Add a short, plain statement to your proposal saying which parts of the work AI assists with and that a qualified person reviews everything, then back it up in your terms of business with clauses on confidentiality and approved tools, human review, ownership of deliverables and liability. Have a solicitor check the final wording before you rely on it.
The proposal and the terms do different jobs. The proposal sets expectations in words a client will actually read; the terms decide what happens if something goes wrong. Below is sample wording for both, matched to how much AI touches the client's work, plus the traps that catch small firms: promising more than you can guarantee, clashing with the client's own procurement terms, and letting the wording go stale as tools change. Treat the samples as a starting point for your adviser, not as legal advice.
First, work out how much AI actually touches client work
Not every use of AI needs a mention. The more directly AI shapes what the client receives, or the more of their information goes into an AI tool, the more explicit you should be. Place each of your uses on this scale:
| Level | Examples | Where to mention it |
|---|---|---|
| Internal only | Your own emails, scheduling, invoicing, research on public information | Internal AI policy; no client mention needed |
| Drafting support | First drafts of reports, minutes or specifications, finalised by a professional | Proposal statement plus a human-review clause |
| Client information processed | Uploading the client's drawings, documents or personal data to an AI tool | Confidentiality and approved-tools clause, plus your data processing terms |
| AI output in the deliverable | AI-generated images, visualisations or text that the client receives as part of the work | Explicit disclosure plus an ownership clause |
| AI talking to the client | A chatbot on your site or AI replies to client messages | Clear disclosure at the point of use, not just in the contract |
If all your use sits in the first row, you may need nothing in client documents. Most professional firms find they have something in rows two and three. For the broader question of whether to tell customers at all, see should you tell customers you use AI?
The proposal statement: three versions
Put this in its own short section, headed something like "How we work" or "Use of technology", rather than burying it in the small print. Pick the version that matches your level from the table.
Light, for drafting support only:
We use AI-assisted software for some drafting and administrative tasks. All work we send you is prepared, checked and approved by a member of our professional team, who remains responsible for it.
Standard, when client information goes into AI tools:
We use a small number of approved AI tools to help draft documents, summarise meetings and check our work. We use only business versions of these tools that do not use your information to train their models, and we don't enter your information into free or personal AI services. Every deliverable is reviewed and approved by a qualified member of our team. If you'd prefer us not to use AI tools on your project, or on particular documents, tell us and we'll agree how to work.
Detailed, when AI output forms part of the deliverable:
Some early-stage visuals in this proposal and in the concept stage are produced with AI image tools and are marked as such. They illustrate mood and direction only and are not design drawings. All technical drawings, calculations and specifications are prepared by our team without relying on AI output for any design or compliance decision.
Notice what none of these say: "we never make mistakes with AI", "AI makes us faster so you'll pay less", or a list of product names. Each of those creates a promise you'll struggle to keep.
The same wording carries over to the moment a client asks directly, often by email halfway through a job. For an illustrative market-research consultancy using the standard statement, a reply might read:
Thanks for asking. Yes, we use two approved AI tools: one produces first-draft transcripts and summaries of the customer interviews, and the other helps draft the findings section of your report. Both are business versions that don't use your information to train their models, and neither decides what the findings are. Every report is checked and signed off by your project lead before it reaches you. If you'd rather we didn't use them on your work, tell me and we'll agree how to handle it. The full list of tools is available if you'd like it.
It answers yes in the first line, says what each tool does in the client's terms, repeats the review commitment and offers the opt-out. A vague answer ("we use various modern tools") invites the follow-up question you were trying to avoid.
Before a statement goes into your proposal template, test it against what the team actually does. An AI assistant on a business plan makes a quick second reader:
Below is the AI-use statement from our proposals, then a list of
every way our team uses AI tools, gathered from each person this
week. Quote every sentence in the statement that our actual use
contradicts or goes beyond, and say which use breaks it.
