DWP: Work Hub
To help users look for a better job, explore skills and training and advise on work that suits the user's individual circumstances.
1. Summary
1 - Name
Work Hub
2 - Description
Work Hub, Part of the Jobs and Careers Service (JCS), helps people explore job opportunities, improve their CV, and identify relevant training by analysing information they choose to provide. It is used to give tailored guidance and support to help users find work or progress in their careers.
3 - Website URL
N/A
4 - Contact email
N/A
Tier 2 - Owner and Responsibility
1.1 - Organisation or department
Department for Work and Pensions
1.2 - Team
JCS Digital Product Team, Working Age Services
1.3 - Senior responsible owner
Service Owner, Working Age Services
1.4 - Third party involvement
Yes
1.4.1 - Third party
Amazon Web Services (AWS)
1.4.2 - Companies House Number
BR019315
1.4.3 - Third party role
AWS provides the Amazon Bedrock service used to process text inputs for features such as CV improvement, skills extraction and conversational responses. Processing is transient and data is not retained or used for model training.
1.4.4 - Procurement procedure type
Call-off via existing agreement
1.4.5 - Third party data access terms
Third parties process data only where necessary to deliver specific functionality. Personal data is minimised and protected by guardrails, and is not retained outside DWP systems.
Tier 2 - Description and Rationale
2.1 - Detailed description
The AI Work Assistant is part of the Jobs and Careers Service, a voluntary digital service that helps users find work or improve their employment prospects. Users interact with the tool through an online interface, including a chatbot that gathers information about their goals, skills and circumstances. The tool uses algorithms and Large Language Models (LLMs) to:
- analyse user inputs such as CVs and skills
- suggest suitable jobs, training opportunities and career pathways
- generate tailored guidance and action plans
The system combines structured workflows (e.g. CV parsing, skills mapping and job matching) with conversational AI to guide users through the service. It does not make decisions about benefit entitlement or automate statutory decisions. The tool is designed for adults seeking employment support. It is not intended for enforcement, compliance, or decision-making that affects legal rights.
2.2 - Benefits
- Provides personalised employment support at scale
- Helps users identify suitable jobs and training opportunities
- Improves accessibility to career guidance
- Reduces reliance on manual support channels
- Supports the government’s objective to increase employment levels
2.3 - Previous process
Previously, employment support within DWP was delivered primarily through work coaches, online job boards and static guidance. Personalised support required manual interaction and was not consistently available at scale
2.4 - Alternatives considered
Alternatives included:
- Fully manual support via work coaches (limited scalability)
- Static digital tools without personalisation
- Rule-based systems without AI
An AI-assisted approach was selected to provide more personalised, scalable support while maintaining human oversight and voluntary usage.
Tier 2 - Deployment Context
3.1 - Integration into broader operational process
The tool supports users directly through the Jobs and Careers Service. It provides:
- job recommendations
- skills assessments
- CV improvements
- tailored action plans
Outputs are presented to users through the digital interface and are intended to guide, not determine, user decisions.
3.2 - Human review
There is no mandatory human review of each interaction, as outputs are advisory only. However:
- outputs are monitored through evaluation and Quality Assurance (QA) processes
- users remain responsible for acting on advice
- human review is used during testing and assurance processes
3.3 - Frequency and scale of usage
The service is new and has been made available to several thousand users as part of a private-beta, we expect to scale it to a national audience over time.
3.4 - Required training
N/A
3.5 - Appeals and review
N/A
Tier 2 - Tool Specification
4.1.1 - System architecture
The tool is a web-based service comprising a user-facing interface, internal service integrations, managed foundation-model services and DWP-managed data storage. It operates within DWP-approved hosting and security arrangements.
4.1.2 - System-level input
user-provided data (e.g. CV, skills, preferences), responses to chatbot questions and other accessible account data where users log in
4.1.3 - System-level output
job recommendations, CV improvements, skills insights and suggested job roles, personalised action plan items
4.1.4 - Maintenance
The service is iteratively developed and continuously monitored, with regular testing, evaluation and updates to improve performance and safety.
4.1.5 - Models
Amazon Bedrock is used as the managed foundation-model service. At the date of publication, the service uses Claude Sonnet 4.6. The model configuration may change as we evolve the service.
Tier 2 - Operational Data Specification
4.4.1 - Data sources
Information provided directly by users in their chat prompt, other user information if the user is logged in
4.4.2 - Sensitive attributes
May include health data and employment-related personal information provided voluntarily by users.
4.4.3 - Data processing methods
Guardrails are in place to remove Personal Identifiable Information (PII) before being processed
4.4.4 - Data access and storage
Data is stored securely in DWP systems. External processing is transient and not stored outside DWP environments.
4.4.5 - Data sharing agreements
No routine sharing with external organisations beyond service providers required to deliver functionality.
Tier 2 - Risks, Mitigations and Impact Assessments
5.1 - Impact assessments
DPIA completed May 2026 and Equality Assessment completed May 2026. The assessments considered the processing of personal information, equality impacts and appropriate controls for the service.
5.2 - Risks and mitigations
Accuracy – AI may produce incorrect or misleading outputs
- mitigated through testing, evaluation frameworks and user disclaimers
Bias and fairness – recommendations may reflect or reinforce bias, including across protected characteristics
- mitigated through guardrails, constrained outputs and demographic bias testing
- mitigated through guardrails and constrained outputs, alongside bias testing using synthetic and paired demographic test cases to assess whether equivalent customer profiles receive materially different recommendations when demographic characteristics or signals are varied. Testing is supported by human review and ongoing monitoring and re-testing as the service develops.
Data minimisation – risk of collecting more data than needed
- mitigated by collecting only necessary data and allowing user control
Transparency – users may not understand AI usage
- mitigated through clear messaging, onboarding and privacy notices