Skip to content

How Corinthia Hotels scaled AI personalisation without losing its voice.

AI THAT SOUNDS LIKE CORINTHIA. EVERY TIME.
Email_Hero_B_Photo_1200x600_v2

 

Luxury brands don't get second chances with tone. We built AI personalisation for Corinthia Hotels that hit 100% brand compliance and zero hallucinations, and it saves the team 150+ hours. Watch how we did it and what we learned.

WATCH TO FIND OUT
  • Meet the seven agents: what each one does, and how they hand the work along from strategy to brief, copy, quality check, edit and final human sign-off

  • What we learned: why one agent can't check its own work, and three other lessons from teaching AI a luxury brand voice

  • See it in action: the same Corinthia copy rewritten for a business traveller and for a guest celebrating a special occasion

  • Winning the business over: the one metric Emily had to hit, and how two awards unlocked more funding

THE CHALLENGE

Why Corinthia needed AI personalisation

At Corinthia, personalisation already happens on property. Staff know returning guests and remember their preferences. The website needed to do the same from the first click.

But 76 behavioural signals × 14 lifecycle stages × 5 personas is more than any team can write by hand, and generic AI would put a luxury brand voice at risk.

THE RESULTS

Scale without losing the brand

For a luxury brand, the one number that had to be right was brand compliance. It was. How Emily used that result to win over Corinthia's leadership is in the session.

 

COMPLIANCE

100%

brand compliance

ACCURACY

0

hallucinations or brand language violations 

EFFICIENCY

150+

hours saved, with 10 variants per workflow run 

WHAT WE LEARNED

FOUR THINGS THAT NEARLY WENT WRONG

Building AI personalisation for a luxury brand threw up problems no brand book prepares you for.

1

Agents can't mark their own homework.

A single copywriting agent kept excusing its own mistakes.

2

Brand books leave out the feelings.

Our first drafts followed every brand rule and still didn't sound like Corinthia. 

3

Give agents too much and they miss what matters.

Overloaded agents drop things, and you can't predict what they'll drop.

4

Context needs editing too.

 

With too little background, the copy barely changed. With too much, the agents lost the point

 

How we fixed each one is in the session.

" We see as AI as supporting creativity, but it should never replace it.  Hospitality is all about people."

Emily Jewell, Senior Global Head of Marketing, Corinthia Hotels


AI Personalisation FAQs

Can AI personalisation be 100% brand compliant?

Yes, with the right controls. Corinthia's AI personalisation variants reached 100% brand compliance and zero hallucinations. That came from verified brand sources, a preferred and forbidden vocabulary framework, a scored quality gate and human approval.

How do you make AI personalisation brand-safe?

Use only verified brand sources and a preferred and forbidden vocabulary list. Separate writing from editing, so no agent checks its own work. Score every variant on nine quality measures and reject anything below 3/5 on accuracy. Keep a human approval step before anything goes live. For Corinthia Hotels, that meant 100% brand compliance with zero hallucinations.

How much time did AI personalisation save Corinthia Hotels?

More than 150 hours of content work. One workflow run produces 10 brand-approved variants.

Do you need personal data for AI personalisation?

No. Corinthia's targeting is built on personas, lifecycle stages and behavioural signals, so it works for visitors who haven't identified themselves.

Which industries can use AI personalisation?

Any organisation with a strong brand or strict content rules. The approach we built for Corinthia Hotels applies equally to membership bodies, professional services and other regulated sectors, where every answer has to come from a tightly governed knowledge base. In areas like healthcare and law, plan for data safety from the start, so personalised content never reveals one person's information to someone else on a shared device.