Skip to content

obielin/public-sector-ai-playbook

Repository files navigation

🏛️ Public Sector AI Playbook

Practical patterns, templates, and decision frameworks for deploying AI responsibly in UK public sector contexts — from healthcare to local government to procurement.

License UK Gov Aligned EU AI Act LinkedIn Twitter


The UK Government published its AI Playbook in February 2025. It tells you what to think about. This repository gives you working artefacts to act on it.

Every template here is ready to fill in. Every checklist has been designed for practitioners, not just policy readers. Every framework is grounded in UK and EU regulation.

If you're a civil servant, council officer, NHS team, or public sector AI practitioner — this is for you.


📂 What's Inside

00-foundations/ — Start Here

The principles and readiness baseline every public sector AI project needs.

Document What It Does
Principles 8 core responsible AI principles mapped to UK/EU regulation
AI Readiness Assessment 25-question assessment across 5 dimensions; produces a readiness score and priority actions

01-use-case-assessment/ — Is AI Right for This Problem?

Document What It Does
Use Case Decision Framework Step-by-step framework: validate the problem, assess risk tier (1–4), determine governance requirements

02-procurement/ — Buying AI Responsibly

Document What It Does
Supplier Due Diligence Checklist 40+ checklist items across 7 sections; sign-off table for legal, DPO, and security
AI Contract Clause Library Ready-to-adapt contract clauses covering transparency, human oversight, data protection, security, IP, and performance

03-sector-playbooks/ — Sector-Specific Guidance

Sector What's Covered
Healthcare NHS-specific: MHRA/SaMD classification, DCB0129, NHS DSPT, national data opt-out, pre-deployment checklist
Local Government ATRS compliance guide, governance board structure, planning AI pattern, scrutiny and democratic accountability
Criminal Justice (Coming soon)

04-governance/ — Running AI Safely

Document What It Does
AI Risk Register 7 pre-populated risk entries (bias, hallucination, data breach, oversight loss, drift, vendor failure, trust) with likelihood/impact scoring

05-deployment/ — Going Live Responsibly

Document What It Does
Model Card Template Complete model card: training data, architecture, performance by subgroup, fairness metrics, monitoring, accountability
Bias Audit Template Pre-deployment fairness evaluation: demographic parity, equal opportunity, predictive parity, intersectional analysis

tools/ — Automation

Tool What It Does
readiness_scorer.py Interactive CLI that scores your AI readiness assessment and saves a Markdown report
governance_report_generator.py Takes a JSON project definition and generates a formatted governance pack

🚀 Getting Started

For a new AI project: Start at the beginning.

  1. Complete the AI Readiness Assessment — or run python tools/readiness_scorer.py
  2. Use the Use Case Decision Framework to determine your risk tier
  3. Follow the governance requirements for your tier
  4. Use the procurement documents if buying AI from a vendor
  5. Complete the Model Card and Bias Audit before go-live

For a specific sector: Go straight to the sector playbook.

For procurement: Use the Supplier Due Diligence Checklist and Contract Clause Library alongside your legal team.


🧭 Regulatory Alignment

All content in this playbook is aligned with:

Framework Coverage
UK AI Playbook (Feb 2025) Core principles, lifecycle guidance, procurement
EU AI Act (2024) Risk classification, high-risk AI requirements, human oversight
UK GDPR / DPA 2018 Data protection, DPIA requirements, lawful basis
Equality Act 2010 PSED, protected characteristics, bias assessment
NIST AI RMF GOVERN, MAP, MEASURE, MANAGE functions
ATRS Algorithmic transparency publication requirements
DCB0129 NHS clinical safety standard
NCSC Secure by Design Security requirements for AI systems
ISO 42001 AI management system standard

🗺️ Roadmap

  • Foundations, principles, readiness assessment
  • Use case decision framework with risk tiering
  • Procurement: supplier due diligence + contract clauses
  • Healthcare sector playbook
  • Local government sector playbook
  • AI risk register template
  • Model card template
  • Bias audit template
  • Python tools (readiness scorer, governance report generator)
  • Criminal justice sector playbook
  • Incident response plan template
  • Post-deployment monitoring dashboard (Streamlit)
  • Case study library (anonymised real examples)
  • Welsh and Scottish devolved context supplements
  • Central government / HMRC specific patterns

🤝 Contributing

Contributions are very welcome — especially:

  • Additional sector playbooks (DWP, HMRC, MOJ, education)
  • Anonymised case studies from real public sector deployments
  • Updates when regulations or guidance changes
  • Translations of templates into Welsh (Cymraeg)

See CONTRIBUTING.md for guidelines.


⚖️ Licence

This repository is licensed under Creative Commons Attribution 4.0 (CC BY 4.0). You are free to use, adapt, and redistribute — including for commercial purposes — with attribution.

This playbook is provided for guidance purposes and does not constitute legal advice. Always consult qualified legal, data protection, and regulatory experts for your specific circumstances.


🙋 About

Created and maintained by Linda Oraegbunam — AI & ML Engineer, PhD Candidate at Leeds Business School researching autonomous AI systems, governance, and sustainability.


Star this repo to help public sector practitioners and policy professionals find it. Every star helps this reach the people who need it.

About

Practical templates, decision frameworks, and governance tools for deploying AI responsibly in UK public sector contexts — aligned with the UK AI Playbook 2025, EU AI Act, and ATRS.

Topics

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages