TD Knowledge & Practice EcosystemTD Consultancy ↗

A proportionate traffic-light framework

Can I Use AI for This?

Classify the task before choosing a tool. Risk depends on the information, consequence, approval, checks and who remains responsible.

Green

Generally lower risk

Non-confidential drafting, formatting, fictional scenarios, terminology explanations, brainstorming, public-report summaries and presentation structures.

Use an approved tool, check the answer and follow policy.

Amber

Use stronger controls

Internal policies, properly anonymised service information, consultation analysis, operational documents, evaluation, research synthesis and professional correspondence.

Confirm approval, minimise information, document purpose and apply proportionate specialist review.

Red

Do not delegate

Safeguarding, diagnosis, prescribing, individual risk assessment, identifiable records, treatment eligibility, service restriction and emergency clinical decisions.

Use established human-led professional and organisational processes.

Interactive decision aid

Work through seven questions

This gives educational guidance, not legal, clinical or organisational approval.

  1. 01Does this involve identifiable or potentially identifiable information?
  2. 02Could the output materially affect a person’s care, safety, liberty or access to services?
  3. 03Is this a safeguarding decision or response to an immediate concern?
  4. 04Is this a diagnosis, prescribing, treatment-eligibility or other clinical decision?
  5. 05Are the tool and this specific use permitted by organisational policy?
  6. 06Can the output be independently checked against authoritative sources?
  7. 07Will a named, suitably qualified person remain responsible for the final work?

Not yet classified

Answer each question

This educational tool cannot provide legal, clinical or organisational approval.

Explore approved lower-risk work

Find a practical use case with clear safeguards.

Open the use-case library →