Practical Guide · 12 min read · Hospitality & Food Service
Most guides on this topic come from software companies. This one comes from someone who has worked in real businesses for thirty years. No affiliate links. No tool lists. Only what actually works — and what doesn't.
Type "AI hospitality" into Google. What do you find? Articles from DISH Digital, gastromatic, Lightspeed, Zenchef — all excellent companies, but all with one thing in common: they're selling you something. Their content is marketing. That's not a problem — it's business. But you should know that before you follow their recommendations.
I don't sell software. I'm Roelof Hulshof: thirty years in hospitality and hotels, on the operator side and the supplier side. I know the language of the restaurateur writing rotas at 11pm. I know the language of the purchasing manager running food cost in Excel. And I know the letdown when an expensive tool nobody uses because the team was never brought in. I don't sell software, subscriptions, or tool bundles. I sell judgement.
This guide is my honest assessment after years of supporting operators and hoteliers through AI adoption. No commissions, no affiliate links, no software partnerships. Just what I've seen in practice — what works, what fails and why.
What you get here: an overview you could compile yourself — if you're willing to invest the next 200 hours in research. I've done that for you. You have better things to do.
Forget the images of robot waiters and fully automated kitchens. Most of that is marketing material for venture capitalists. What actually works in German, Austrian, Swiss, Dutch and Belgian businesses today is far less dramatic — and far more useful.
AI is at its core pattern recognition and prediction at scale. An AI system analyses data — historical bookings, weather forecasts, event calendars, POS data — and recognises patterns no human could spot in any reasonable time. From that it makes predictions and suggestions. That's it. No consciousness, no magic, no autonomous decisions.
For hospitality, three types of AI are relevant:
Content AI (ChatGPT, Claude, Gemini): writing, translating, reformulating. Social media posts, review replies, menu descriptions, onboarding documents for new staff. Immediately usable, no technical knowledge required.
Process AI (Planday, gastromatic, specific POS integrations): automating recurring decisions — shift scheduling, order suggestions, reservation management. Requires clean data and setup time.
Analytics AI (embedded in POS systems, BI tools): sales analysis, contribution margin calculation, demand forecasting. Turns your till data into actionable insights.
"AI isn't a product you buy. It's a capability you build."
The key distinction: good AI implementation isn't a purchasing process. It's a learning process. Businesses that understand this get real value. Businesses that buy a tool and wait for it to "do AI" will be disappointed.
These five areas aren't theoretical. They're based on what I've seen and measured in real businesses. Not all of them fit every operation — but at least two of them are relevant for you.
No-shows are a serious problem: on average 15–20 % of reservations don't arrive. Every empty table is a direct contribution margin loss. AI-powered reservation systems address this in three ways: automated reminders by SMS or email with optimised timing, predictive models for no-show probability based on historical patterns, and personalised communication based on previous visits.
Beyond that, AI chatbots can handle reservation enquiries around the clock — on your website, via WhatsApp, via Google. The guest gets an immediate answer. You don't have to answer the phone at night. This isn't experimental any more — it's standard in well-run businesses.
Rota planning is one of the most time-intensive management tasks. On average, managers spend 4–6 hours a week on rotas — and they still don't always get it right, because a big concert nearby was missed or the weather turned better than expected.
AI-powered systems combine historical sales data, weather data, local event calendars and booking patterns to produce realistic demand forecasts. The system suggests shift staffing. The manager decides. That's the important nuance: AI suggests, the human decides. The people dimension stays central — AI doesn't know the personal circumstances of your staff, their conflicts, or their development.
More important than the scheduling itself is whether people stay at all — onboarding determines whether someone is still there after eight weeks. More on that with real numbers in our staffing deep dive.
Food waste is the biggest hidden problem. Studies show 30–40 % of purchased food ends up in the bin. That's not a management failure — it's a forecasting problem. Too much ordered because the weekend picked up unexpectedly. Too little because a big catering event cleared out the stock.
Predictive ordering systems combine current booking levels, historical consumption data, weather forecasts and local events to optimise order quantities. Important: you don't need expensive software to start. A structured spreadsheet combined with ChatGPT analysis is a valid first step. Anyone with clean consumption data from the last three months can start immediately.
Online reviews are the new word of mouth. 87 % of people read reviews before visiting a restaurant. The problem: responding professionally to every review takes time. On Google, TripAdvisor, TheFork — and in a way that matches your brand voice and speaks to the next guest.
This is where content AI is immediately valuable. With a good template and ChatGPT or Claude, you can produce 20 review replies in 10 minutes that sound convincing, personal and brand-consistent. The same applies to social media content, menu descriptions and translations for international guests. Time saving: 5–8 hours a week — time that goes back to your guests.
This is the area where chefs are most sceptical — and simultaneously the one that delivers the fastest practical results. Not because AI is creative. Because it can calculate.
Scaling recipes to new quantities, seeing which dishes carry the business and which are dragging it, turning leftover seasonal ingredients into a suggestion: those are blunt tasks that eat time but need no creativity. ChatGPT gives you a recipe idea for three kilos of leftover Jerusalem artichokes in seconds. An AI-powered costing tool shows you which menu items earn their place and which quietly kill margin — once you see that, you cut three dishes and sharpen two. More numbers, more levers and realistic effect sizes are in our food cost deep dive.
I've supported dozens of businesses through AI adoption. The successes share common patterns — but so do the failures. These three mistakes come up most often.
