An AI diligence statement is a short, honest note that says how AI was used in a piece of work, and confirms that a named person has checked it and takes responsibility for it. Think of it as the methods section for your work with AI: what the AI did, what you did, and how you checked the result. This article explains what to include and how to write one. It draws on Anthropic's Claude Academy tutorial on the subject, which belongs to its wider AI Fluency work, and the examples and advice are our own.
Why write one
- Trust. A specific statement shows you have thought about how you use AI. A vague or missing one is what damages trust, especially if someone finds the AI involvement out later.
- Expectations are changing. In some fields, such as finance, life sciences, and education, openness about AI use is increasingly expected, and sometimes required. Check what applies to your sector and your clients.
- A record of your role. It documents your checks, and makes clear who is accountable for the final result.
- Shared learning. When AI use is visible, a team can agree standards instead of everyone working it out privately.
The five things a good statement includes
| Element |
What to say |
| 1. What the AI did | Name the tasks: drafting, summarising research, analysing data, editing, formatting. A line such as "AI contributed to this work" tells the reader nothing. |
| 2. Which tool | Name the model or product. "Claude" is more useful than "an AI assistant". If you used more than one, say so. |
| 3. How you directed and checked it | What context you gave it, and how you verified the output: checking claims against primary sources, cross-checking figures, or having a colleague review. |
| 4. What you changed | Where you corrected, added to, or overrode the AI's work using your own expertise. This is where your judgement shows. |
| 5. Who is responsible | A clear statement that you stand behind the final output. This turns a disclosure into a commitment. |
An example
"I used Claude to produce a first draft of the CRM audit findings from the configuration export I supplied, and to suggest wording for the recommendations. I checked each finding against the live system, corrected three where the draft misread a field setting, and rewrote the priority order based on what I know of the client's sales process. I take responsibility for the content of this report."
Drafting one with AI
Any AI tool can help you draft a statement in about ten minutes, provided you bring real detail.
- Describe the project and the AI's part in it, task by task. The more specific you are, the better the result.
- Review the draft against what happened. Did it capture every task, and your actual checks? Add specifics, such as who else reviewed the work and what kind of errors you corrected.
- Adapt it for the audience. A full report might carry two or three paragraphs about your review process, while an executive summary or a slide might need two or three sentences. A regulatory submission needs the most.
- Check the statement itself. It is a claim about your process, so make sure it is true. If it says you verified every citation, you must have.
Do not copy a generic statement between projects. You can reuse the structure, but the content should describe what you actually did each time. A pasted boilerplate line says very little and can overstate your checks.
Making it normal in your organisation
Statements only work if people feel safe writing them. If disclosing AI use feels risky, people either stop using AI or stop being honest about it, and both outcomes are worse than open, responsible use.
- Build statements into templates and workflows, so they are not an afterthought.
- Have leaders include them on their own work, so disclosure looks like a professional habit, not an admission.
- Avoid policies that treat any AI use as suspect. The aim is responsible use, and responsibility needs honesty.
- If your organisation has no policy yet, start writing statements anyway. Your practice can shape the policy that follows.
Where this fits with AI and Zoho
This section is our own view. If an AI assistant has read or changed records in Zoho on your behalf, the checking habits in a diligence statement matter even more: ask for an audit trail, review what actually changed, and say who signed it off. Our guides cover the practical side: Connecting Claude to Zoho: A Safe Setup Checklist and Prompting Guide, and Using AI with Zoho: Why We Recommend Claude, and When Something Else Fits Better.
Need help? As a Zoho Partner serving the UK and Ireland, we can help you set up AI safely with your Zoho data, and agree the checking and disclosure habits that go with it. Email
zoho@1cloudconsultants.com or
raise a support request with 1 Cloud Consultants.