The Wrong Questions

Almost none of it on the harder question of which decisions AI is quietly influencing, and who is positioned to catch it when the output is wrong.

Full credit to Berkus and Berkonomics for the framing. What follows is that governance question applied specifically to hotel ownership, asset management and operating leadership — where I'd argue the gap is, if anything, wider, because hospitality has embedded AI into daily operating decisions faster than most industries have built the accountability structures to match.

WHERE HOSPITALITY BOARDS SPEND THE HOUR

Sit through most owner or asset-management reviews of a property's AI posture and the agenda is predictable: which PMS or RMS vendor is being used, whether the chatbot is PCI-compliant, whether guest data sits in the right region, whether the brand's AI tools have been formally approved. These are legitimate questions. A hotel handles more personal data per square foot than almost any other hospitality-adjacent business — passport numbers, payment details, dietary and health disclosures, loyalty histories. Data governance is not theatre.

But an hour spent on encryption standards, as Berkus notes, can pass without anyone once asking whether an algorithm is quietly setting tonight's rate, deciding who gets called back first, or telling housekeeping how many rooms to skip.

That is the gap. Not a lack of rigor — a misplaced one. The technical review feels thorough because it produces documents: vendor security certifications, DPA addenda, a signed AI-use policy. None of those documents answer the operating question a general manager should be able to answer without hesitation: where, today, in this hotel, is a machine's output changing a guest's or an owner's outcome — and who is watching it do so?

THE FIRST-ORDER QUESTION, RESTATED FOR A HOTEL

Adapted directly from Berkus's framing, the question every ownership group and hotel leadership team should be able to answer for at least one material workflow is this:

What decisions in this hotel are being influenced by AI output, and who catches it when the output is wrong?

Not who owns the AI strategy. Not who signed off on the RMS contract. Not which committee reviewed the vendor's SOC 2 report. Those questions matter — they are second-order. The first-order issue is operational accountability, and in a hotel it sits closer to the floor than the boardroom: revenue management, front office, F&B, housekeeping, and increasingly, the algorithms sitting quietly inside each of those departments' software.

SIX PLACES IT'S ALREADY HAPPENING

A short, honest inventory — the kind Berkus recommends boards build for any business — looks like this for a typical full-service hotel:

Look down the third column. In most properties I review, it reads more like a hope than an answer. “Revenue manager, if they still look” is not a control — it is a habit that erodes the moment the RMS starts being right often enough that nobody double-checks it. That erosion is exactly the failure mode Berkus describes: errors that scale faster than the human review process was ever designed to catch, because the humans quietly stopped reviewing.

WHY HOSPITALITY IS MORE EXPOSED, NOT LESS

• Decisions repeat thousands of times a day. A mispriced room or a mis-triaged complaint is not one error — it is a pattern replicated across every similar guest, every similar night, until someone notices the trend rather than the incident.

• The people closest to the AI output are often the least empowered to challenge it. A front-desk agent overriding a CRS-suggested rate, or a line cook questioning a demand forecast, needs both the authority and the confidence to do so — and hospitality's operating culture does not always grant either.

• Seasonality and property-level idiosyncrasy make “drift” hard to see. A model trained on a full calendar year can look accurate in aggregate while being quietly wrong for three consecutive weeks around a local event the algorithm has never seen before.

• Ownership and operator incentives are not always aligned on who owns the answer. Brand-mandated tools, third-party RMS platforms and property-level staff can each assume accountability sits with someone else — which is another way of saying it sits with no one.

WHAT A USEFUL BOARD DISCUSSION ENDS WITH

Berkus is precise on this point, and it translates directly: a useful board or ownership discussion on AI should end with names, thresholds and response times — not enthusiasm about capability or reassurance about compliance. For a hotel, that means walking out of the room with answers to four questions for at least one workflow, before the next one is added to the agenda:

• Who has the standing authority to override this system's output today, without asking permission first?

• What would evidence of drift actually look like — and is anyone assigned to look for it on a schedule, rather than by accident?

• If the output is wrong, how many minutes or hours does it take to reach a human with both the context and the authority to act?

• Is that escalation path written down anywhere a new hire could find it — or does it live only in one manager's head?

THE PERTLINK VIEW

Within our own Human Experience Orchestrator (HXO) framework, this is precisely the layer we ask operators to design deliberately rather than inherit by accident: the point at which algorithmic recommendation meets human judgment, and someone — by name, not by department — is accountable for what happens next. Technology selection, data privacy and vendor diligence remain necessary work. But they are the second-order questions Berkus describes, dressed in hospitality's own vocabulary of PMS integrations and brand standards.

The first-order question doesn't require a vendor briefing to answer. It requires an honest hour with department heads, a plain-language inventory of where the algorithms already sit, and the discipline to name a human being next to each one. Good governance, in a hotel as in any boardroom, was always about accountability before complexity. AI has not changed that principle. It has only made the cost of skipping it arrive faster than most owners are used to.

Source & Attribution

This paper is a hospitality-specific extension of the governance argument made by Dave Berkus in “Are boards asking leaders the wrong AI questions?”, Berkonomics, published 30 July 2026 (berkus.com). The core framing — first-order accountability questions over second-order technical ones — originates with Berkus; the hospitality application, workflow inventory and HXO framing are Pertlink's own.

Created with the help of various AI tools – but always with a HITL.