Operators buy AI software because it sounds modern — without first defining which problem they want to solve. The result: an expensive tool nobody uses, because it solves the wrong problem or solves the right problem worse than the existing process. AI adoption always starts with one question: Which task costs me the most time that I could hand to a system? The answer leads to the right tool. Not the other way round.
AI is only as good as the data it receives. If you don't have a clean POS system, don't capture reservation data systematically, don't track food costs consistently — AI won't benefit you. Garbage in, garbage out. That's not a technical challenge, it's a management challenge. Before investing in AI tools, an honest stock-take is worthwhile: What data do I actually have, in what quality, and how far back does it go? That's the real starting point.
The most common killer of AI projects: implementation when the team isn't on board. Tools get sabotaged, ignored or simply not used when the team feels the decision was made over their heads. People who make changes with their team get further than people who buy the best tool alone. My advice: never introduce an AI tool as a "cost-saving measure". Introduce it as a "tool that takes work off your plate". The difference: who benefits from it? If the answer doesn't include the team, it'll fail.
AI adoption doesn't have to be complex. Here's the framework I recommend to operators and hoteliers — simple, sequential, with clear decision points.
Phase 1 · Weeks 1–4
Stocktake & Focus
Phase 2 · Weeks 5–8
Pilot
Phase 3 · Weeks 9–12
Measure & Decide
Not sure where to start? The free self-assessment gives you a clear picture in 10 minutes — with an observation that lands.
To the self-assessment →No affiliate income, no partnerships. Just my personal assessment after dozens of consulting mandates in the industry.
ChatGPT / Claude
Recommended — entry point for every businessFor content, reservation enquiries, recipe ideas, staff onboarding documents, review replies and internal communication. Entry from £0, ChatGPT Plus from £20/month. This is my first recommendation for everyone — regardless of business size. Learning curve: 1–2 weeks to productive use.
DISH Digital
Situation-dependentGood for online visibility and reservation management, especially if you don't yet have a strong online presence. But: you pay for a bundle of features of which you might use only 30 %. First question: which specific problem should DISH solve? If you know the answer, DISH is good. If not, wait.
Planday / gastromatic
Recommended for shift schedulingBoth systems are solid. gastromatic is more strongly rooted in the DACH region, with better German-speaking support and good integration into German payroll processes. Planday has the more compelling UX and better mobile experience for staff. If shift scheduling is your main problem: both are worth it. My tip: demo both systems, then let the team decide.
AI-generated websites & logos
CautionSound cheap, end up costing more. The problem: generic AI designs look like generic AI designs — and guests pick up on authenticity very quickly. Your brand is not an experiment in cost optimisation. For texts and content: yes. For brand personality and visual identity: invest in a real designer.
The entry point is cheaper than most think. ChatGPT Plus costs £20/month. Specialist tools start at £50–150/month. The biggest cost factor isn't the software — it's time: for setup, training and adaptation. If you realistically plan 2–4 weeks of onboarding, the investment pays back for most businesses within 3–6 months.
No. The relevant AI tools in 2026 are as accessible as sending a WhatsApp message. What you need: curiosity, patience for the first 2–3 weeks, and willingness to question processes. If you can write an email, you can use ChatGPT. This isn't a technical skill — it's a communication skill.
With the problem that costs you the most time. Not with the coolest tool. For most people that's: social media content, reservation management or food cost analysis. Start with one, do it well — and build from there. If at step one you don't know which problem you want to solve, start with the free self-assessment.
That's the most common objection — and the most legitimate. My advice: never introduce an AI tool as a "rationalisation measure". Introduce it as a "tool that takes work off your plate". The difference: who benefits. When the team understands the new tool handles the repetitive tasks — and gives them more time with guests — buy-in is significantly higher. Involvement beats instruction. Always.
In the next 5 years: no. AI takes over repetitive tasks — booking confirmations, standard replies, reports, rota suggestions. Guests still expect human contact. That's our industry's strength — and AI won't change it. What AI will change: businesses that use AI sensibly will be able to deploy their staff better — where the human dimension matters.
I've seen the industry go through many phases. The arrival of POS systems, when everyone said it was too complicated. The first online reservation portals, when operators insisted their guests would never book online. The smartphone revolution, which suddenly put review platforms in the hands of every guest.
AI is a similar threshold — but with a difference: it's coming faster, and it touches more areas simultaneously. Businesses that start now will have a structural advantage in three years. Not because they have more technology. But because they've spent longer learning how to make it work for them.
Hospitality and food service are and will remain a people business. The guest doesn't come because of the algorithm — they come because of the atmosphere, the taste, the feeling of being welcome. AI is the tool that gives us more time to deliver exactly that. Less time on rotas. Less time on orders. More time with guests.
This isn't a technological shift. It's a chance to get back to what our industry has always been: hospitality.
This guide is the overview. For those who want to go deeper into the money questions or the people questions, here are the two detailed pieces — both with real industry numbers, clear levers for this week and honest AI applications.
Food cost below 30 % — the 5 levers that really work.
Industry benchmarks by outlet type. The four sins — waste, portion size, single-supplier lock-in, menu tail. Realistic AI effects (1–3 % in the first year, 5–8 % in two years) rather than marketing promises. Three levers for this week.
Staff shortage? Check your onboarding first.
Why people stay, why they leave, why they applied in the first place. Three questions nobody asks. Plus: four practical AI applications for induction — 30-day plan, training snippets, pulse mails, mentor matching. With example calculations.
Next step