Trust in an AI world: what customers forgive

The tolerance line is clearer than most banks expect. Customers accept automation at the front of a conversation, provided it is labelled and provided there is a route past it. What they do not accept is discovering after the fact that no person was ever involved.

The second thing they forgive is a wrong answer, if it is corrected quickly. The thing they do not forgive is a confident wrong answer that nobody in the organisation is accountable for. That distinction should shape where AI sits in your service design.

For marketing teams the practical rule is narrow: use the tools for volume, speed and translation, keep a named human on judgement, and never let a model speak on risk, pricing or complaint without review.