Every bank board I have spoken with is asking the same question about AI: “Are we keeping up?”
This is the wrong question. The gap between what boards are asking and what they should be asking is where the next category of technology failures will arise.
THE RIGHT QUESTION IS: WHAT BREAKS WHEN THE AI IS WRONG?
Not if. When.
Every AI system produces errors. The question is not whether your fraud model will misfire, your credit decisioning will produce a disparate outcome, or your customer-facing AI will generate a response that creates regulatory exposure. It will. The question is what happens next.
Most banks cannot answer that clearly. They have invested in model development and model governance, but they have not invested in model failure architecture.
WHAT MODEL FAILURE ARCHITECTURE ACTUALLY MEANS:
- Knowing in advance which AI-driven decisions are reversible and which are not.
- Having a human escalation path that works at the speed the AI operates.
- Understanding which failures produce regulatory exposure versus operational disruption versus reputational damage.
- Testing failure modes before they appear in production, not after.
This is not a technology problem. It is a governance problem. And it belongs on the board agenda — not buried in a model risk report that three people read.
THE BOARDS GETTING THIS RIGHT ARE ASKING DIFFERENT QUESTIONS:
Not “are we using AI responsibly?” — that is a compliance question.
“What is our most consequential AI decision right now, and what does failure look like?” That is a governance question.
The difference matters more than most boards currently understand.
Interested to know, what is your board asking?