TD Knowledge & Practice EcosystemTD Consultancy ↗
AI for Drug & AlcoholPractice
Menu

For services, leaders and commissioners

AI Governance Toolkit

Practical starting templates for introducing, approving, monitoring and reviewing AI use across drug and alcohol services.

Take the readiness assessment
Use these as starting points

These resources do not replace legal advice, information-governance advice, a formal DPIA, clinical-safety processes, organisational policy or procurement rules. Adapt and approve them through your local governance routes.

Template 01

AI acceptable-use policy

Set clear boundaries for approved tools, permitted tasks and human accountability.

Important

Starting template only. It requires local legal, information-governance, safeguarding, workforce and senior approval.

  1. Purpose and scope
  2. Approved tools and accounts
  3. Permitted uses
  4. Prohibited uses
  5. Confidentiality and data minimisation
  6. Human oversight and accountability
  7. Verification and references
  8. Record keeping and transparency
  9. Training
  10. Incident reporting
  11. Monitoring and review

Template 02

AI use-case register

Keep a visible record of proposed, approved, paused and retired uses.

Important

A register supports oversight; it does not itself approve a use case.

  1. Use-case name
  2. Team or department
  3. Purpose and intended benefit
  4. AI system and account type
  5. Information involved
  6. Risk classification
  7. Approval route and status
  8. Responsible owner
  9. Human-review method
  10. Measures and incidents
  11. Review date
  12. Retirement or exit plan

Template 03

AI risk assessment

Identify foreseeable risks and controls before an AI-assisted workflow begins.

Important

Use alongside formal local risk, data-protection, clinical-safety and safeguarding processes.

  1. Task and affected people
  2. Necessity and alternatives
  3. Confidentiality and data protection
  4. Safeguarding and clinical implications
  5. Bias and inequalities
  6. Accuracy, hallucination and omission
  7. Transparency and explainability
  8. Accountability
  9. Supplier and model-change risk
  10. Human oversight
  11. Monitoring, stop criteria and residual risk

Template 04

AI output verification checklist

Give staff a repeatable check before AI-assisted material is used or shared.

Important

The depth of review must match the potential consequence of an error.

  1. Purpose and audience confirmed
  2. Factual accuracy checked
  3. Original evidence opened
  4. References confirmed to exist
  5. Important omissions considered
  6. Dates and jurisdiction checked
  7. Bias and stereotypes reviewed
  8. Certainty matches evidence
  9. Confidentiality rechecked
  10. Tone and accessibility reviewed
  11. Named human approval recorded

Template 05

AI incident and error log

Capture errors, near misses and concerns so the organisation can learn.

Important

Follow existing incident, safeguarding, clinical-safety, cyber-security and data-breach procedures where relevant.

  1. Date and reporter
  2. System and version if known
  3. Task and intended use
  4. Error, concern or near miss
  5. Information involved
  6. Potential and actual impact
  7. Immediate action
  8. People and processes notified
  9. Root-cause questions
  10. Learning and control change
  11. Owner and follow-up date

Template 06

Manager approval checklist

Support a proportionate decision about a proposed team use.

Important

Approval must sit with the correct accountable route, not automatically with a line manager.

  1. Problem and expected benefit
  2. Why AI is appropriate
  3. Approved system confirmed
  4. Information classification
  5. Risk assessment completed
  6. Human owner named
  7. Verification process tested
  8. Staff training completed
  9. Service-user impact considered
  10. Equality and accessibility considered
  11. Pilot boundaries and stop criteria
  12. Review date agreed

Template 07

AI procurement checklist

Bring practice, governance and evaluation questions into purchasing from the start.

Important

Use with organisational procurement rules, DTAC where applicable, legal advice and formal due diligence.

  1. Defined need and alternatives
  2. Evidence of benefit
  3. Intended users and affected people
  4. Data flows and processing locations
  5. Security and access controls
  6. Regulatory and clinical-safety position
  7. Bias and performance across groups
  8. Human oversight
  9. Interoperability and accessibility
  10. Costs and resource assumptions
  11. Model updates and change control
  12. Audit rights, exit and data return

Template 08

Supplier and vendor questionnaire

Ask suppliers for information needed to assess suitability and govern change.

Important

Supplier answers require verification and contractual clarity.

  1. System purpose and model provider
  2. Data processing locations
  3. Training use and opt-out
  4. Retention and deletion
  5. Security certifications and testing
  6. Data-processing agreement
  7. Sub-processors
  8. Accuracy and limitations
  9. Bias testing
  10. Auditability and logs
  11. Model-update notification
  12. Human control
  13. Accessibility
  14. Incident notification
  15. Exit, portability and deletion

Template 09

AI transparency statement

Explain clearly where and why AI is used, its boundaries and who remains responsible.

Important

Check whether a formal Algorithmic Transparency Recording Standard record or other disclosure is required.

  1. What the tool does
  2. Why it is used
  3. What information it processes
  4. What it does not decide
  5. Human involvement
  6. How output is checked
  7. Known limitations
  8. How people can ask questions
  9. How concerns and corrections are handled
  10. Last review date

Template 10

Staff AI-use declaration

Create proportionate transparency for significant AI assistance in professional work.

Important

Do not turn declarations into surveillance; agree when they are necessary and how they will be used.

  1. Work item and purpose
  2. System used
  3. Nature of AI assistance
  4. Information supplied
  5. Checks completed
  6. Sources verified
  7. Changes made by the author
  8. Named responsible person
  9. Date
  10. Any limitation or concern

Template 11

AI implementation checklist

Move from a defined problem to a controlled pilot and evidence-led review.

Important

Start small. A tool demonstration is not evidence of safe organisational implementation.

  1. Problem and non-AI alternatives
  2. Stakeholder and lived-experience involvement
  3. Use-case and workflow mapping
  4. Governance ownership
  5. Supplier and information assurance
  6. Policy and training
  7. Pilot group and boundaries
  8. Baseline and success measures
  9. Incident and stop process
  10. Equality and accessibility checks
  11. Review decision
  12. Scale, change or stop

Template 12

AI review and audit template

Test whether an approved use still operates as intended and remains justified.

Important

Review after significant supplier, model, policy, data or workflow changes as well as on schedule.

  1. Use-case and owner
  2. Current system and version
  3. Purpose still valid
  4. Usage and adoption
  5. Benefits against baseline
  6. Output quality
  7. Errors and incidents
  8. Equality and accessibility
  9. Staff and service-user feedback
  10. Supplier or model changes
  11. Controls working
  12. Decision: continue, change, pause or stop
  13. Next review

Template 13

DPIA preparation questions

Help teams gather the information an information-governance lead may need for a formal DPIA.

Important

This is not a DPIA and does not replace Data Protection Officer or information-governance advice.

  1. Purpose and necessity
  2. People and information affected
  3. Data sources and lawful basis questions
  4. Special-category and criminal-offence data
  5. Data flow and processors
  6. Automated or significant decisions
  7. Transparency
  8. Accuracy and bias
  9. Security and retention
  10. Rights and access
  11. Risks to people
  12. Alternatives and controls
  13. Consultation and sign-off

Move from policy to practice

Check your organisational starting point.

Take the readiness assessment →