Five questions to ask a vendor about their AI feature.
What it reads, where it runs, who reviews it and how you switch it off. Ask before you turn it on; your examiner will ask the same things.
Most core, lending and service vendors now ship an AI feature, and it often arrives switched on by default. Before your staff start relying on it, get five answers in writing. They should fit on one page.
Five questions before you switch it on
- {{ q.n }}{{ q.t }}
"If the answer needs a sales call, it isn't an answer yet."
What good answers look like
We hold our own work to the same rules. Your data stays in your workspace. Member data is checked before any AI reads it, and uploads with social security numbers, account numbers or credentials are blocked. The assistant proposes; a person decides. Nobody confirms their own work, and every action leaves a record you can hand to an examiner.
A vendor does not need to match that exactly. It does need to tell you, in plain words, where its feature sits against each rule.
Write the answers down
Keep the five answers next to the feature, with a named owner on your side. That list is the first step of the routine in our whitepaper on responsible AI use, and it is the page an examiner will ask for.