The people closest to the work belong in the design of the workflow.
We worry about the colleague who quietly keeps everything working and then hears that AI is going to make the process easy. They know the exception that never made it into the manual. They know which answer will trigger a difficult customer conversation. If we arrive without taking that knowledge seriously, we have misunderstood the job before touching the technology.
Our approach to capability transfer starts with respect for what your team already knows. We want the person closest to the process to challenge the first version. Give them permission to choose the awkward example and explain why an answer would not be usable. Their objections can be the most valuable design input in the room.
Invite the people who know
That is an opportunity for collaboration. Ask the person closest to the process to explain where it becomes difficult. Invite them to choose an example that would expose a weak answer. Let them define the difference between a plausible output and a useful one. Their role becomes clearer when their judgment shapes the test.
Training can then focus on the work itself. A general introduction to AI may help establish language, but confidence grows through relevant practice. Someone needs to see how to prepare an input, interpret an exception, challenge an explanation and recover when the output is unsatisfactory. These are practical skills that can be developed through a familiar task.
Our approach to capability transfer starts with respect for what your team already knows.
Consider an illustrative operations manager who is asked to use a new cash dashboard. Showing the final chart explains very little about how to maintain it. Working through a changed collection assumption together can reveal which field matters, what the chart is saying and when finance needs to be involved. The person learns the relationship between the tool and their responsibility.
Build confidence through use
An AI Champion needs more than enthusiasm. We would look for curiosity, patience and the confidence to say that an answer needs checking. They also need protected time, a backup and a decision-maker who listens. Otherwise the new role becomes another job added to the person who was already holding the process together.
It is also important to make disagreement safe. If a person believes an AI output is wrong, the process should give them a practical way to challenge it. The aim is thoughtful use, not unquestioning adoption. A team that can identify and explain a failure is developing useful capability.
Leaders influence this culture through the questions they ask. Asking only how much time AI saved can encourage people to hide the effort required to correct it. Asking what was learned, where review was difficult and what changed in the process gives a more complete picture. It also produces better information for the next improvement.
Our capability-transfer ambition depends on this human part of the work. We want people to understand enough to carry the process forward and recognise when it needs attention. That requires explanation, practice and a design that respects the knowledge already in the business.
Who in your team would you trust to challenge a confident answer and help a colleague learn from it? What would they need from you to do that well? We see those questions as part of implementation. The aim is a team that can keep going when we leave.
Examples in this essay are illustrative, rather than accounts of client engagements.