STATEMENT: [paste]
ACTUAL USE: [paste the list]For an illustrative landscape design studio using the "standard" statement, the reply flagged one clash: "'We don't enter your information into free or personal AI services' is contradicted by 'junior designer uses a free online upscaler to sharpen clients' site photos'." Site photos show a client's house and garden, so the studio had two honest options: move the upscaling to an approved tool, or drop the sentence. It moved the task. The check took ten minutes and found the one line a client could have caught them out on.
Clauses for your terms of business
The terms carry the commitments. These six cover most situations. Adjust them to your practice and have them reviewed; the wording is illustrative.
1. Use of AI and professional responsibility
This says AI use doesn't change who's responsible or the standard of care.
We may use software tools, including artificial intelligence tools, in providing the services. Our use of such tools does not reduce our responsibility for the services, which are subject to the same standard of care as the rest of this agreement.
2. Confidentiality and approved tools
We will only enter your confidential information into AI tools that we have approved for business use, that are bound by confidentiality and data-processing terms, and that do not use your information to train their models. A list of approved tools is available on request.
"Available on request" matters. Tools change mid-project; a list inside the contract goes out of date, a list you maintain doesn't. Here's what one looks like filled in, for an illustrative five-person HR consultancy:
| Tool and plan | Approved for | Trains on our content? | Approved by, date |
|---|---|---|---|
| ChatGPT Business | Drafting policies and reports; summarising client documents | No, by default on this plan | Director, 3 Mar |
| Gemini in Google Workspace (Business Standard) | Email drafts; meeting notes in Meet | No, by default on business plans | Director, 3 Mar |
| Transcription tool, business plan | Recording interviews with the interviewee's agreement | No, per the vendor's terms (checked 14 Jul) | Operations lead, 14 Jul |
| Not approved | Any free or personal AI account for client material | n/a |
The last row is as useful as the first three. It's the line a client's procurement team will look for, and the one staff need when they're tempted by a free tool on a deadline.
3. Personal data
If client documents contain personal data, your AI tools act as processors on your behalf, so data-protection law governs how they handle it. Point to your existing data processing terms rather than rewriting them here, and check your AI vendors' agreements match; what to check in an AI vendor's data processing agreement lists the terms to look for.
4. Ownership of deliverables
This is where firms over-promise. Whether purely AI-generated material can be protected by copyright is unsettled and differs between legal systems, so don't warrant that the client will own rights you may not hold. Grant what you can and say so honestly. Who owns AI-generated content? explains the uncertainty in plain terms.
On payment, we grant you [ownership of / a licence to use] the deliverables as set out in clause X. Where a deliverable includes material generated with the help of AI tools, we grant you such rights as we hold in that material, and we will tell you on request which elements were AI-generated.
5. Client restrictions on AI use
You may ask us not to use AI tools on your project or on specified material. We will confirm any restriction in writing. Restrictions may affect the programme or the fee, and we will tell you before proceeding if they do.
6. Material you supply
Clients increasingly send AI-written briefs, specifications or images. This clause keeps responsibility for them where it belongs.
You are responsible for the accuracy of information you provide and for having the right to let us use it, including any material you or your advisers created with AI tools.
When the client's own contract sets the rules
Larger clients often send their own terms, and those usually override yours. Read them for three things before you sign: a ban or consent requirement for generative AI, a duty to disclose the tools you use, and confidentiality wording strict enough to rule out any third-party software. If a client's terms forbid AI use without consent and you've already built it into how you work, raise it before signing rather than hoping nobody asks. A one-line email confirming "we use the following approved tools under the standards in our proposal; please confirm this is acceptable" turns a risk into a record.
A long set of supplier terms is a reasonable job for an AI first pass, as long as a person reads the result against the document. On a business plan, paste the terms with a narrow question:
Below are a client's supplier terms. List every clause that could
restrict us using AI tools on their work: bans or consent requirements
for AI or software, confidentiality limits on third parties,
sub-processor or subcontracting rules, and data-location rules. Quote
each clause with its number. Check the definitions section too, and
say if a defined term changes what a clause covers.
