An audit that brings foresight and behavioural science to your metrics. You find where users lose trust, lose agency or get confused, and leave with a blueprint your team can act on.
Your metrics describe what people did. They rarely say why, or what is about to change. The audit works on the signal behind them, with three questions you can ask of any model you run.
Intent gap
Nest knew the what. It optimised energy, but missed life changes like a newborn. This is a proxy problem. You don’t fix it with more data, you fix it with better signal.
Foresight gap
That is the drift you never see in a log. The audit tests the model against a few different futures and edge cases, and asks whether it can recover. This is where designers and data scientists work best together.
Agency gap
If they can’t, you lose your best source of truth: the people living the reality your model is trying to predict. No override, no correction.
A structured walkthrough of your product, mapped against the moments where users lose trust, lose agency or get confused by what the system does.
Not a slide deck of observations. A ranked set of concrete changes that your product and development teams can both act on, each tied to the moment that needs it.
Best for product and development teams with an AI feature in production, or close to launching one.
Who is behind it
Ten years designing AI solutions and advising businesses on strategy, backed by a PhD.
The method is in my book, The Anticipatory Design Playbook, and in the free tools you can try before you book anything.