AI tools like ChatGPT and Copilot are now part of many workplaces. We’re seeing organisations use them to draft policies, create guidance documents, summarise information, and support learning design.

We’re also hearing a lot of understandable concern. Questions about ethics, environmental impact, privacy, bias, and misuse are not hypothetical. They matter. And they deserve serious attention when combining AI and EDI.

As an EDI training and consultancy organisation, our role isn’t to hype AI or dismiss the risks. Our role is to help organisations do people related work well, lawfully, and with integrity. Because whether an organisation chooses to use AI or not, the accountability for fairness, inclusion, accuracy and compliance still sits with humans.

This blog sets out a balanced, practical position: AI can be used as a support tool in some contexts, but it comes with real limitations and risks. If it is used in policy and training work, it needs clear boundaries and strong governance – AI and diversity and inclusion is still in its infancy ultimately.

Why this matters for EDI policies and guidance

EDI policies aren’t generic documents. They shape workplace culture, influence decision making, and affect whether people feel safe, respected and treated fairly. They also carry legal and reputational risk when they’re unclear, inaccurate, or misaligned with practice.

AI can produce a confident, professional looking draft quickly, but it cannot understand your workforce, your organisational context, your lived reality, or your responsibilities. It also cannot be relied on to produce consistently accurate or unbiased content.

That means AI outputs may:

• miss key context or nuance
• reproduce stereotypes or biased assumptions
• sound inclusive while embedding exclusionary norms
• give incorrect or outdated information
• create a false sense of completeness because the writing looks polished

Using it without care is a problem. However, using it with clear limits and proper reviews built in, it can help greatly. It can help teams get started, organise content, or improve readability.

A balanced view: benefits, concerns, and a realistic workplace approach

Some organisations find AI tools useful for early drafts, structure, and language clarity.

At the same time, many people (including within our own sector) raise serious concerns about the wider harms associated with AI, including environmental cost, the ethics of how some systems have been trained, and the misuse of generative tools. These concerns are legitimate and should form part of any organisational decision about whether and how AI is used.

So the point here isn’t “AI is good” or “AI is bad”. The point is: if AI is being used in your workplace, your safeguards need to match the level of risk, especially when the content relates to people, equality, and inclusion.

Where AI can be useful (with the right controls)

AI can be helpful for:

• creating an initial structure for a policy or guidance document
• generating plain English versions of existing content
• suggesting headings, FAQs, or implementation checklists
• drafting communications that then get edited for tone, accuracy and inclusion
• ideas generation for training activities that are then reviewed and improved by human experts

In these cases, AI should be treated as a drafting assistant, not an authority.

Where we recommend extreme caution (or avoiding AI entirely)

We strongly recommend avoiding AI use for:

• decisions about recruitment, performance, capability, disciplinary processes, redundancy or safeguarding
• content based on confidential casework or sensitive employee data
• any situation where AI output could directly influence someone’s access to support, opportunity, or fair treatment without robust human oversight

For AI and EDI specifically, we advise extra caution when drafting:

• equality and diversity policies
• dignity at work / anti-bullying policies
• reasonable adjustment guidance
• grievance, disciplinary or complaint-related documents
• inclusion guidance involving protected characteristics, trauma, or lived experience

These areas require accuracy, sensitivity and context. They cannot be “automated” safely.

A simple checklist for responsible AI use in policy and guidance writing

If your organisation uses AI at any stage of drafting policies or training content, ask:

1. What is AI doing here — is it drafting, summarising, advising, or influencing a decision?
   AI should support writing, not make judgements about people. AI and gender equality, for example, can be fraught with issues.
2. Are we entering any personal, confidential or sensitive information?
   If yes, don’t. Don’t paste in case details, employee information, or internal complaints.
3. Who is accountable for accuracy and sign off?
   Name the person responsible. Build in fact checking, legal checks where needed, and senior ownership.
4. Has this been reviewed through an inclusion lens?
   Check for stereotypes, assumptions, inaccessible language, and “one size fits all” wording.
5. Does the policy reflect how things actually work here?
   Generic policies fail in practice. Make sure procedures, responsibilities and reporting routes are real and usable.
6. Does it support trust?
   A policy is only effective if people believe it will be applied fairly and consistently.

The key message: people stay in the loop

In EDI work, the “human” part is not an add on. It’s essential. Inclusion depends on judgement, context, relational practice and accountability – All things which AI cannot provide.

If an AI drafted policy hasn’t been meaningfully reviewed, adapted, and owned by real people, it isn’t ready to be used, even as part of an AI and equality collaborative effort.

How we can help

Organisations often ask us, as EDI specialists, to support the development or review of EDI related policies and guidance. If your organisation is using AI to speed up drafting, we can help you make sure the final version is:

• accurate, clear and fit for purpose
• legally and ethically aware
• inclusive in language and intent
• implementable in day to day practice
• aligned with your organisational values and culture

The goal isn’t to produce more policies faster. It’s to produce policies that protect people, build trust, and improve workplace experience, something that AI and EDI can struggle with.

The takeaway

AI is topical, and it’s being used widely. It also raises serious concerns and can create real harm if used carelessly.

If your organisation chooses to use AI in policy or training work, do so with caution, clear boundaries, and strong human accountability.

Use AI for support, not authority. Keep people at the centre.