Every case changes two experiences. read both before you build.
Each AI case does something to your customers and something to your people. The method reads both, per case, before a line of code exists: because half the expensive mistakes in enterprise AI are cases that worked technically and failed on one of these two fronts.
The customer read
Every case is classified by the kind of value it gives customers: table stakes the market now expects, performance that does existing work better, a genuine differentiator, or, the category nobody budgets for, value that subtracts. The Harborne readout carries a live example: an automated settlement case that scores well on cost and speed while removing the human contact at the exact moment customers value it most. The read asks that question before the build rather than in the renewal data two years after.
Differentiators carry a clock. What wins preference this year reads as expected next year: delight decays into table stakes, and the read times that decay so you build difference while it is still different, and buy parity as parity.
The employee read
The same case is read for what it does to the people doing the work. Some cases absorb toil and hand judgement back: those raise the job. Some remove the judgement, or the work through which juniors learn it: those hollow the role and, years later, the pipeline into your senior positions. The read flags both, per case, and where automation takes the learning, delivery replaces it by design, structured supervision inside live work, before the automation ships.
The blocker is the system people work in, and it is fixable.
Twenty years of change management was built on one assumption: that people resist change, and ACM manages that resistance. The data disproves it. In most organisations your people are already ahead of the system around them: skills built, use cases found, ways of working changed. These aren’t resistant employees. These are employees waiting for the organisation to catch up.
Diagnose the system, not the people.
If skills are already in place but value isn’t compounding, the gap is structural: incentives, decision rights, manager behaviour, what gets recognised in performance reviews. Organisational factors drive 2x the AI impact of individual mindset and behaviour.
Managers are the unlock.
When managers visibly use AI themselves, employees report 17 points more value from it. Most ACM programmes treat managers as a comms channel. We treat them as the intervention.
Redesign what gets rewarded.
Only 13% of AI users say they’re rewarded for reinventing work with AI when results aren’t immediately there. People won’t change what they do until you change what you reward. The hardest part isn’t the comms or the training: it’s redesigning performance management, decision rights and recognition.
This work runs inside the engagements.
In Decide, both reads run per case and the warnings land in the Dream Book with everything else. In De-risk we equip the managers and pilot the new behaviours. In Scale, change is embedded as each capability is built, not bolted on after: and where a case took the learning, the preceptorship pattern puts it back.
If we recommend training, it’s because the diagnosis showed training was the answer. Usually it isn’t. Most ACM teams sell training. We sell system redesign.
Adoption and change management used to manage resistance. This version removes the reasons for it.