AI shouldn’t replace your best people. It should change what they spend their time doing.
Design around the value of human expertise, and examine the work that keeps it occupied.
Expertise is more than task completion.
Experienced people bring context that is difficult to reduce to a process diagram. They notice an unusual detail, understand a relationship and recognise when the obvious answer is incomplete. A useful AI initiative begins by understanding where that judgement matters.
Then look at everything surrounding it. Finding information, preparing a first draft, reconciling records and collecting missing details may take up attention that could be spent on interpretation, advice or a better conversation.
Redesign the whole task.
Automating one step does not automatically improve the experience. If the result needs extensive checking or arrives in the wrong place, the effort may simply move from preparation to review. The whole task needs to be considered, including what happens when the system is uncertain.
Involve the people doing the work from the beginning. Ask what would make a prepared result useful, what they would need to trust it and which decisions they want to retain. Their answers are design inputs, not a final training exercise.
Make review a real part of the system.
Human involvement is meaningful when a reviewer has the context, time and ability to change the outcome. A simple approval button is insufficient if the evidence behind a recommendation is hidden or difficult to inspect.
Design the review around the decision. Show relevant sources, distinguish known facts from inference and make missing information visible. Give people a clear way to reject, correct or escalate an output.
From our work: review in Sponsor Hub.
The UHC Sponsorship workspace brings sponsor research, dated club evidence and a draft proposal into a shared process. The team can inspect the preparation, while executive review remains part of approving the proposal. A sponsorship target or a researched prospect is kept distinct from signed support.
That separation gives the reviewer a specific job: check the evidence and the proposed commitment. The system’s contribution is the preparation and visibility around that decision; the person remains responsible for the commitment.
Measure the work that becomes possible.
Time is a useful measure, but it is only part of the picture. Ask whether the change improves the quality of preparation, reduces avoidable rework or gives a person more attention for a difficult case. Agree on these measures before implementation.
The ambition is to make expertise more available to the business. That requires a thoughtful operating design, a practical system and a willingness to improve both as people use them.
Intelligence in practice.
