The 9pm chat that could cost a credit union a member

A member in arrears opens your app at nine at night. They wouldn't ring in the day. But at nine, alone, they'll type. The assistant answers the first message. The second. On the third, it routes them to a generic FAQ. They give up. They stay in arrears another month. Nobody at the credit union ever knows the conversation happened.

It's the scenario our free AI Disruption Risk Index for UK credit unions uses to make its top risk concrete. The report calls it member-channel automation and scores it 89, the highest of any risk in the edition. Not the newest risk on your register. The highest. And the reason isn't the technology. It's who it fails.

 

A line on the register that AI quietly rewrote

 

On most registers this one has sat for years in a quiet, low-amplitude form. The wording is familiar: increased automation of the member channel introduces error, exclusion, or member disengagement, particularly for older or vulnerable members. A service concern. Reviewed annually, if that. AI didn't add a new line. It changed what the existing one means.

Three forces did it, at once. Expectation drifted: a member who banks with a challenger by day arrives at your app expecting it to answer at nine at night, because theirs does. Supply outran governance: the vendors behind your app and chat rolled AI into their roadmaps as a feature release, often with no disclosure to whoever owns risk, so the capability is live before it ever reaches an agenda. And AI is at its weakest exactly where you can least afford it, with the atypical member. The older one. The one in difficulty. The one whose income or name the model has seen less often. In a credit union, that is not a rounding error. That is the person you exist for. The member the assistant loses at nine at night is the member Consumer Duty most wants you to reach.

 

Point the same capability the other way

 

Here is what the score hides. The exact capability that drops that member is, turned around, the report's second-biggest opportunity: proactive member wellbeing, scored 84. Same data, same model, opposite outcome. The only variable is whether you are steering it or being steered by it.

The signals are already in the account. The salary that stops landing. The savings that start to drain. AI's only real trick is reading them at the scale of every member, every day, instead of whenever a member of staff happens to look. So the board question was never whether to automate the member channel. That decision has largely been made for you, upstream, by your vendors. It is whether you can see, override and evidence every AI tool standing between you and a member. Without those three, automation is something happening to your members. With them, it is something you are doing for them.

 

Control a 12-person team can actually hold

 

None of this needs a data-science function you don't have. The worked example in the report is a £28m credit union with twelve staff, the kind of team where risk is often one person's job, and only part of it. What good looks like there is unglamorous and entirely doable: a named owner for member-facing AI, a quarterly look at what each tool does and to whom, and that list in front of the board. That is the whole distance between catching the nine-o'clock failure and reading about it in a complaint three months later.

The hard part is staying current, because the tools change underneath you. The controls most credit unions still lean on, annual surveys, quarterly committees, a manual review, were built for a register that moved at human pace. The AI driver doesn't. A vendor pushes an update, the behaviour shifts, and last quarter's check is fiction. Drova RunSafe exists to hold that picture steady as the ground moves, so the AI in front of your members stays something you can see, override and defend at a board meeting. Not more oversight. Oversight that keeps up.

 

Why this is the one to get right

 

This is the top-scored risk in the credit union edition, and not because credit unions are behind. It is because the member channel is where the whole promise lives. A bank loses a nine-o'clock conversation and logs a metric. You lose a member who trusted you to be different, in the exact moment they needed you to be. Getting it right is not a technology project. It is the oldest job the movement has, in its newest form.

The full register, all four top risks scored against the objectives they threaten, with the opportunities sitting beside them, is worked through that anonymised £28m credit union in the UK edition. It is free.