[paste terms]For an illustrative translation agency reviewing a retailer's terms, the reply might list clause 9.2 ("no subcontracting of the Services without prior written consent") and clause 14.1 ("Supplier shall not disclose Confidential Information to any third party"), and conclude that AI tools aren't mentioned. The agency's own read of the schedule found what the summary skipped: "Sub-processor" was defined as "any third party, including any software or cloud service provider, that processes Client Data". Read with that definition, uploading source documents to an AI translation tool needed consent. That's the pattern to expect. The quoted clauses are usually accurate, and the conclusion drawn from them is where it slips. Treat the output as a map of where to read, and put anything that matters in front of your solicitor.
Here's how skipping that step tends to surface. An illustrative four-person payroll bureau signed a manufacturing client's supplier terms, which said no third party could process the client's confidential information without written consent. A year later the bureau had started recording client calls with an AI transcription tool. Nobody connected the two until the client's annual supplier questionnaire asked for "all sub-processors handling our employees' data", and the honest answer included a tool the client had never approved. The bureau disclosed it, asked for consent, and got it with one condition: no recordings of calls where individual salaries were discussed. It was an awkward week that a two-line email at the start would have avoided.
Check your professional indemnity insurance at the same time. Ask your broker whether AI-assisted work is covered on the same basis as other work and whether the insurer expects particular controls, such as human review. Does your business insurance cover AI mistakes? has the questions to put to them.
Worked example: an architect practice's fee proposal
Take an illustrative seven-person architecture practice. It uses AI to produce minutes from recorded design meetings (with attendees' agreement), draft first versions of design and access statements, check specifications against an internal checklist, and generate early mood images at concept stage. It does not use AI for any structural, fire or regulatory compliance decision.
That puts it in rows two, three and four of the table, so its proposals now carry the "standard" statement plus the "detailed" paragraph for concept images, and its terms include all six clauses. Two practical additions came out of the exercise. The meeting invitation now says "this meeting will be recorded and summarised with an AI tool; tell us if you'd rather it wasn't", because recording consent belongs at the meeting, not in a contract signed months earlier. And every AI mood image is captioned "AI-generated illustration" in the concept report, so nobody mistakes it for a design.
In the first quarter, one client, a developer with strict confidentiality terms, asked that their site survey never be uploaded to any AI tool. The practice agreed, noted it on the project sheet, and added about two hours to that stage's fee for manual specification checks, explained in advance under clause 5. Another client asked for a discount because "AI does the writing now"; the firm explained that the drafting saving was already reflected in its fixed stage fees. The architect-specific workflows behind these uses are covered in how architects use AI for fee proposals and briefs.
Keeping the wording true as your tools change
Contract wording about AI is only protective while it matches what you actually do. Three habits keep it honest:
- Keep an approved-tools list with an owner. The contract points to it, so it must exist and be current. Note for each tool its plan, whether it trains on your content, and who approved it.
- Avoid absolutes. "We never", "all" and "guarantee" fail the first time someone makes an ordinary mistake. "We only use approved tools" and "every deliverable is reviewed" are commitments you can meet and prove.
- Review once a year, and whenever you add a new kind of AI use. Moving from drafting support to AI-generated visuals is a new row in the table, and it needs new wording before the first client sees one.
Absolutes creep in when someone edits the wording in a hurry. Three before-and-after lines, all illustrative:
| Promise you can't keep | Commitment you can prove |
|---|---|
| "All AI output is checked for accuracy." | "Every deliverable is reviewed and approved by a named member of our team before we send it." |
| "Your data never leaves our systems." | "We only enter your information into the approved business tools on our list, available on request." |
| "We don't use AI on your documents." | "We don't use generative AI to draft or analyse your documents unless you agree." (Spellcheck and search in your office software may use AI, so an outright "no AI" is rarely true.) |
If you work in marketing, design or content, the angle is slightly different because AI output is often the product itself; AI clauses for agency contracts covers approvals and ownership for that kind of work in more depth.
