teams, or workflows.

3: Lack of sustained habits

Previous investments in generic online courses, video-based platforms, or standalone workshops frequently result in uneven skill levels across teams and limited long-term behaviour change. Many organisations struggle to build sustained habits via one-off engagements. As one learning and development leader described it: “We ran 2 months of lectures, trainings, workshops, inspiration… now we need to convert to action.”

These challenges are particularly acute in organisations with diverse roles, varying levels of digital confidence, and limited internal capacity to design structured AI learning journeys.

Intervention

100 School positions itself as an AI training partner for organisations, designing upskilling programmes that focus on behaviour change, habit formation, and practical application.

Engagement typically begins with a diagnostic stage, conducted with HR, learning and development, or transformation leads. This explores the organisation’s AI maturity level, existing patterns of AI use across teams, cultural attitudes towards AI, and the organisation’s underlying goals for adoption. Rather than assuming that AI training is always the right solution, 100 School works backwards from business needs, identifying where AI capability can genuinely add value and which employee groups should be prioritised.

The core intervention is 15 Days of AI, a fully online, challenge-based learning programme designed to establish a shared baseline of AI capability across an organisation. Participants invest approximately 15 minutes per day over 3 working weeks (Monday to Friday), building practical AI skills through daily application rather than passive content consumption. The total time commitment is under 4 hours per person.

Key features of the programme include:

• daily, time-bound learning challenges that require participants to create a tangible work artefact each day, not watch a video or complete a quiz
• tool-agnostic, principle-first content that teaches transferable frameworks (such as structured prompting, use-case identification, and workflow integration) rather than features of any specific AI tool
• peer interaction and social accountability through integrated Slack or Microsoft Teams channels, where participants share their work, ask questions, and learn from colleagues
• leadership participation to model engagement and signal organisational legitimacy, because when directors and senior managers complete challenges publicly it creates cultural momentum
• low-friction delivery with no login required, no LMS, and no additional software, content is delivered directly through email and web, removing barriers to participation

The programme is structured across 3 phases:

1: Discover (Days 1 to 5)

Participants build foundational understanding through frameworks, task mapping, and their first AI-generated outputs.

2: Augment (Days 6 to 10)

Where they apply AI to real work tasks such as email systems, data analysis, and workflow prototypes.

3: Systemise (Days 11 to 15)

Where they build personal AI toolkits, refine techniques, and complete capstone projects.

The programme is deliberately designed as a gateway intervention. Its primary purpose is to raise the floor of AI capability across the workforce, generate insight into where further specialised training may be required, and identify potential AI champions within the organisation.

Journey

Client organisations typically engage with 100 School through a staged AI upskilling journey.

The journey begins with the 15 Days of AI programme, which serves 3 functions simultaneously:

• establishing a shared baseline of AI understanding across participants
• surfacing variation in confidence, engagement, and capability
• identifying potential AI champions within the organisation

Throughout the programme, data on participation, engagement, and outputs is collected in a lightweight and non-intrusive way. Rather than relying on formal exams or quizzes, which many employees find discouraging or unrepresentative of real learning, 100 School measures learning by creation: the meaningful indicator is whether participants apply AI to real work tasks and produce tangible outputs, not whether they pass a knowledge test.

Following completion, 100 School provides clients with a post-programme insights report, enabling HR and learning teams to understand where skills are concentrated, where gaps remain, and which employees may benefit from further development. This report includes participation data, engagement patterns, and identification of AI champions: individuals who demonstrated consistent daily practice, peer support, and readiness for advanced training.

In some cases, this leads to additional, targeted interventions, such as advanced training for identified champions, role-specific deep-dives, or strategic advisory on embedding AI into team workflows. In other cases, organisations use the insights internally to inform future learning strategies without further external training. The journey is adaptive rather than prescriptive, and further training is offered only where it aligns with organisational needs.

Outcome

In a recent engagement with a global professional services firm, 122 employees from the people team (primarily HR professionals across compensation, talent acquisition, learning and development, and HR business partners) participated in the 15 Days of AI programme. Key outcomes included:

• 88% activation rate achieved, with participants logged in and engaged from day one, and zero unsubscribes throughout the programme
• 67% created 3 or more AI artefacts (prompts, assistants, dashboards, workflow prototypes), with participants submitting over 1,400 real work outputs across the programme
• confidence with AI grew significantly across all experience levels, with the proportion of participants reporting confident, regular use of AI in their work rising from 49% to 86% by the end of the programme
• 30 AI champions were identified, ready for advanced training, peer coaching, and identifying automation opportunities within their teams

The Slack channel created for the programme generated over 150 substantive posts, with questions answered by peers within hours and directors publicly sharing their daily progress. Directors and senior managers were enrolled as learners alongside their teams and shared their daily progress in the same shared Slack channel as all other participants. This peer visibility from leadership is what created the cultural momentum referenced.

