Making personalization feel handcrafted with the Agent Platform at Corinthia Hotels

Deciding to build

Corinthia had long recognised the gap. On property, it already treated guests differently from the moment they arrived, a spa guest's first point of contact was the spa itself, not the front desk, but the website treated everyone the same, built to capture wedding parties, spa visitors and plain room bookings alike with one generic experience. Closing that gap, so a guest felt the hotel already knew what they were looking for, beyond the room booking itself, became the business case for the project. For Emily Jewell, Senior Head of Global Marketing, that gap came down to something more fundamental than data.

Two things had to be true before that was possible. First, the data: different hotels and restaurants across the group had long run on different booking engines, with nothing feeding a single view of the guest. Corinthia spent roughly two years unifying its marketing and operations systems before personalization at this depth was even on the table.

Second, the content design and delivery mechanisms had to understand the customers’ mental model of the data Corinthia held.

"Responding to a guest's specific needs at each moment of interaction without becoming invasive is the hard part with hyper-personalization," says John Prior, Consulting Partner at MSQ DX. "You should never simply show what you know without knowing why you're using it and what the guest's expectation will be. Personalization succeeds where it feels natural and useful to the guest for you to anticipate what they need, based on their relationship with your brand."

Seven agents, one standard

The system works like a small, tireless content team. Seven AI agents each own a piece of the workflow, from setting content strategy through to writing variants and, most importantly, checking quality.

That last part is the difference. A dedicated quality-control agent reads everything the others produce and catches brand and grammar slips, down to style conventions as specific as Oxford comma preferences, before anything ships, so consistency is enforced by the system itself, not by a person reviewing every line. In a single run, the agents turn out a full set of persona and lifecycle variants, work that was never realistic to do by hand at this scale.

"AI processes are unpredictable by design, so it isn’t enough to rely on one-shot prompting to deliver reliable outcomes," John adds. "Multi-stage safeguards need to be built into the workflow, using adversarial agents to check each other’s work, with clear data boundaries at system integration points. The right system means that human intervention becomes an exception rather than a blocker."

None of this is guesswork. The system maps 76 distinct behavioral signals across 6,384 points to shape which variants get built and how they line up with each stage of the journey. That granularity is what lets Corinthia personalize around real guest behavior rather than a handful of broad segments.

Ten times the output, none of the compromise

That combination, seven specialist agents plus a human approval gate, is what turned the numbers above into more than a projection. The most recent cycle came back with a 100% first-time approval rate and zero brand violations, results that carry more weight in luxury than almost anywhere else, because the audience is unforgiving. Across every output, zero hallucinations. Variant production ran ten times faster than the old manual process, and planning alone saved more than 150 person-hours, time the team now spends on guests rather than proofing copy.

The project has been recognised twice by the industry: Innovative Use of AI in Marketing at the Hotel Marketing Association Awards 2026, and a Gold Standard win in the Artificial Intelligence in Travel Marketing category at The Travel Marketing Awards 2026.