Involve sceptics early in AI initiatives
Sceptics are more valuable than advocates during the design of an AI initiative — they see the failures most clearly; involve them early in roles that protect against the failures they fear, rather than sidelining them as resistant to change.
The instinct when announcing an AI initiative is to bring enthusiasts into the planning group — people who will be energetic advocates, who can help demonstrate early wins, who will not slow the momentum by questioning fundamentals. The instinct is mistaken. Sceptics are more valuable than advocates during the design phase, because they are the people most likely to notice what will actually go wrong.
The working rule is to bring sceptics into the work, specifically in roles that protect against the failures they fear. A staff member who worries the AI will produce slop is a useful person to put in charge of the quality-control function for AI output. A staff member who worries about job displacement is a useful person to put in charge of articulating which parts of the work should explicitly not be automated. The scepticism gets converted into scope definition; the sceptic becomes an owner of the corners of the initiative where their concerns are load-bearing.
Why this works
Two mechanisms make the move effective.
First, it addresses the underlying pattern in Unvoiced staff resistance is the primary failure mode of AI initiatives. Sceptics who are involved are sceptics who are visible; their concerns are on the table rather than expressed through later non-compliance. The initiative loses the “silent killer” distribution of the pattern because the resistance has a name and an address.
Second, it improves the initiative itself. Sceptics who are specifically asked to imagine the failure modes often see them more clearly than enthusiasts who are asked to imagine the successes. The initiative that survives sceptical design is stronger than the initiative built only from advocate enthusiasm.
How to apply
The move is not “soften the sceptic into an advocate”. The move is “give the sceptic a role that honours the substance of their concern”. The scepticism is the information, not the obstacle. A sceptic who has been converted into an advocate has lost their usefulness; a sceptic who is now running the quality-control function is using their scepticism for the initiative rather than against it.
This does not guarantee the initiative will succeed. It does mean the failure modes the sceptic named will be on somebody’s deliberate watch, rather than allowed to occur and explain themselves later. That is the specific improvement the heuristic produces — and it is often enough to change the trajectory of an initiative that would otherwise end in the pattern this heuristic is trying to prevent.