What should our firm say about AI in a client tender or panel questionnaire?

Answer plainly and specifically: name the tools your firm has approved, describe the tasks they are used for, state who reviews the output before it reaches the client, and set out where client information is stored and processed. A general reassurance that you "use AI responsibly" will not satisfy a procurement team, and an overstated answer creates a promise you then have to keep. The safest questionnaire response is a short, factual description of your actual practice.

What these questionnaires are usually trying to establish

Client procurement teams, in-house legal departments and panel managers are not testing your enthusiasm for technology. They are checking four things.

• Confidentiality. Does their material leave your control, and if so, to whom, and under what terms?
• Data location and retention. Where is the data held, how long for, and is it used to train anything?
• Accountability. Who is answerable for the work product, and how is it checked?
• Consistency. Do your answers match your engagement terms, your privacy notice and what your fee earners actually do?

Some questionnaires also ask whether AI has reduced your costs, and whether that saving is passed on. Answer that honestly. If your pricing has not changed, say so and explain how you price. A vague claim of efficiency invites a fee negotiation you have not prepared for.

Get your facts straight before you draft the answer

You cannot answer a data question about a tool you have not checked. Before anyone writes a word, gather the following from your suppliers and, if necessary, in writing.

1. The full list of AI tools in use across the firm, including anything embedded in software you already licence. Office suites, dictation tools, document comparison tools and case management systems increasingly include AI features.
2. Whether each tool is on your approved list, and what happens to tools that are not.
3. Where each supplier holds and processes data, including any subprocessors.
4. Whether your content is used to train models for other customers.
5. Retention periods and deletion arrangements, including what happens at the end of the contract.
6. Access controls: who inside your firm can see a given matter, and how identity is verified.

If the answer to any of these is "we are not sure", that is the finding, not a gap to be papered over. Resolve it before the questionnaire goes back.

Answering the common questions well

"Do you use AI on client matters?" Say yes or no, then qualify. For example: yes, for document review, chronologies, first drafts and research support; not for advice to the client, which is settled by the responsible fee earner.

"Which tools?" Name them. Firms sometimes resist this, but a client asking about confidentiality is entitled to know who holds their papers. If you would name your document management supplier, name your AI supplier.

"Is our data used to train models?" Answer from the contract, not from memory. If the position differs between tools, say so tool by tool.

"Who reviews the output?" Describe the supervision arrangement you actually operate: who checks, against what, and how it is recorded on the file. Reference your file note practice if you have one.

"What is your AI policy?" Attach it. If it is three pages and readable, that is better than ten pages of principles. If you do not have one, write one before you answer.

"How do you handle errors?" Explain that AI output is treated as unverified until checked against the source, and describe how a problem is escalated and recorded.

Keep a maintained set of answers in one place, owned by one person, so that two partners responding to two clients in the same month do not give different accounts of the same firm.

What not to say

Avoid these, in order of the trouble they cause.

• Claims you cannot evidence: "fully compliant", "bank-grade security", "human oversight of every output" if that is not literally true.
• Rule and article numbers you have not checked. Refer to the Solicitors Regulation Authority's expectations of competence and confidentiality, and to UK GDPR and the Information Commissioner's Office, in general terms.
• Promises about accuracy. No supplier can guarantee that output is correct, and you should not adopt a guarantee on their behalf.
• Silence about tools you would rather not mention. If a fee earner has been pasting client material into a consumer chatbot, the questionnaire is the moment to find out, not the moment to conceal it.
• Blanket denials. "We do not use AI" is increasingly untrue for most firms once embedded features are counted, and an untrue answer to a client is a serious problem.

If a client's requirements go further than your current position, say what you do now and what you would agree to for their matters, such as restricting a particular tool or requiring named-partner review.

Where Alesis fits

Alesis is an AI assistant for UK law firms, made by L25 Limited and used through the web browser, with one conversation for a matter. If you are completing a questionnaire, the answers you need about us are these: the firm's information is held in the UK and processed only in the UK and the EU, kept apart from every other firm, and never used to train anything for anyone else, and inside a firm people see only the matters they are on, with seniority alone granting no view. Every account proves its email address, and can add two-factor authentication and passkeys. Alesis prepares drafts for a qualified person to review and sign off, does not file, serve or email anything, and does not provide legal advice.