The pilot may have worked because someone at head office phoned every store on Mondays.
A retailer was rolling an AI stock-planning tool out to sixty stores.
The business case was strong. AI improves stock planning. The tool creates consistency. The pilot worked.
Underneath it sat a second model, which nobody had written down. Managers must use and trust the recommendations every Monday morning, even when local judgement feels faster and safer. Consistency depends on manager rituals and feedback loops.
And the pilot had a phone call in it.
The visible case was sound. The risk sat entirely in the invisible model. The verdict: test the Monday behaviour in six stores before scaling.
Rolled out as a technology, the investment becomes a dashboard. Rolled out as a weekly operating behaviour, it becomes the model.
Every business has two models. Most teams test one.
Shows how value is meant to move. Customers, channels, revenue, costs. It gets modelled, stress-tested and signed.
Shows what people have to do for any of it to happen. It gets a word in the business case and a nod in the meeting.
Alterkind maps the second one.
How the judgement works
Name the bet.
For this idea to work, [actor] must [behaviour] at [moment], despite [pressure].
Keep it filmable. If you can't point a camera at it, it isn't a behaviour yet.
Find the collision.
People are already doing something, and it makes sense to them. The old route is local logic. Nobody is being difficult.
The idea asks for a new behaviour while the current system rewards, protects and normalises the old one. Especially under pressure.
Weigh exposure against evidence.
What breaks if the behaviour doesn't happen. What proof exists that it will.
High exposure with weak evidence is the danger zone, and it's where most plans are sitting when they arrive here.
It ends in a decision
Proceed.
The behaviour is realistic. Go.
Change it.
The bet doesn't hold as it stands. Test it, simplify it, support it, redesign it, sequence it or localise it.
Stop.
The plan is asking for a behaviour that isn't going to happen.
Three answers. The middle one has six shapes, and we'll tell you which one you're in. The most valuable is stop.
Start with ninety minutes
The Review
Every behaviour the plan depends on, not just the one that worries you.
Priced against the decision.
Or run it yourself. The Canvas is free. Get the canvas
Then, if the answer is change it
Two days to name the behaviour. Most people arrive with four and believe they have one. Reframe tells you which, and it's what decides whether there's a week of work here or a year.
Then a sprint. One behaviour, five days. Find what's holding it in place, design against it, and by Friday know whether what your plan assumes about people holds up.