The Best Way to Test an AI Strategy: Invite the Critics to the Table

The best way to test how resilient an AI strategy is? Invite its critics to the table.

In strategy meetings, the people who get heard are usually those who have championed AI adoption and succeeded with it. They are also happy to tell you what works. The experts who question the benefits of a pilot, or challenge its logic, hold precise knowledge that they do not share — because no one asks.

I ran into this phenomenon in recent research (Saup et al., Journal of Decision Systems 2026). The researchers noticed a surprising limitation in their material: the interviewees were almost without exception AI advocates. The company’s strong top-down governance channeled dissenting views into formal governance routines. The researchers themselves note that the absence of resistance in the data does not mean it is absent from the organization.

So what helps? User-centered design — it brings the development work right up against the user’s everyday reality.

You sit next to the expert and observe how the work is actually done. What slows it down? Which AI outputs does the user doubt and bypass? That is extremely valuable data.

When users see that the issues they raise — hallucination risks, illogical results — are taken seriously, the critic turns into AI’s quality assurance. This trust decides whether the AI investment produces strategic value or remains a fine experiment.

I would argue that user-centered design is essential in AI implementations too. Agree, disagree?

Aspenly · Thinking