We pitched an AI complaint-monitoring product to two bank compliance leaders in the same week. Both rejected it within minutes. Two calls is a small sample, so treat this as a pattern to test, not a law of the market.
The first leader said their bank would not use AI in that area. That was the whole objection. There was no question about price, no request for a case study, and no interest in a pilot. The second leader was more polite but reached the same place. They pointed to Copilot, which the bank already licensed, and to a complaints platform they had bought the year before. For a team handling low complaint volumes, they saw no reason to add another tool.
Looking back, the problem was our opening, not the product. We led with AI. In a bank compliance function, that word signals risk before it signals value. Compliance leaders are measured on whether they can show a regulator a controlled, repeatable process. When they hear “AI,” they think of model behaviour they cannot explain, customer data leaving the perimeter, and a decision they would struggle to defend in an audit.
So we changed the order. The first sentence now names the control problem: complaints logged late, categorised inconsistently, or missed entirely, and the regulatory exposure that follows from each. We describe how every complaint is captured, how each classification traces back to a source record, and where a named person signs off. The technology comes after that, if it comes at all.
Some practical points from this:
Name the regulatory exposure in the first sentence, not the third.
Describe the audit trail before the capability. Who reviewed what, and when.
Ask about current complaint volume before you pitch. If volume is low, an existing platform may be enough, and saying so earns more trust than arguing with it.
Expect the word “Copilot.” Have a clear answer on where your tool sits relative to what they already license.
Don’t defend the label. If someone says no to AI, ask what they would need to see before a control like this could pass review.
Two calls do not prove that compliance buyers reject AI. They do show that a label can end a conversation before the value gets heard. Next, we are testing the control-first opening on a wider list. We will report the results, including the ones that do not go our way.