AI in Hospitality: Cutting Through the Hype with Crunchtime

There's no shortage of AI talk in the hospitality industry right now. Every vendor has a pitch, every conference has a panel, and every LinkedIn feed is full of bold predictions. But what's actually happening on the ground? What's delivering real results, and what's quietly been shelved?

In a recent event hosted with Crunchtime at The Tree House Hotel, we brought together John Raguin from CrunchTime, Tim Cross from Caffè Nero UK, and Andrew Winter from SSP Group plc for an honest conversation about where AI stands in food & beverage operations today. Here's what came out of it.

The Hype Doesn't Always Match the Reality

One of the most striking examples discussed was the wave of AI-powered drive-through ordering that swept across the US. Operators piled in, costs were committed, and then, almost universally, it was abandoned. Ambient noise, high error rates, and operational complexity made it unworkable. The irony? Most operators didn't even know their human error rates at drive-through, so when the AI underperformed against an invisible benchmark, it looked far worse than it was.

The lesson isn't that AI doesn't work. It's that jumping in without a clear use case, proper data, and realistic expectations is a fast track to failure.

Where It Is Working: Forecasting

If there's one area where the panel agreed AI is genuinely moving the needle, it's forecasting. John from CrunchTime shared that adoption of their AI forecasting tool went from just 1 to 2% of locations to around 50% in the space of twelve months, not because of a big marketing push, but because operators started seeing it outperform their manual methods.

The context matters here. A stat raised during the conversation stopped the room: 82% of UK-based operators are already using some form of tech-based forecasting, but average accuracy sits at just 62%. Most GMs, as John put it bluntly, just take last week's numbers and repeat them, regardless of the weather, local events, or seasonal shifts. AI can fix that. And when it does, the ROI is significant, especially at scale.

For Caffè Nero, the focus is on product availability, making sure the right stock is in the right store at the right time. Too little and you lose revenue. Too much and you waste it. AI-driven sales forecasting, fed into their operational systems, is the tool they're building toward.

Data Quality is Non-Negotiable

A recurring theme throughout the conversation was data. Specifically, bad data.

John shared a candid example from CrunchTime's own business: when they first switched on AI for customer support, it was accurate just 3% of the time. The problem wasn't the AI, it was that it was pulling from outdated internal documentation. Garbage in, garbage out.

The panel's advice for anyone starting their AI journey: get your data house in order first. For forecasting specifically, CrunchTime has found that 400 days of clean historical data is the sweet spot, enough to generate meaningful accuracy without the noise that comes from loading decades of potentially inconsistent records.

Change Management is the Hardest Part

Ask John what causes CrunchTime implementations to go sideways and the answer isn't a technical one. It's change management. Operators call in not because the system is broken, but because the numbers don't match what they expected, often because a calculation has changed, or a process has shifted, and nobody prepared the team for it.

Andy from SSP put it well: for too long, businesses have led change through technology rather than through business transformation. SSP is now deliberately investing in change management capability, empowering local teams, setting clear KPIs, and keeping the human in the loop rather than asking people to simply trust a black box.

What's Coming Next

Looking ahead, the panel sees the biggest shift coming in how managers interact with data. Instead of sitting in the back office scrolling through reports, John envisions managers asking their phone a question, "How many covers did we do last Tuesday?", and getting an accurate, instant answer. Voice-based, conversational AI that brings managers out front where they belong.

It won't happen overnight. The consensus was five to seven years before this becomes the norm. But the direction is clear.

The Takeaway

AI isn't a silver bullet, but it's not smoke and mirrors either. The operators winning with it right now are the ones who started with a real business problem, built on clean data, and brought their teams on the journey. Everything else is just noise.