Should I tell my employer I used AI?

Short answer: check your organisation's policy first. If it doesn't settle the question, being open is usually the safer choice. How much you need to say depends on what the AI did and who is relying on the result.

This guide covers when disclosure matters most, how to mention it without sounding apologetic, how to use AI without looking lazy or losing your own skills, and what to do if you've been using a tool you weren't supposed to. It isn't legal or HR advice. Your organisation's policy and approved tools still take priority, and your manager or HR team can tell you how it applies to you.

Start with your policy

Many organisations now have a written AI policy, or rules about AI in a wider IT or data policy. Some say which tools are approved. Some require AI-assisted work to be labelled. Some ban AI for particular tasks, such as client advice or anything involving personal data.

Look on the intranet or in the staff handbook, or ask your manager. If the policy sets a rule on disclosure, follow it. If there's no policy, or it doesn't cover your situation, the rest of this guide will help you decide.

Why being open is usually the safer default

Accountability stays clear. When you send something, you're responsible for it, whoever drafted it. Saying "AI helped with this, I checked it" makes it clear you took that responsibility. If something turns out to be wrong, it's easier to trace why and fix it.

Trust lasts longer. Sensible, open use of AI is rarely what causes friction. What tends to unsettle people is finding out afterwards, especially if the work had an error or didn't sound like you. Being upfront avoids that.

Mistakes get caught. AI tools produce confident errors. If a reviewer knows AI was involved, they know to look harder at the specifics: figures, names, dates, quotes and sources. That extra look protects you.

It helps your team learn. When people are open about how they use AI, good practice spreads and risky practice gets spotted. One of the ten principles in the UK Government's AI Playbook is simply "You are open and collaborative". It was written with public-sector projects in mind, but the idea carries over to any team.

When disclosure clearly matters

Some situations call for openness even if nobody has asked.

Customer-facing work. If a customer would reasonably expect a message, report or proposal to come from a person's own judgement, or would be surprised to learn a chatbot wrote most of it, say so internally at least, and follow your organisation's rules on telling customers.

Your professional judgement. If someone relies on your work as your own expert view (a recommendation, an assessment, a technical opinion, advice), they're entitled to assume you did the thinking. AI can help you organise and draft. The judgement has to be yours, and if AI shaped the conclusion, say so.

Assessments, qualifications and apprenticeship evidence. Assessment rules may prohibit or restrict AI, and some require you to declare any use. Apprenticeship portfolios, coursework and assessed projects usually have to be your own work. Check your training provider's and assessment organisation's rules before you use AI at all. If you're an apprentice, the Evidence It course shows how to use AI as an interviewer that draws your experience out, so the writing stays yours. The lesson Be interviewed, not ghost-written is a good place to start.

Regulated work. Finance, legal, health, social care and other regulated fields often have specific rules on record-keeping, advice and accountability. If you work in one, check with your compliance or professional standards team before using AI for anything that forms part of regulated activity.

Anything that affects a decision about a person. Hiring, performance, discipline, eligibility. If AI helped summarise or organise information for one of these decisions, the decision-maker should know.

When it matters less

Not every use needs an announcement. Using AI to check your spelling, suggest a better word, or reword a routine internal email is closer to using a spellchecker. Unless your policy says otherwise, you probably don't need to mention it.

A useful test: if the person relying on this found out how it was made, would they feel misled? If yes, say something. If they'd shrug, you're probably fine.

How to say it well

Disclosure doesn't need to be long or apologetic. One plain line is usually enough. It works best when it says what AI did and what you did.

• "Drafted with AI, checked by me."
• "I used Copilot to summarise the three contracts; I've checked every figure against the originals."
• "First draft from ChatGPT on our business account. The recommendations are mine."
• "AI helped with structure. The analysis and conclusions are my own."

At Fernway Group, a fictional office equipment company used across this site, Sam Price, an account manager, uses the company's approved Copilot to draft a proposal for a new customer. He sends it to Tom Elliott, the sales lead, with a note: "First draft from Copilot using our standard proposal and the customer's brief. I've rewritten the pricing section and checked the product specs against the catalogue." Tom knows exactly where to look, and trusts the rest.

Avoid lines that shift responsibility away from you, like "the AI says" or "according to ChatGPT". You're the one sending it.

Using AI without looking lazy

People worry that admitting AI use makes them look as if they cut corners. What usually creates that impression is output that is generic, padded or wrong. Good AI-assisted work looks like good work.

• Keep the thinking yours. Decide what you want to say before asking AI to help say it. Give it your points, your structure, your conclusion.
• Use it where it's strongest. First drafts, restructuring, summarising material you've read, checking for gaps, suggesting a clearer way to put something.
• Edit properly. Cut the filler, fix the tone, add the specific detail only you know. If a paragraph could have been written about any company, it needs more from you.
• Check before you send. Every figure, name, date and claim. The free Prove It course teaches a quick routine.

Using AI without becoming dependent on it

There's a quieter risk: letting skills you need fade because AI does them for you. If you always let AI draft difficult emails, you may find it harder to write one yourself when the tool isn't available, or when the situation needs a personal touch.

• Practise the core skills of your job yourself, regularly, without AI.
• Write the first version yourself sometimes, then ask AI to suggest improvements. You'll learn more than by editing its draft.
• Notice when you can't explain the output. If you couldn't defend a piece of work without the AI's help, you don't understand it well enough to send it.
• Treat AI as a colleague to check with, not a replacement for knowing your job.

If you've been using a tool that wasn't approved

Plenty of people have used a personal AI account for work before realising it wasn't allowed. If that's you, here's what to do.

Stop using it for work. Switch to your organisation's approved tool, or ask whether one exists.

Tell someone. Your manager is usually the right first person. Be factual: which tool, roughly how long, and what kind of work.

Flag anything sensitive straight away. If you put in personal data, customer information or anything confidential, tell your data protection contact as well, the same day. Your organisation may have to report a personal data breach to the ICO no later than 72 hours after it becomes aware of it (the organisation decides whether a breach is notifiable, not you), as the ICO's guide to personal data breaches explains. Speed matters more than blame.

Don't try to cover it up. Deleting chats may not remove every copy the provider holds, and it makes the conversation harder later.

It is almost always better for your manager to hear it from you than to find out another way. If your team doesn't have clear AI rules yet, your honesty might be what prompts someone to write them. Managers can find help on handling this well in the guides for managers.