HN Original: Decisions You Can Trust: A Conversation with IDeaS at ITB Berlin

At ITB Berlin, Simone Puorto spoke with Klaus Kohlmayr, Chief Evangelist at IDeaS, about the trust gap in AI adoption, why process silos are just as damaging as data silos, and what the revenue manager’s role looks like in 24 months. The conversation was grounded and honest, including a candid moment about MCP that is worth reading to the end.

The trust gap is real, and it is holding the industry back

IDeaS research has identified a significant gap between intention and confidence when it comes to AI in hospitality. Hotels say they want to use AI. When it comes to actually trusting the outputs enough to act on them, the level of confidence is a fraction of what would be needed for meaningful adoption.

Klaus was direct about why. Clean data and sound processes are prerequisites, not nice-to-haves. Slapping an AI interface on a business intelligence tool does not fix what is underneath it. If the data is fragmented and the processes are inconsistent, the AI output will reflect that, and no amount of polished interface will change it. IDeaS’s stated position is simple: decisions you can trust. Anything they build in AI has to meet that standard before it ships.

Credible hallucinations are the real risk

The hallucination conversation in hospitality tends to focus on the obvious failures: a chatbot inventing a temple that does not exist, or, as came up in the interview, a client receiving a booking confirmation that included an Olympic pool in every room. Those errors are easy to spot and easy to dismiss.

Klaus raised a more serious concern: the credible hallucination. A small error in a data output that looks plausible, gets incorporated into a pricing or distribution strategy, and compounds over time. The damage from that kind of error is invisible until it is not. It is why IDeaS is spending significant time on data cleanliness and aggregation before adding AI layers, and why their new Rate Data Advantage product uses 400 times more data points than a standard shopping tool. The foundation has to earn the trust before the interface can.

Night audit in 2025 looks like night audit in 1985

Klaus has a son working in a hotel. The night audit process his son runs today uses the same reports, the same procedures and produces the same issues that Klaus encountered when he worked in hotels 40 years ago. That observation carries weight coming from someone in his position. The operational layer of hospitality has not kept pace with the technology sold to it. Bringing AI into an unchanged operational model risks scaling the inefficiencies rather than replacing them.

People silos, not just tech silos

The integration challenges between systems remain unsolved. But Klaus pointed to something that gets less attention: the silos between commercial teams. Revenue management, marketing, sales and distribution frequently operate with limited shared visibility. IDeaS has just launched a product called Spotlight that begins to address the gap between revenue management and marketing specifically, giving marketing teams predictive data on where spend will generate the highest uplift. It is a start, but the broader problem across the full commercial organisation remains open.

A note on MCP, CLI and the pace of change

Klaus referenced a recent episode of the Lex Fridman podcast featuring Peter Steinberger of OpenClaw: OpenClaw: The Viral AI Agent that Broke the Internet. Steinberger’s position in that conversation was that MCP may already be losing relevance, with CLI emerging as the more durable approach for agent integration. Klaus was candid: he does not know how long MCP will survive as the dominant protocol, and nobody does. The space is moving fast enough that the standard answer from six months ago may not hold today.

His advice was to take a responsible, thoughtful approach rather than chasing every new capability as it emerges or waiting for clarity that may never fully arrive. The framework he described: AI should first show you something, then tell you something, then do something for you. The industry is moving along that path, but the journey is gradual and needs to be deliberate.

24 months

Asked how far away the future is where a revenue manager supervises agents rather than doing the work those agents will handle, Klaus gave a specific answer: 24 months. Not everything at once, and not uniformly across the industry. But the shift from execution to supervision, from doing to directing, from individual decisions to managing a team of agents, is closer than most people in the room at ITB are planning for.