"What should I tell my bank about AI?" The honest starting point is that you may never be asked, and that a few owners will be asked at the worst moment, in the middle of a renewal, with a relationship manager waiting on a number. This article is for the owner of a company with ten to two hundred people and a bank line of credit or term loan. It explains why a lender may raise the subject, what the question usually means underneath, what to have ready on paper, what not to say, and the situations where it is wise to tell the bank first. It is about the reporting and controls side of the relationship. If you also want a written rulebook for staff, do I need an AI policy for my small business covers that part. If you are about to borrow to buy a building or expand, AI and an SBA loan: what to have on paper covers the application side.
I was COO and then President of a public company that closed more than forty acquisitions, and I ran the integration on most of them, so I have sat on the company side of a good many lender conversations. I am a lawyer by training. Now I write fixed-fee AI plans for CEOs of private-equity-backed and owner-led mid-market companies through 013 Labs. This is not legal advice, and it is not lending or tax advice. Where your loan documents or the law could apply, talk to your own counsel.
Why a lender may ask
I cannot tell you what your bank will do, and banks differ. These are the places where the subject can come up.
A credit review or renewal. A lender looks again at the company from time to time, and the file may carry questions about how management runs the business and produces its numbers. AI is one more thing that can change that.
Reporting questions. If you send financial statements, aging reports, inventory counts or a compliance certificate, the bank relies on them. A question such as "how are these prepared?" is ordinary, and a tool that helps prepare them is a fair thing to mention.
A borrowing base. If your line is tied to receivables or inventory, the bank cares a great deal about how those two figures are built, because they decide how much you can draw.
Fraud controls. Lenders and treasury teams worry about payments sent to the wrong place. AI makes convincing fake emails, invoices and even voices cheaper to produce, so a question about how your team confirms a change of bank details would not be odd.
What they are really asking
Under all four sits one pair of questions. Is the reporting reliable? And are the risks managed by someone? The bank is not examining your technology. It wants to know that the figures it lends against were not produced by something nobody checked, and that nothing new has quietly weakened your control over cash.
That is good news, because it means a short, concrete answer works better than a long, impressive one.
What to have ready
Four things, kept together in one folder, none of which needs more than a page or two.
One page on how you use AI and your data rule. List the tools staff actually use and the task each one does, including any that people signed up for on their own. Add one sentence on what must never go into an outside tool, for example customer lists, pricing, contracts or bank details. The article on whether AI makes a business worth more when you sell shows why the same page helps with a buyer later.
Who checks the numbers before they go to the bank. Name the person, and say what they check. If a tool helped prepare a report, the answer is still a human who compared the result with the ledger or the count before it left the building. If nobody does this today, that is the thing to fix before the question comes, not after.
Whether anything changed in how reports are produced. Write down the date and the change in plain words: which report, which step, what the tool does now that a person did before, and who confirmed that the two methods agreed. If nothing changed, write that too, with the date you checked.
Payment and bank-detail controls. This is where the fraud pattern bites. A familiar scam is a message that looks like it comes from a supplier or a senior colleague, asking you to pay to new bank details. Tools that write convincing messages make it easier to run. The control is dull and works: any change to payee bank details is confirmed by calling a number you already held on file, not one in the message, and by a second person, before the first payment goes. Write down who does it and how it is logged. Ask your bank what its own steps are for large or changed payments, and use them.
For the bookkeeping side, can AI do bookkeeping for a small business sets out where a person still has to check. For anything customer-facing, can AI answer customer emails for a small business covers the checking step. If the bank has asked for a business plan, can AI help me write a business plan covers what a tool may draft and what the numbers must come from.
What not to say
Do not overclaim. A banker who hears a big claim has to write it down, and you will be asked about it next year.
"The AI makes our numbers more accurate." Unless you compared the output with the ledger and kept the result, you do not know that. Say what you checked.
"We have fully automated our reporting." It tells the lender that no person is looking. Say who reviews.
"We don't use AI." Often untrue, because staff use free tools. If it turns out to be false, it harms trust more than the tool did.
Anything you cannot name a person for. If you say someone approves changes to payment details, be ready to say who.
Promises about the future. Do not forecast savings, headcount changes or results you have not measured. Say what you did and what you found.
Anything that sounds like you are quoting your loan terms from memory. Check the document, with counsel, before you describe it.
When to tell the bank first
Some changes are better raised before the bank finds them in a report. If a tool changes how inventory is counted or valued, how receivables are recorded, aged or reported, or how a borrowing base certificate or other required report is prepared, think about telling the bank before the next report goes out, and not after. The same goes for a change in who approves payments, or in how bank details are changed.
Whether you must tell the bank depends on your credit agreement: its reporting terms, any notice requirements and any definitions that cover how figures are calculated. I cannot tell you what yours says. Check it, with your own counsel, before the change goes live. If you decide to tell the bank, a short note works: what is changing, the date, who checked that the old and new methods agree, and who the bank can call. For a distributor, the piece on whether AI can forecast inventory for a small distributor shows why the inventory figure deserves this care.
Picture a building-supplies distributor of seventy people with a revolving line tied to receivables and stock. The finance manager has started using a tool to draft the monthly aging summary. Before the next report, the owner asks for three things: a note of the change with the date, a comparison of last month's summary done both ways, and the name of the person who will tick it against the ledger. The owner then checks the credit agreement with counsel, finds nothing that blocks the change, and sends the relationship manager two short paragraphs ahead of the report. At renewal the banker asks how reports are prepared, and the owner hands over one page and the folder. The conversation takes ten minutes. This is a hypothesis, a composite and not a client.
Mistakes that repeat
Waiting to be asked. Having no name against the check on the numbers. Changing how a report is produced and saying nothing. Overclaiming accuracy or automation. Letting a tool near payment details without a call-back rule. Describing the credit agreement from memory instead of checking it with counsel. Hiding free tools staff already use.
Frequently asked questions
What should I tell my bank about AI?
Be ready with a short, accurate account: which tools you use and for what, your data rule, who checks the numbers before they reach the bank, anything that changed in how reports are produced, and how payment and bank-detail changes are confirmed. Say only what you can back up with a name or a record.
Why would my bank ask about AI?
It may come up in a credit review or renewal, in questions about how reports are prepared, in a borrowing base discussion or in a fraud-control conversation. Underneath, the bank wants to know the reporting is reliable and the risks are managed.
Do I have to tell my bank when I start using AI?
It depends on your credit agreement and your situation. I cannot answer that for you. Check the reporting and notice terms with your own counsel, especially if a tool changes how inventory, receivables or a required report is produced.
What should I not say to my bank about AI?
Do not overclaim accuracy or automation, do not say you do not use it if staff do, and do not promise savings you have not measured. Say what you checked and who checked it.
How does AI affect payment fraud?
It can make fake messages, invoices and voices more convincing and cheaper to produce. The practical control is to confirm any change in payee bank details by calling a number you already held on file, with a second person involved, before paying.
Should I send my bank a document about this?
Not unless asked or unless a change calls for it. Keep the page and the folder ready, and send a short note ahead of a change that affects a required report.
Is this legal or lending advice?
No. It cannot be, because I do not know your loan documents. Ask your own counsel what applies to you.
Not sure where you stand? Take the 3-minute AI readiness scorecard.