Conversational AI's Hotel Value Lies in Who Gets to Ask, TrustYou CEO Says

A hotel group producing twelve guest-experience reports a year was, functionally, asking twelve questions a year. That calculation, from TrustYou CEO Benjamin Jost, frames his argument that conversational AI's real value in hospitality is not speed but the collapse of what he calls "skill walls" between hotel data systems.

Jost, who also chairs the HSMAI Europe AI Advisory Board, laid out the case in an essay acknowledging his commercial interest: TrustYou builds software in this category. His starting point is a question most hotel managers never file — "Show me tomorrow's arrival list, sorted by customer lifetime value" — because the arrival list sits in the PMS, lifetime value sits in the CDP, and joining them requires someone with access to both systems, fluency in both data models and a free afternoon.

"By the time that person is found, tomorrow has happened," Jost wrote. The question never enters a backlog, appears as an unmet requirement or becomes a budget line. It simply goes unasked.

How the org chart followed the tooling

For thirty years, analytics in hotels was a department and a job title, Jost argues, because every data-holding system required a trained operator. Revenue management became a discipline — "one of the most quantitatively sophisticated corners of any industry" — because rate and demand data arrived with tooling that demanded a specialist first. Guest experience generates at least as much data, he notes, but never got its department. Housekeeping, F&B and loyalty each depend on whoever in the group is good with exports.

"The shape of a hotel's org chart is largely the shadow cast by the shape of its data tools," Jost wrote. "We hired around our software's requirements for thirty years and called the result an organization."

What changes when asking gets cheap?

The first consequence of conversational AI, Jost says, is volume, not speed. When the cost of asking falls to the cost of typing a sentence, the unconscious filter — is this worth someone's afternoon? — comes off, and a group asking twelve questions a year starts asking thousands.

The productivity sits in follow-ups: a manager who can ask three or four in ten minutes reaches an understanding that a manager holding one report never reaches. "The first answer is rarely the useful one," Jost wrote. "The useful one is three questions further in, and the old model priced those three questions out of existence."

The second change is bigger. The walls between silos were skill walls, not technical ones; the data could always be joined, but the people who understood two systems at once were rare, expensive and fully booked. Once conversational AI dissolves that wall, the questions that surface are cross-silo by nature:

• "Correlate ADR with guest satisfaction across my portfolio, and tell me what happens to ADR if I fix the three topics guests complain about most."
• "Show me tomorrow's arrivals, sorted by lifetime value."
• "Which properties had a cleanliness drop that begins after a change in housekeeping hours, and not before it?"

Each maps onto a decision somebody has to make this week, Jost notes. Rate decisions are guest experience decisions; loyalty decisions are operational decisions. "We have been optimizing the wrong set for three decades," he wrote.

Two preconditions

Cross-silo questioning still requires connected data: a shared identity that resolves across systems, agreed definitions and somewhere the data can meet. If the PMS and the feedback platform disagree about what counts as a property, no interface reconciles that. Conversational AI moves the data work upstream, Jost says — not less work, and "anyone selling it as less work is selling something."

The second precondition is access. Deploying the capability as an add-on for three named users at head office gives them same-morning answers while twenty general managers keep requesting reports. "The queue has not been removed. It has been given a better interface and moved inside the add-on," Jost wrote. "Analysis stopped being the constraint some time ago. Permission did not."

He cites one trial group where the manager, facing a decision on keeping the tool, declined an extension — she wanted to see how much she would miss it. The evening before a call, she built an internal presentation from a blank page using nothing but questions put to the data, and by trial's end had asked more of it than anyone in any group TrustYou worked with.

What buyers should test

Jost's procurement advice: ask vendors who may ask — whether every login-holder can query the data on signing day, and whether pricing meters by volume or by role. Volume metering is fair, he concedes, though a general manager who exhausts an allowance in week three joins a different queue, unsolved by anyone in the market, his company included.

His measurement standard: don't track response times. Count distinct questions asked this month, how many crossed a system boundary, and how many came from outside head office. If those are near zero, "you have bought a faster way to produce the reports you already had." Jost adds that the same shift is arriving on the guest-facing side of the desk, and the two conversations will eventually need to share one memory — work he expects the groups that dissolve the first skill wall to recognize fastest.