There is a specific moment that anyone who has worked in a senior advisory role will recognise. You have thought carefully about a problem. You have held the scientific evidence alongside the commercial reality, noticed where they conflict, worked out what a defensible path through that conflict looks like. You walk into the room and give your recommendation. And you watch it fail to land — not because it was wrong, but because the people in the room had no way to tell whether it was right.

This is not a communication failure. It is not a confidence problem. Understanding what it actually is changes what you think can be done about it.

When a decision-maker allocates weight to an advisory contribution in a senior forum, they are rarely doing so because they have directly evaluated the reasoning behind it. They are doing so because the visible signals available to them, such as how the recommendation is framed, how uncertainty is handled, how the professional holds their position under challenge, are consistent with a prior expectation that this person’s reasoning can be relied upon. The quality of the reasoning is not the direct input. The inferred reliability of it is.

This is what makes the gap between a good recommendation and its reception so resistant to the usual explanations. The decision-maker isn’t being incurious. The professional isn’t communicating badly. The gap exists because integrative reasoning across scientific, regulatory, commerc ial, and operational domains simultaneously cannot be directly observed in the conditions under which senior decisions are actually made.

Why not? That’s the question this essay addresses.

The instinct is to treat it as a communication problem. The reasoning wasn’t visible because it wasn’t made visible — held internally when it should have been articulated, framed in specialist terms when it should have been translated. Fix the communication and you close the gap. This isn’t wrong. It’s incomplete.

The reasoning isn’t fully visible even when the professional is actively trying to make it visible. Because the incompleteness is not a failure of expression. It is a property of what expert knowledge actually is.

The philosopher Michael Polanyi, writing about the nature of expertise, put it in a sentence that hasn’t been improved on: “We can know more than we can tell.” What this means in practice is that expertise accumulated through experience doesn’t sit in articulable, transferable form. It is embedded in the practitioner. The experienced clinician who identifies the atypical presentation cannot fully reconstruct the process by which they did so. The senior negotiator who reads a room accurately can’t fully decompose what they read. The knowledge is real, operative, and consequential. It is not, or is only partially, available for export.

For advisory professionals working across multiple domains simultaneously, this tacitness operates at two levels that compound each other.

The first is within each domain. Integrating regulatory risk into a recommendation draws on tacit understanding of how regulators actually think about benefit-risk, what scientific uncertainty means in context for different decision types, which constraints are typically binding and which have room. That understanding cannot be fully transmitted even to a knowledgeable observer who shares one domain but not all of them.

The second level is the integration process itself. The practitioner who routinely produces defensible cross-domain recommendations has developed, through experience, implicit strategies for weighing competing considerations. Recognising when constraints are genuinely irreconcilable versus when they only appear to be, knowing when sufficient synthesis has been achieved to support a directional stance. This is even less articulable than domain-specific knowledge, because it concerns the process of integration rather than any single domain’s content.

Neither level can be fully recovered by an outside observer in real time. Not because the professional is withholding anything. Because the thing being observed resists decomposition at its most expert.

There is a second dimension to this, and it has nothing to do with the professional.

Even a cooperative professional who is committed to transparency and trying hard to make the reasoning visible runs into a ceiling that isn’t of their making. Senior decision forums are not designed for the full reconstruction of complex cross-domain reasoning. Time pressure, competing demands, incomplete information, finite cognitive bandwidth. Following the scientific argument, tracking the regulatory constraint, weighing the commercial implication, and evaluating how they were integrated requires exactly the kind of slow, deliberate, effortful thinking not available in high-load environments.

When evaluative bandwidth is constrained, something else takes over. Decision-makers develop, through experience, an implicit model of what reliable reasoning looks like built from prior interactions, pattern recognition, the accumulated texture of contributions that turned out to be right. That model becomes a template against which new contributions are assessed. Contributions that match it are attributed as reliable. Contributions that don’t match it register differently, regardless of their actual quality.

This template is not arbitrary and it is not unreliable. It uses visible signals including structural coherence, calibrated confidence, consistency across contexts, which do correlate to some extent with reasoning quality. The problem arises when the template misreads what it is seeing: when the features of genuinely sophisticated reasoning don’t match the pattern the template expects. Explicit acknowledgement of competing constraints can read as indecision. Proportionate representation of uncertainty can read as fragility. The recommendation that is carefully qualified because the evidence genuinely warrants qualification looks no different, under this kind of assessment, from the recommendation that is simply hedged by someone who isn’t sure.

Putting these two things together gives a cleaner account than either alone.

The reasoning isn’t fully visible because expert integrative reasoning is inherently tacit — a property of what the knowledge is, not a failure of expression. And even what is made visible cannot be fully assessed, because the forum operates under the cognitive constraints that make exhaustive evaluation unavailable as a practical matter. These are not separate problems. They are the same structural gap viewed from two sides.

This is why the usual responses don’t reach the problem. Better evidence improves the quality of the reasoning, which matters. Better communication improves what of the reasoning can be seen, which also matters. Neither addresses the evaluation mechanism itself — the process by which a decision-maker constructs an expectation of reliability from the signals available. That mechanism operates whether or not the reasoning is good, and whether or not the communication is clear.

The tacit basis of integrative reasoning cannot be made explicit. That is not a limitation to be overcome as it is a feature of expertise at senior levels. But the visible structure of the reasoning, the constraints that were held, the assumptions that were made, the trade-offs that were navigated, are not wholly invisible. However, they can be made more available than they typically are.