When experienced professionals hesitate over AI: what remains unresolved?

An experienced physician listens to a presentation about a new AI application. She asks two questions, declines a trial for the time being and returns to the ward. What do we now know about her attitude? We know about a decision in a particular situation. We still need to understand her reasons. Immediately attributing it to an unwillingness to change settles on an explanation before the conversation has begun.

For leaders in hospitals, pharmaceutical companies and medical device businesses, that distinction has practical consequences. An unresolved clinical question needs a different response from a lack of time to learn a system. Concern about one's place in a future working arrangement calls for another conversation again. I propose translating reservations into conditions that can be examined. The approach below is my organisational inference; the cited studies did not test its effectiveness.

What research can contribute to the conversation

Negash and colleagues studied eight focus groups with 39 physicians in Germany, conducted between June 2022 and January 2023 and published in 2025. Evidence, human oversight and patient relationships shaped participants' accounts. This qualitative sample identifies possible reasons; it establishes neither their prevalence nor a causal effect of age. [1]

Schaffernak and colleagues interviewed 22 ophthalmology professionals in Germany, Austria and Switzerland in July and August 2024, publishing in 2025. Openness to AI could coexist with unresolved technical and organisational requirements. Self-selection and differing experience limit transferability. These interviews also cannot establish a representative ranking of barriers. [2]

Those boundaries belong in any discussion of experienced professionals. The papers provide no basis for dividing people into innovative youngsters and sceptical older colleagues. An objection also cannot reliably reveal a hidden fear. For a particular organisation, useful knowledge emerges when someone explains their own condition and it becomes possible to establish whether that condition can be met.

Name the decision precisely

I would first clarify what the person is being asked to agree to. Watching a demonstration, working through a protected practice case, using a system in daily work and approving its output involve different commitments. An invitation that folds all four into the same request makes any answer difficult to interpret.

A useful explanation might be: this meeting concerns an assessment of a fictional example. We will decide about deployment later, using additional information. That boundary creates a comprehensible decision. It promises no safety that has yet to be established. It explains what agreement would actually cover, so the person can respond to a clearly defined proposal.

A refusal needs an object, too. Does it concern the specific application, the proposed timing or the task? Someone who has had a good experience with one application may reasonably decline another. Someone without time for a trial today may still consider its clinical value plausible. These possibilities should remain open during the conversation.

Translate the objection into a condition

For me, the most productive next question is: what information or change would affect your decision? The answer should connect to something observable where possible. A general concern about reliability might become a request to examine particular exceptional cases. An objection about workload might become a requirement for a process that avoids entering the same information twice.

The person leading the conversation should allow the answer to determine the category. Someone describing a missing interface should not then have to explain their workload limit in psychological terms. Equally, an explicitly expressed concern about one's role deserves a factual response: which tasks are expected to change? Which expectations will continue? Which consequences have already been decided, and which remain open?

An objection that can be examined is useful to a supplier as well. It describes a requirement more precisely than an instruction to build more trust. If the application does not currently meet that requirement, the supplier can state a defensible limit. Recognising an unsuitable use while it remains unsuitable is part of building a company in a regulated environment.

A constructed hospital example

A fictional hospital wants to use AI to prepare internal summaries of educational materials. A senior physician initially declines to participate. The discussion reveals that she is expected to approve the content afterwards but has been given no additional time to do so. The version of the source materials is also unclear. Her objection concerns two specific working conditions.

The project lead could therefore prepare a bounded exercise: explicitly fictional educational material, a recorded source version and a reserved review session. The questions the exercise should answer are agreed in advance. The physician can then assess whether the draft makes review easier and whether important qualifications remain visible. Participating does not by itself approve the eventual process.

Suppose she finds the preparation useful but needs longer to review it than expected. That produces two useful findings. The application may have value; the proposed staffing does not yet fit the work. A participation rate alone would conceal this distinction. A short record of the outcome can preserve it without labelling someone as generally resistant to change.

Take responsibility for answering

A conversation about reservations also creates obligations for leadership. Once an objection has been made examinable, it should be clear who will obtain the answer and when another decision would be useful. If the source-version question remains unresolved, another general presentation about AI contributes little. The next action follows from the identified gap.

I would capture three sentences: the outstanding condition is …; this person or function will clarify it …; we will then decide about this bounded step … . This is a proposal for collaboration, rather than a scientifically validated scoring instrument. Its purpose is to make commitments and unresolved points traceable, without requiring an elaborate new administrative process.

Some objections cannot be resolved in a week. A reasoned deferral is then a possible decision. Other conditions may turn out to have been met already once someone explains the arrangement. That should also become visible. A fair process allows both sides to change their assessment without losing credibility, because the reason for the change is available to everyone involved.

Recognise progress in the problem clarified

I would assess whether this approach helps locally by examining the decisions it enables. Are objections becoming more specific? Can recurring problems be grouped? Do employees receive answers that address their actual questions? Are missing evidence, missing capacity and a deliberate professional judgement distinguished? These would be useful observations during the first use of the approach.

Leaders should also revisit their original explanation. Several experienced people identifying the same additional review step is a different situation from three entirely separate concerns. If a promised change fails to happen, the next meeting should address that fact. Repeated reservations may then serve as a precise reminder of an outstanding commitment.

From my perspective as a physician and founder, the central achievement of these conversations is a more accurate account of the work. An experienced colleague needs room to express a considered view without being expected to display enthusiasm. They should be able to explain the conditions under which they can professionally support the next step. For an organisation, that information provides a practical basis for bringing implementation and responsibility together.