This social accountability was not driven by pressure, but by visibility. When leadership modelled participation, it created cultural momentum that sustained engagement through to programme completion.

In a separate engagement, a UK-based technology company enrolled 319 employees in the programme as a voluntary, company-wide initiative. Outcomes included:

• 70% of participants reporting a positive shift in mindset towards AI
• 50% actively building or using AI in their work
• 21 AI champions identified for advanced training

Champions are identified through 3 data points:

• total lessons completed, because each lesson requires creating a real AI artefact, this is effectively a measure of how many tangible outputs they have produced
• consistency of on-time completion
• Slack/Teams engagement, how actively they share progress, reflections, and tips with fellow participants, which indicates how well they are modelling AI usage and contributing to a collaborative AI learning culture at their organisation

The identification is data-driven, not subjective.

Examples of outputs created during the programme included automated tracking tools, workflow optimisation for translation and reporting, and AI-enabled dashboards for senior leadership.

These outcomes demonstrate how structured AI training can generate both individual-level benefits and wider team-level ripple effects when participation is supported by leadership engagement, peer learning, and daily practice.

AI upskilling impact

100 School’s training approach aligns with the AI skills framework used in the AI Skills for the UK Workforce report, spanning:

• non-technical AI skills - confidence building, use-case identification, critical evaluation of AI outputs, and the ability to articulate how AI applies to specific professional contexts
• technical skills - structured prompting, workflow integration, data analysis with AI tools, and building AI-assisted systems appropriate to role and function
• responsible AI awareness - appropriate use, understanding limitations, data governance, and organisational compliance considerations

The emphasis on habit formation, real-world application, and shared learning positions client organisations supported by 100 School within the adoption and integration stages of AI capability development, with clear pathways towards scaling. The programme’s tool-agnostic approach, teaching transferable principles rather than features of any specific AI product, ensures that skills remain relevant as the AI tool landscape continues to evolve.

At impact level, benefits are observed across:

• individuals - increased confidence and practical competence
• teams - shared use cases, peer learning, and emerging community of practice
• organisations - improved productivity, more consistent AI adoption, and data-driven insight into workforce AI capability

2.4 Public sector and inclusion-focused models

2.4.1 NHS primary care: AI upskilling through practice-level adoption

Key lessons

1: AI adoption in healthcare must be supported by practical training

Hands-on demonstrations and practice-based training sessions help clinicians understand how AI tools can support real clinical workflows.

2: Governance support is essential for scaling AI adoption

Regulatory and clinical safety requirements can slow the adoption of AI tools. Centralised support structures, such as shared clinical safety officers, can help practices navigate these requirements.

3: AI training should involve the entire practice workforce

Effective implementation requires training not only clinicians but also administrative and reception staff, whose workflows interact with AI-enabled systems.

Challenge

Primary care within the NHS is facing growing operational pressure due to rising patient demand, workforce shortages, and increasing administrative responsibilities. General practitioners must manage high volumes of consultations while also completing documentation, generating referrals, reviewing test results, and responding to patient enquiries. These tasks often occur within tightly constrained consultation times. These challenges are illustrated by the example of James O’Riordan Medical Centre.

At James O’Riordan Medical Centre, these pressures reflect wider challenges across NHS primary care. Clinicians frequently need to divide their attention between patient interaction and documentation tasks such as writing consultation notes or preparing referral letters. This can reduce the quality of patient engagement while also increasing administrative burden for clinicians.

At the same time, the introduction of AI tools into healthcare environments is subject to strict governance requirements. Any new system must comply with data protection rules, clinical safety standards, and NHS regulatory frameworks before being used in practice. While these safeguards are necessary to protect patient safety, they can slow the adoption of new technologies.

Although AI technologies offer significant potential to improve clinical efficiency, adoption across primary care remains uneven. Many practices are interested in using AI but lack structured training opportunities and clear guidance on safe implementation.

The challenge therefore involves not only introducing AI technologies into healthcare settings but also developing the skills, confidence, and governance structures required for responsible AI adoption across the healthcare workforce.

Intervention

At James O’Riordan Medical Centre, several AI-enabled tools have been introduced to support clinical and administrative workflows. The most significant intervention has been the adoption of an AI clinical scribe, which automatically generates consultation notes and referral documentation.

