Responsible AI for trans inclusion

Joanne Lockwood | SEE Change Happen Ltd | 31 October 2026

Why this matters

When we ask AI about people, we need to look beyond whether an answer sounds convincing. We also need to notice the picture it paints. Does it help us understand the situation, or turn someone’s everyday life into a diagnosis, a dispute or a problem to manage?

This briefing is for colleagues using AI to support people, policy and organisational decisions. It sets out what to look for in an answer, how to retain human responsibility, and where the Trans Inclusion Toolkit can help.

My starting point is simple: use AI to support informed judgement, while keeping the people affected in view. Sound evidence and respectful treatment belong together. Neither removes the need to understand a genuine constraint, listen to concerns or make a difficult decision.

An answer may cite a real source and still give it more weight, or wider relevance, than the question justifies. That is a reason to examine how we use AI, rather than assume that every polished answer deserves our trust.

Accuracy is the beginning

A manager asking how to welcome a colleague using a new name needs practical help with respectful communication and privacy. An unsolicited discussion of surgery or court rulings can distract from that task. A question about a particular legal obligation, on the other hand, needs accurate legal context. Avoiding it to sound reassuring would be just as unhelpful.

The test is relevance. Ask what the person needs to know, what the evidence supports and what the proposed action will mean in practice.

What responsible use looks like

Answer the question in front of you. Keep an everyday conversation everyday. Ask for additional context when it could change the answer, rather than collecting personal details simply because someone is trans.

Make the evidence checkable. A citation should support the actual claim. Check the source, date, jurisdiction and relevant passage. Separate what a source says from an interpretation or a suggested course of action. A confident summary is not a substitute for reading the part on which a consequential decision depends.

Treat people as participants. Listen to those affected without expecting one person to represent a whole community. Lived experience can reveal harm, barriers and overlooked options. It should inform a decision alongside other relevant evidence, without being presented as a legal ruling or dismissed as merely subjective.

Be fair without creating false balance. Take concerns seriously and apply consistent standards of evidence and respectful treatment. Distinguish a belief from specific behaviour, and an allegation from an established fact. Giving every assertion equal weight can obscure rather than clarify the issue.

Keep responsibility with people. An AI-generated draft does not approve a policy, resolve a complaint or determine someone’s rights. A named person should check the important points, make the decision and be able to explain it. Where the evidence is missing or the consequences are serious, seek appropriate specialist advice.

A workplace example

The question: “A colleague has told us she is using a new name. What should we do before Monday?”

A useful starting point would be to agree privately how she wants the change communicated, identify the systems and practical arrangements that need updating, and clarify who will do what. There is no need to make medical history the centre of this conversation.

If a separate issue raises a genuine legal, safeguarding or operational question, address it on its own facts. Keep the discussion proportionate and avoid circulating personal information beyond those who need it.

This is an illustrative approach, not a statement that every organisation or situation has identical requirements. The point is to help the manager take a sensible next step without turning a colleague into a case file.

AI-assisted writing needs the same care

AI can help organise information and shape a draft, but the author still needs to recognise the argument as their own, verify claims and respect the people whose experiences or work appear in it. Do not invent testimony or present a generated persona as a real person’s account. Check consent, attribution and the recipient’s rules before using AI-assisted material.

Different organisations may draw different boundaries around AI-assisted work; those choices should be clear to the people doing the work.

How the Trans Inclusion Toolkit helps

I created the Trans Inclusion Toolkit to help organisations work through complex people and policy questions with more structure. It combines organisational assessment, practical resources and AI-assisted tools, while leaving decisions with the people responsible for them. [1]

Understand your starting point

Readiness checks and the organisational diagnostic help teams identify where they need to look more closely. They provide a starting point for discussion about policy, responsibilities and practical arrangements, rather than treating a well-written policy as proof that everything works. [1]

Examine the basis of an answer

Quinn, the Toolkit’s AI assistant, draws on the Toolkit’s case summaries, the judgments it holds, and its own guidance, playbooks and scenarios. Where a source supports it, Quinn gives paragraph references so you can check the point against the text. The published method describes summaries linked to full judgments, with a second-reader check. Corrections to descriptions of the law are checked by two independent readers. [2]

That gives users a route back to the evidence. It is still important to read the relevant passage, consider its application and recognise what a source does not establish.

Explore how a decision might work

The Toolkit brings together authorities, playbooks, personas, scenarios and a risk library. These resources support consideration of different circumstances and practical consequences. Its scenario tools and documented decision workflows help teams rehearse issues and record the basis for their choices. [3]

Personas and scenarios support reflection; they do not replace consultation with the actual people affected. A documented rationale also needs to stand up to scrutiny in practice.

Turn understanding into action

The Toolkit includes support for policy review, implementation planning and organisational briefings, alongside consultancy and training where direct help is more appropriate. Teams can use these routes to move from a broad concern to specific work, ownership and review. [3]

Use it with judgement

The Toolkit holds a selection of judgments, not every potentially relevant decision. Law and guidance change. Its AI outputs need checking. The Toolkit is not legal advice; for a specific case, take qualified legal advice. A specialist evidence base can help organise the work; it does not guarantee an error-free answer or a fair outcome. [2]

The value I want colleagues to look for is practical: can you see what an answer relies on, understand its limits, consider who is affected and decide on a responsible next step?

Five questions before you act

Use these questions when reviewing an AI answer, a draft policy or a suggested response to a colleague.

1. Is it answering our actual question? Notice unnecessary legal or medical framing, but also check that genuinely relevant constraints have not been left out.
2. Can we verify the important claims? Follow the sources and check that they apply to this setting and date. Treat missing evidence as a reason to investigate, not fill in the gap.
3. Whose experience or concern is missing? Consider privacy, dignity and practical effects for everyone involved. Ask people rather than assuming how they feel.
4. What remains uncertain? Be clear about which facts, interpretations or decisions still need a competent human reviewer.
5. Who owns the next step? Record the action, responsible person and review point. Check what happened rather than assuming that issuing guidance solved the problem.

A useful next conversation

Choose one ordinary workplace question and one more complex policy scenario. Review the answers with colleagues using the five questions above. Discuss what helped, what was missing and what you would need before acting.

The aim is to build confidence through scrutiny, not to reward an answer simply because it agrees with us. That is how AI can contribute to better decisions and more positive people experiences.

Explore the Toolkit, its assessments and practical resources at thetransinclusiontoolkit.co.uk. For situations that need discussion or supported implementation, the site also sets out our consultancy and training options.

Joanne Lockwood

SEE Change Happen Ltd

Sources and further reading

[1] The Trans Inclusion Toolkit. Overview and assessments. https://thetransinclusiontoolkit.co.uk/

[2] The Trans Inclusion Toolkit. How we use case law, reviewed 31 October 2026. https://thetransinclusiontoolkit.co.uk/how-we-use-case-law/

[3] The Trans Inclusion Toolkit. Tools and resources. https://thetransinclusiontoolkit.co.uk/toolkit/

Sources checked 31 October 2026. This briefing supports discussion and responsible practice; it is not legal advice.