UK Hospitality's AI Debate Gets a Webinar — and a Sharper Warning

I had the webinar open on the laptop on the corner of the bar, volume low, while the breakfast crew plated behind me. The Caterer’s AI For Guest Experience panel went live mid-morning with the lineup you would expect: Peter Moore, CEO of Lolly; David Turner, CTO of Compass UK & Ireland; Dean Culpan, GM at Miiro Templeton Garden in London; and Will Francis, the digital marketing and AI advisor the trades book when they want someone who will push back. (Interpretation flag: the framing that follows — that Francis was the sharper voice in the room and the rest was the polite version of the same conversation — is my reading, not The Caterer’s.)

The five-lessons write-up is the primary record (The Caterer, five lessons). Search visibility, allergen data, hyper-personalisation, GDPR consent, and a case study from the South Place Hotel where Culpan’s earlier multilingual AI-driven SEO drove booking conversion up by as much as 120%. All real, all useful, and almost none of it the part an operator should actually write down.

The line worth writing down is in the companion piece The Caterer ran the same day, framed entirely around Francis (hospitality must be wary of over-using AI). The moment the customer realises the premium interaction has been outsourced to a chatbot, the premium goes with it. “If it had been a chatbot, the premium-ness of that would have gone down because it was a fancy place,” Francis said, on the example of a hotel WhatsApp concierge whose perceived value came from the guest knowing a real person was at the other end. “It wouldn’t have felt fancy.” He used the phrase “low grade.” That is a strong phrase for a sponsor-friendly webinar and he used it on purpose.

The polite panel

Strip Francis out and the panel was the industry conversation operators have been having since 2023.

Turner’s contribution was the corporate-IT version: get the data right, get the consent right, then the personalisation works. His specific example — accurate allergen data lets the system recommend safer menu items — is the bit a Compass CTO is professionally obliged to highlight, and he is right about it. Allergen data is the structured, high-stakes, repeatable problem where AI as a back-of-house tool earns its keep. He framed GDPR not as a constraint but as the trade where the customer gets “meaningful value” for the data they share.

Moore, from Lolly, was the EPOS-vendor voice — his firm sells the POS-and-payments layer where guest data actually lives, so his interest is structural, bullish, on brand. Culpan played operator-who-has-actually-used-this. The 120% conversion lift from multilingual AI-driven SEO at his previous property is the number that travels. His one-line on personalisation — “use AI, but don’t be creepy” — was the cleanest soundbite of the morning.

None of it is wrong. All of it is the polite version.

The unpolite version

Francis’s argument cuts the panel in two. AI, in his framing, is “just a tool,” excellent at “big spreadsheets for numbers and analysing them, or understanding customer data, customer habits, customer behaviours.” That is the back-of-house claim every operator should agree with. The harder claim is the customer-facing one: “people want to feel handled by a human, but with AI in the background.” The chatbot in the front seat doesn’t fail to add value — it actively subtracts the premium the venue charges for.

That is a sharper warning than the panel asked him to make, and it is the version that connects to the structural argument running through TableTransfers. Eitan’s four margins of a restaurant — gross, operating, brand, enterprise — names the problem cleanly. The chatbot at the host stand may shave operating margin by a few points of labour. It does its damage on the brand-margin line, the one no daily report surfaces. A concierge quietly swapped for a voice agent is still answering the messages. The brand premium it used to anchor is not.

What the voice-agent vendors don’t price

A consumer-facing restaurant voice agent — the category Famulor and its peers sell — runs around $80 to $180 a month for a venue handling three-to-five-hundred inbound calls, against a human front-of-house cost the same vendors put at $1,800 to $2,500 a month (Famulor pricing context). Read that as vendor framing, not a benchmark — those are the numbers the vendors make their own case with. But the pressure is real: at a ten-to-twenty-times labour-cost differential, the voice agent is on the table in every operator P&L review this quarter.

Francis’s warning is the brand-margin counterweight to that operating-margin maths. The cheap voice agent is genuinely cheap. The premium that pays for the room rate above it may not survive it. The webinar priced one of those numbers.

What I am writing down

Three things, before the lunch brief.

One. Watch the webinar back not for the slides but for the seven minutes around Francis’s intervention. The Caterer’s follow-up AI: Building Trust in the Workplace on 18 February is the workforce-trust version of this; the guest-trust version is already on the record.

Two. The operator question on AI in the guest-facing seat is not “does it work” — Culpan’s 120% number says back-of-house AI plainly does — but “does it cost more in brand margin than it saves in operating margin.”

Three. Francis’s phrase is the one I will quote back at the next vendor demo. People want to feel handled by a human, but with AI in the background. The webinar’s other four lessons are how you execute it. His is the one that tells you when you’ve executed it wrong.

— Naomi covers hotel F&B for TableTransfers. Tips: naomi@tabletransfers.com.