During a patient consultation, the system records the conversation and produces a structured summary of the consultation. Within seconds of completing the appointment, clinicians receive a draft summary that can be quickly reviewed and approved before being saved to the patient record. Referral letters can also be generated automatically based on the consultation content.

This allows clinicians to focus fully on the patient during consultations rather than dividing their attention between communication and documentation.

The introduction of the tool was supported by practice-level AI training delivered by the technology provider. Initial training sessions typically lasted around one hour and introduced staff to the functionality of the system, how it integrates with clinical workflows, and how to review AI-generated outputs safely.

Training is ongoing rather than one-off. Practices periodically reconnect with the provider to receive updates and demonstrations of new features, typically every 6 to 8 months as the technology evolves.

Importantly, training includes all staff members within the practice, not only clinicians. Administrative staff and reception teams also participate because the technology affects processes such as referrals, appointment management, and patient communication.

Additional AI tools used within the practice include:

• automation systems for filing routine blood test results
• digital triage platforms that collect patient information before consultations
• AI-powered reception systems capable of answering patient calls and directing them to appropriate services

Training for these tools is also typically delivered through demonstrations, webinars, and practice visits organised by the technology providers.

Journey

Across NHS primary care, the integration of AI tools within the practice has developed incrementally through local experimentation and professional networks rather than through centrally coordinated NHS programmes.

Many clinicians across primary care first encountered AI-enabled tools through demonstrations organised by technology providers and professional events. These demonstrations allowed practices to explore potential applications of AI in clinical workflows before adopting the technologies within their own GP practices.

Regional training sessions across South West London have supported knowledge sharing. For example, in Croydon, a training event introducing AI tools was attended by around 40 healthcare professionals from approximately 50 GP practices. These sessions allowed clinicians to observe demonstrations, discuss practical use cases, and explore potential adoption within their own organisations.

However, regulatory requirements can create barriers to adoption. Each AI system must undergo clinical safety checks and approval processes before it can be used in clinical settings. To address this challenge, the Croydon federation funded a dedicated clinical safety officer responsible for managing regulatory approval processes across multiple GP practices.

By centralising this function, practices are able to adopt new technologies more efficiently while maintaining compliance with NHS safety standards.

As clinicians begin to observe tangible improvements in workflow efficiency and documentation quality, interest in AI tools continues to grow across practices.

Outcome

The use of AI tools within the practice has produced several important benefits.

One of the most significant impacts has been the reduction of administrative workload during consultations. AI-generated consultation summaries allow clinicians to give their full attention to patients while ensuring that clinical documentation is completed accurately and efficiently.

This improves both workflow efficiency and the quality of patient interaction. Clinicians are able to focus on listening to patients rather than simultaneously typing notes during consultations.

AI tools also improve the consistency and completeness of documentation. Automatically generated summaries and referral letters help ensure that important clinical information is recorded and reduce delays in administrative processing.

More broadly, these tools allow clinicians to spend more time focusing on complex medical cases rather than routine administrative tasks.

Despite these benefits, adoption across the wider NHS remains uneven due to regulatory complexity, limited training provision, and the absence of coordinated national programmes for AI workforce development.

AI upskilling impact

The experience at James O’Riordan Medical Centre highlights how AI upskilling in healthcare is currently emerging through practice-level learning rather than structured national programmes.

The training delivered within the practice supports 3 categories of AI skills aligned with the AI skills framework used in this report:

• technical AI skills - staff learn how to operate AI-enabled tools, integrate them into existing clinical systems, and manage AI-generated outputs such as consultation summaries and referral documents
• non-technical AI skills - clinicians develop the ability to interpret AI outputs, incorporate them into clinical workflows, and make informed decisions based on AI-assisted information
• responsible AI skills - training also emphasises data protection, patient privacy and regulatory compliance, clinicians must understand how AI tools interact with patient data and ensure that outputs are reviewed and validated before being used in clinical decision-making

Evidence from the practice suggests that face-to-face demonstrations are the most effective way to introduce AI tools, as clinicians are able to observe real-world use cases and understand how the technology fits into clinical workflows. Once initial engagement is established, shorter online sessions and update briefings can support ongoing learning as technologies evolve.

Overall, AI tools are viewed not as replacements for clinicians but as support systems that reduce administrative workload and enable healthcare professionals to focus more on patient care.

Based on the AI adoption pathway used in this report, the practice currently operates at an early operational stage, where AI tools are being integrated into workflows and supported by informal training and peer learning, but where wider system-level training programmes have yet to emerge.

2.4.2 Good Things Foundation: Addressing the AI literacy gap through inclusive digital learning

Key lessons

1: Resistance and misconceptions should be expected

People approach AI with ethical concerns and anxiety, but with structured support they can progress. Often, a short, supported intervention can have a significant positive impact on someone’s comfort and confidence in using AI.

