Welcome to Level 1
Hello, and welcome! Before we touch a single prompt, I want to answer one simple question with you. What is an AI assistant, really, and where does it fit in a working day behind the bar, on the pass or at reception?
By the end of this level you'll be able to explain it in plain words. You'll sort your own tasks into three piles. And you'll be able to quote some real, dated numbers on how the trade is using AI.
So what does an AI assistant actually do?
Tools like ChatGPT, Gemini, Claude and Microsoft 365 Copilot are built on something called a large language model. That sounds fancy. The idea underneath is simpler than you'd think.
You know how your phone suggests the next word when you're texting? An AI assistant does the same thing on a far bigger scale. It has learned patterns from huge amounts of text, so it predicts which words are likely to come next. That's why it's brilliant at writing. A polite reply, a punchy social post or a tidy checklist comes out in seconds.
Here's the catch, though. Predicting words that sound right isn't the same as knowing what's true. The assistant hasn't tasted your soup. It hasn't seen your diary, your recipe cards or your supplier's spec sheets, unless you show it.
Why does it sound so sure when it's wrong?
This is the big one, so let's slow down.
When an AI makes something up and states it confidently, people call it a hallucination. In September 2025, researchers at OpenAI published a paper explaining a key reason why it happens. The way these models are trained and tested tends to reward guessing over admitting "I'm not sure."
Think of a multiple-choice quiz. If you leave an answer blank, you get zero. If you guess, you might get lucky. So the model learns to guess, and to guess smoothly.
Picture a keen new starter on their first shift. They're friendly, fast and eager to please. A guest asks if the gravy is gluten-free, and rather than say "let me check," they say "yes, it's fine!" with a big smile. Lovely manner. Dangerous answer.
That's your AI assistant. Great helper. Never the final word on facts.
Three piles for your everyday tasks
So how do you decide when to use it? Let's sort your work into three piles.
Pile one: AI can help. These are writing and thinking jobs where a mistake is easy to spot and cheap to fix. Drafting a friendly reply to a booking enquiry. Brainstorming names for a winter cocktail. Turning your messy notes into a neat team briefing. Ideas for a quiz night theme. You still read it before it goes anywhere, but the risk is low.
Pile two: needs close checking. Here, AI can save you time, but errors matter. A first draft of next week's rota from your notes. A prep list for a 60-cover function. A reply to an unhappy guest's review. A rough stock order. You check every line against what you actually know: who's available, what's in the walk-in, what really happened on the night.
Pile three: must stay human. Some things are too important for a guess. Allergen information is the clearest example. Whether a dish contains nuts, sesame or celery has to come from your recipes, your ingredient labels and your kitchen's own records. It should never come from an AI's best guess. The same goes for final decisions about people, like disciplinary matters, and for anything that could put a guest at risk.
We'll go much deeper on allergens and the rules around them in later levels. For now, just hold on to this: AI drafts, you check, you decide. That three-step habit runs right through this course.
How is the trade actually using AI?
You might be wondering whether anyone in hospitality is really doing this. They are, and the numbers are worth knowing. Notice that each one comes with who, when and where. That's how you should always quote a statistic.
In the UK, NIQ and Sona published a survey of hospitality leaders on 18 December 2025. It was run in July and August 2025, and the respondents ran more than 13,000 venues between them. The most common uses were data tasks like sales forecasts and reports (49%), creating marketing materials (47%) and writing emails (43%). Staff scheduling came in at 27%, and stock optimisation at just 12%. So far, the trade is mostly using AI for writing and admin. Fewer businesses use it for operations.
In the US, Toast surveyed 676 restaurant operators with 16 or fewer locations in April 2026. Nearly nine in ten said they were using AI.
And for hotels, the h2c Hotel Chain Tech Report, published on 14 October 2026, found AI in use at 91% of the hotel chains surveyed. But only 13% reported a measurable return on that investment so far.
What does that last figure tell you? Using AI and getting real value from it are two different things. Plenty of businesses have the tools. Far fewer have worked out how to use them well. That's exactly the gap this course is here to close for you.
Your turn
Grab a pen. Write down ten tasks from your last shift. Now put each one into a pile: AI can help, needs close checking or must stay human. Be honest. If you're unsure, it goes in the checking pile.
Quick checklist:
- Can I explain in one sentence what an AI assistant does?
- Have I sorted ten of my own tasks into the three piles?
- Is allergen information firmly in my "must stay human" pile?
- Can I quote one statistic with who ran it and when?
Key takeaways
- An AI assistant predicts likely words. It's great at drafting, but it doesn't know your venue's facts.
- It can sound confident when it's wrong, partly because training rewards guessing over admitting uncertainty.
- Sort tasks into three piles: AI can help, needs close checking, must stay human.
- Allergen facts always come from your own records, never from AI.
- Always quote statistics with who ran the survey, when, and where.
- Lots of businesses use AI, but far fewer have seen measurable value yet. Using it well is the skill.