What clients and owners ask about AI wording
Am I legally required to tell clients I use AI?
It depends on your contracts, your sector and where your clients are. A client's terms may require consent, professional bodies may set disclosure rules, data-protection law applies when personal data goes into AI tools, and if you sell to customers in the EU the EU AI Act requires people to know when they're dealing with a chatbot. Outside those, disclosure is often a trust decision rather than a legal duty. Check with a solicitor for your situation.
What if a client says no AI at all on their project?
Agree it in writing and make it workable. Ask whether the restriction covers everything or only their documents and data, since internal admin rarely matters to them. Record the restriction on the project file so every team member sees it, and consider whether the extra time changes your fee. Your terms can say that such restrictions may affect the fee and programme, so the conversation isn't awkward later.
Should I list the specific AI tools I use in the contract?
Usually not in the contract itself, because tools change faster than contracts do. Refer instead to a list of approved tools you keep up to date and provide on request, and commit to the standards those tools must meet, such as business plans that don't train on client content. That keeps the contract accurate when you switch from one assistant to another mid-project.
Further reads
- How to Write an AI Disclosure Statement for Your Website — Write the public-facing version for your website.
- Should You Charge Clients Less When AI Speeds Up Your Work? — Handle the fee question clients raise once they know.
- Should Freelancers Tell Clients They Use AI? — The same decision from a sole practitioner's side.
- How to Write an AI Usage Policy for Your Small Business — The internal policy your contract wording should match.
- How to Catch Made-Up Figures in AI-Drafted Proposals — Stop invented figures slipping into proposals.
- Can AI Review a Contract? What Small Business Owners Should Know — What AI can and can't spot in contracts you receive.
- Barriers to AI Adoption in Small Businesses and How to Clear Them — Eight practical barriers that keep small firms from adopting AI, the cheapest way past each, and the few barriers you should respect rather than clear.
- Does a Five-Person Business Really Need an AI Policy? — Why a five-person business needs a one-page AI policy rather than a handbook: six triggers, a complete template, and when one page stops being enough.
- AI Itineraries: What a Travel Agent Must Check Before Sending — A seven-part checklist for AI-drafted itineraries, with why each check matters, how to verify it, the red flags of an unchecked draft, and a sign-off record.
- How to Write Event Proposals With AI in Half the Time — Where the hours in an event proposal really go, the kit to build once, four prompts to run in order, and a timed worked example.
- AI Readiness Checklist for Accountants, Solicitors, Consultants — Twenty checks, grouped and scored, that tell an accountancy, law or consulting firm whether it is ready to pilot AI or has gaps to fix first.
- AI Engagement Letters for Accountants: Faster Drafts, Clear Scope — Which parts of an engagement letter AI should draft, five prompts with sample outputs, and the partner checks that keep scope clear and terms untouched.
- How Consultants Use AI to Write Proposals in Under an Hour — A proposal kit you build once and a timed five-block hour: brief, draft from the kit, price it yourself, let AI object as the buyer, then check.
- AI Acceptable Use Policy for a Small Professional Firm — A clause-by-clause checklist and one-page sample policy for accountants, lawyers and consultants who handle confidential client work.
- Reusing Past Client Work With AI Without Leaking Client Data — How consultants and small firms turn old client reports into a reusable AI library: contract checks, three reuse tiers, sanitising, and access controls.
- AI Proposal Software vs a General AI Assistant: Which to Pay For — What proposal tools add beyond the writing, what they cost per seat, and a question-by-question table for choosing between them and a chat assistant.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: EU AI Act transparency obligations (Article 50), official FAQ on the EU digital-strategy site. This tutorial is not legal advice.