Stop asking what AI earns; ask who it earns for

Every hotel owner meeting now contains the same question. What is AI earning us?

It sounds reasonable. It is the wrong question.

Show the same AI tool to an operator, an asset owner and a guest, and you will hear three different answers. Each is correct on its own scorecard. None of them agrees on what “return” actually means.

That gap kept surfacing as I prepared to moderate a panel on AI and human intelligence at IHIF Asia. Hospitality is not short of AI. It is short of agreement.

So before asking what AI earns, hotels need to ask something tougher: Who does it earn for?

Most AI business cases stumble for one of two reasons. Neither is technical.

The first is sequencing. Hotels switch the tool on before anyone defines what “before” looked like. No baseline, no agreed process, no definition of success. Six months later, everyone has an opinion on whether it worked. Nobody has proof.

The second is ownership. Operator, owner and guest each define return differently. Left unreconciled, the same result is celebrated in one meeting and challenged in the next.

Payback periods differ too. An operator on a management contract and an owner with a long hold will never read the same timeline the same way. But timing is a symptom. Sequencing and ownership are the cause.

So let us go around the table.

For the operator, return usually comes down to one number: gross operating profit.

That is not cynicism. It is the contract. Management fees are commonly linked to gross operating profit, so whatever lifts GOP lifts the operator’s income.

This is where AI gets interesting, and awkward. Where the cost of an AI tool is booked quietly decides whose number improves. Capitalize it, and the cost is spread over years below the GOP line. Expense it, and it hits GOP in full, today.

A senior hotel finance leader I respect a great deal described the tug of war precisely. Operators lean toward capitalizing. Owners resist, because it flatters operating performance while they pay a higher fee on a higher GOP. In practice, most AI is subscription software that rarely qualifies as capital anyway. So many hotel teams simply expense it and avoid the discussion.

Either way, the return has been shaped before the tool has done anything.

Then comes the attribution trap. Operators love to credit AI with a lift in average daily rate or revenue per available room. But those numbers move with the market, with events, with competitors and with the weather. Claim them, and you open a debate that never closes.

Owners are not against AI. They are against numbers they cannot verify.

That same finance leader made a point every revenue team should hear: Stop leading with metrics the market controls. Start with measures the AI can genuinely influence, and that both sides can see.

Channel mix is the obvious candidate. Owners have wanted a stronger direct share for years. If AI personalizes the website, answers pre-booking questions at any hour and turns lookers into bookers, the shift to direct shows up in every monthly report.

But the real prize is what the direct guest spends. SiteMinder’s 2025 Hotel Booking Trends found that hotel websites generated an average of $516 per booking, against $312 through OTAs. The difference came from higher room categories, longer stays and extras.

That is a key performance indicator an owner will sign: direct share up, total spend per guest up. Agree on both baselines before go-live, and the conversation moves from whether the AI worked to how much it delivered.

Owners also carry the risk, and AI can protect them from it. A tool that monitors major corporate accounts and triggers credit chasers early may save very little in labor. Avoid one significant bad debt, and it has paid for itself.

That is a return too. It just rarely appears in the business case.

The guest does not care about GOP. The guest’s return is time, recognition and ease.

Hotels have learned this the hard way. When the industry borrowed self-check-in kiosks from the airlines, the logic was flawless on paper. Fewer front desk hours, faster arrivals, lower cost per check-in.

Reality disagreed. Guests arriving after a long journey did not want a machine. They wanted a welcome. Many kiosks ended up bypassed, underused or quietly removed. The labor saving was real, but so was the damage to the arrival experience, and to the loyalty and spend that follow it.

That is the line between automation that removes friction and automation that removes the hotel.

AI that answers a pre-arrival question at 2 a.m., remembers a pillow preference or settles a billing query without a queue adds to the guest’s return. AI that forces a guest through a script when they want a person takes it away.

Measure the guest’s return where the guest records it. The KPI is review score uplift. Set the baseline before go-live and track it by theme in the review text. It is visible to operator and owner alike, it moves with the experience rather than the market, and every revenue manager knows a stronger score supports a stronger rate.

And here the perspectives collide. The direct guest who spends more is exactly the guest a clumsy bot sends back to the OTA next time.

The test is simple. Does this make the stay easier, or just cheaper to run?

Then there are the people doing the work. Here the industry needs some honesty.

AI labor savings are real only at scale. A cluster finance team reconciling bank and supplier accounts across ten properties can genuinely shrink. So can a large chain’s rate monitoring desk. A single independent hotel with one revenue manager cannot save half a person.

For most hotels, the case is redeployment, not reduction. Hours freed from reconciliations, rate shopping or first-pass contract reviews become hours for guests, owners and better decisions.

Which leads to an uncomfortable conclusion. The least exciting AI may be the most defensible. Back office tools rarely win awards. But their costs are clear, their baselines are easy to set and their benefits are hard to dispute.

The chatbot gets the headline. The reconciliation tool gets the return.

The fix is not more technology. It is sequencing.

Before any AI contract is signed, the operator and owner should sit at the same table and agree three things: The baseline today, the metric the tool can genuinely influence, and whose return it is meant to improve.

If the guest pays the price, it is not a return. If the owner cannot verify it, it will be disputed. If the operator’s GOP is the only winner, it will be resisted.

AI will keep getting better. The question in the owner meeting will keep getting louder.

So stop asking what AI earns. Start asking who it earns for.

 

Philip Schaetz is founder and managing director of CUBE. Prior to establishing CUBE in 2018, Philip enjoyed a long career in revenue management, distribution expertise and sales and marketing strategy in senior global positions for some of the world’s largest hotel & resort brands.This column is part of ISHC Global Insights, a partnership between CoStar News and the International Society of Hospitality Consultants.