2: Organisations should start before perfect clarity is available

AI is evolving rapidly. Waiting for complete certainty risks widening the gap.

3: Digital inclusion programmes are becoming more essential in the age of AI

As AI becomes embedded in online services, AI literacy, digital literacy and media literacy need to be addressed together.

Challenge

Good Things Foundation is a digital inclusion charity. As AI adoption grows, the organisation is observing a shift in the digital inclusion landscape. AI is reshaping online experiences and demands new skills to participate confidently and safely in digital environments. The minimum specification for suitable devices is increasing, while new types of interfaces and issues around personal data security are emerging. All this adds an extra layer of complexity for those who are digitally excluded and could also cause people who are currently digitally confident to feel left behind.

The key challenge identified by Good Things Foundation is a growing AI literacy gap. AI literacy is defined as the ability to use AI safely, confidently and effectively. The gap refers to a widening societal divide between people who can access and benefit from AI and those who cannot. Good Things Foundation highlighted that AI literacy remains limited among communities that are already digitally excluded. This is particularly concerning because these groups may be among those who could benefit most from using AI tools.

There is potential for AI to create a more equal digital playing field for people with low or limited digital skills, for example through translation tools and chatbots. However, individuals who are already digitally excluded may experience increased complexity, reduced confidence and greater barriers to participation because of concerns around privacy, data use, bias, environmental impact, and increased exposure to AI-created misinformation.

As AI becomes embedded in more services, skills requirements shift, and trust and safety become bigger challenges, inclusive AI programmes are even more essential. Existing AI training materials often assume prior knowledge and are typically aimed at workforce preparation rather than foundational literacy. As such, they are frequently inaccessible to people with low or no digital skills.

Intervention

To address the AI literacy gap, Good Things Foundation developed the AI Gateway, a free, open-access AI literacy platform. The AI Gateway:

• uses plain English
• provides simple explanations
• avoids technical language
• includes captioned videos
• can be accessed using translation tools such as Google Translate
• is delivered in bite-sized modules for flexible learning

The learning design includes explainer videos, interactive e-learning topics, practice activities, and knowledge-check quizzes.     

Learners can follow a structured progression route or build their own journey. The content starts with understanding what AI is, environmental and societal considerations, generative AI risks and responsible use, example prompts for first interactions with text generators, and knowledge checks. The Gateway focuses on everyday use cases (for example creating a recipe, planning a trip, preparing for an interview) rather than professional or technical scenarios.

Learning journey

The AI Gateway is used in 2 main ways:

1: Supported learning via the National Digital Inclusion Network

Organisations within the network provide guided support before, during and after Gateway completion.

2: Independent learning

Because the platform is open access, individuals may use it independently without structured follow-up.

The Gateway acts as a foundational introduction, a confidence-building tool, and a launchpad for higher-level AI learning. It is explicitly entry-level and designed to support individuals who would otherwise struggle to engage with more advanced AI training.

Outcome

As of the interview date with Good Things Foundation, the AI Gateway had recorded just over 13,000 topic completions.

While large-scale impact metrics are not yet available, qualitative feedback highlights early benefits:

• increased confidence among library staff
• staff recommending the Gateway internally to improve AI support provision
• learners reporting independent AI use at home
• tutors using the Gateway as preparation before Copilot sessions

One tutor reported that learners began using AI independently after completing Gateway content, describing this as a significant step forward.

AI upskilling impact

The primary impact observed relates to:

1: Confidence and demystification

The Gateway helps demystify AI, particularly for individuals with low digital skills. Learners move from hesitation and misconceptions toward active experimentation.

2: Transferable skills

Although examples are domestic in nature, learners develop transferable skills such as writing prompts, interacting with text generators, evaluating outputs, and understanding risks and responsible use. These skills can later be applied to job search or workplace contexts.

3: Strengthening digital inclusion infrastructure

Staff in community organisations also require support. Many digital inclusion practitioners lack confidence in navigating AI’s ethical and practical complexities. Without strengthening this infrastructure, exclusion risks may intensify. Digital inclusion providers are already under financial strain, and without funding for digital inclusion infrastructure, barriers may worsen.

3. Cross-case synthesis

3.1 What works across the case studies

The 10 case studies show clear and consistent patterns in how organisations build AI skills. These patterns appear across different sectors, including construction, healthcare, creative industries, and professional services. They also apply to both large organisations and smaller firms.

The first pattern is how organisations start. In most cases, AI use begins in an informal way. People try tools on their own, often without training or clear guidance. Over time, this changes. Then, organisations begin to introduce more structured learning. This includes clear training pathways, shared resources, and support from managers. This shift from informal use to structured learning is seen across all types of organisations.

The second pattern is the strong link between training and real work. Training is most effective when it is based on tasks that people already do. This includes writing reports, analysing information, managing projects, or supporting customers. When training is linked to real work, people can see the value more clearly. This helps them build confidence and apply what they learn.

The third pattern is that training works best when it is part of everyday work. In stronger examples, AI learning is not separate from the job. Instead, it is built into tools, workflows, and systems. People learn while doing their work, rather than stopping work to attend long courses. This makes learning more practical and easier to sustain over time.

The fourth pattern is the role of leadership. Organisations with strong leadership support tend to make faster progress. Leaders help set clear expectations about AI use. They also create a safe space for people to learn and try new tools. This is especially important in sectors where risks are higher.

The fifth pattern is the mix of skills needed. This reflects the AI skills framework set out in the AI Skills for the UK Workforce report. Across all case studies, organisations develop a combination of technical, non-technical, and responsible AI skills. Most workers do not need advanced technical skills. Instead, they need to understand how to use AI tools, how to check outputs, and how to apply them in their role. This shows that non-technical and responsible AI skills are essential across the workforce.

The final pattern is how learning is delivered. Organisations do not rely on one-off courses. Instead, they use ongoing learning. This includes short training sessions, peer learning, and practice over time. This approach works better because AI tools and practices are changing quickly. Continuous learning helps workers keep up and stay confident.

3.2 How this links to the PRIMES principles

Patterns identified across the case studies align closely with the PRIMES principles. The case studies provide concrete evidence of how these principles operate in practice across different organisations and sectors, reinforcing their relevance and applicability.

Practical

Across the case studies, training is most effective when it is linked to real tasks and decisions. People learn by applying AI in their day-to-day work, rather than through abstract or theoretical training.

Reachable

The case studies show that training works better when it is accessible and easy to engage with. This includes using simple language, flexible formats, and approaches that support people with different levels of experience.

Integrated

Across the evidence, organisations achieve stronger results when training is built into everyday work. This includes linking learning to tools, workflows, and organisational systems, rather than treating it as a separate activity.

Modular

The case studies show that shorter, flexible learning formats support engagement. Training is often delivered in small parts, allowing individuals to learn at their own pace and alongside their work.

Expandable

Across organisations, training often starts small and then grows over time. The evidence shows that approaches can be extended across teams and functions as confidence and capability increase.

Sustainable

The case studies show that effective AI training is ongoing. Organisations update learning over time and support responsible AI use, helping ensure that skills remain relevant as tools and practices change.

4. Limits and transferability

4.1 Why these case studies are illustrative

These case studies are designed to show how AI upskilling works in practice. They are illustrative rather than representative.

Each case reflects a specific organisation, sector, and stage of AI adoption. This includes differences in size, resources, digital maturity, and workforce needs. For example, large organisations often have more formal systems and dedicated training resources, while SMEs rely more on targeted and practical approaches.

The case studies are based on qualitative interviews. This provides detailed insight into how organisations design and deliver training. However, they do not represent all organisations across the UK.

The evidence shows what is possible in real settings. It does not suggest that all organisations will follow the same path or achieve the same outcomes. AI is also evolving quickly. Tools, skills, and ways of working are changing over time. This means that approaches may need to adapt as technology develops.

4.2 Where models may and may not generalise

The case studies highlight approaches that can inform wider practice. However, these approaches do not transfer in the same way across all organisations.

Some elements are widely applicable. These include linking training to real work, supporting ongoing learning, and combining technical, non-technical, and responsible AI skills. These patterns appear across all case studies and can be adapted to different settings. Other elements depend on context.

Organisational scale is important. Large organisations can invest in structured programmes, internal platforms, and dedicated teams. Smaller organisations are more likely to use focused, use-case-led approaches and external support.

Sector context also shapes what is possible. In regulated sectors such as healthcare, defence, and professional services, training must align with strict rules and governance. In less regulated sectors, organisations may have more flexibility to experiment and adapt quickly.

Digital maturity affects how easily organisations can adopt AI. Organisations with strong data systems and digital infrastructure can integrate AI into workflows more quickly. Others may need to build these foundations first. Workforce needs also vary. Some groups require more support, time, or flexible learning formats. Inclusive design is important to ensure that training is accessible to all.

Overall, while specific models may not generalise directly, the core principles identified across the case studies remain relevant. These principles can be adapted to suit different organisational and sector contexts.