Can a restaurant be full every day and still lose money? Although it may seem paradoxical, it’s a reality more common than we think in an industry where controlling revenue doesn’t always reveal profitability. Rising costs, fluctuations in raw material prices, and the lack of time to analyze the numbers push many hospitality operators to run their businesses almost blind. And it is precisely here that artificial intelligence begins to prove that it can be much more than a tech trend: a tool capable of detecting margin erosion, anticipating needs, and helping to make better decisions.
We spoke with Daniel Fontanet, cofounder of MindChef, the startup that won our First AI Projects for SMEs Contest. In this interview we analyze why some of the best-selling dishes can also be among the least profitable, how technology can free the restaurateur from much of their administrative burden, and what role AI will play in the restaurants of the future. Because, as MindChef argues, the goal isn’t to replace the professional’s experience, but to help them stop putting out fires constantly and focus on what really matters: making their business profitable and sustainable.
MuyPymes: You say that a large part of hospitality is still managed “by eye.” Where do you detect today the greatest lack of data: costs, margins, purchases, waste, demand forecasting…?
Daniel Fontanet: There’s something we see constantly when we start talking with a restaurant: the obsession with controlling the POS. How much have I billed today? How many tables did I have? Which dishes did I sell the most? What was my day with the highest revenue?
Of course, those are important data. But many times we are looking at the tip of the iceberg while underneath there is a much deeper problem. You can know exactly how much you billed today and not know how much money you really made.
The data that matters most isn’t necessarily another data point, but the ability to relate data and turn it into decisions: how much it truly costs to sell each dish, what’s happening with my margins, how costs are evolving, which suppliers are changing terms, where I’m seeing shrinkage, or what’s ceasing to be profitable.
And there’s another problem: many restaurants have that information, but it’s scattered. It’s in the POS, in an invoice, in a delivery note, in a spreadsheet, in the owner’s head, or nowhere at all.
The delivery note is a good example. In many restaurants it’s reviewed once a month, if there’s someone administrative who can do it. In other cases, the driver hands it over and it ends up in a drawer or even in the trash.
A delivery note today can tell you that the dish you’re going to sell this evening no longer leaves you the same margin as yesterday. That’s why we believe the big challenge in hospitality isn’t having more information, but stopping to manage the business by looking only in the rear-view mirror.
MuyPymes: One of the sector’s big problems is knowing the real cost of each dish. To what extent do restaurants today know the real margin of what they sell, and what mistakes do you see most often?
Daniel Fontanet: There are restaurants that have their dish costs very well controlled, but we also encounter many where the cost of a dish is a snapshot of a moment that no longer reflects reality, usually since they opened.
A supplier can change conditions without warning. A product can rise in price, its format can change, or shrinkage can increase. And there is an especially important factor: kitchen staff turnover. If you change the cook, you can change the way a dish is prepared, the quantities, the use of the product, or even the size of a portion. And when that happens, often there isn’t a data point that immediately tells you: “Something has changed here.”
There’s also a very common error: thinking that selling a lot means earning a lot. We meet restaurants obsessed with knowing their top-selling dishes, when the truly important question is: which dishes contribute most to the business’s profitability?
Because you can have a dish that sells constantly and, after subtracting all its costs, leaves you with almost no margin. And then an uncomfortable but necessary question emerges: how many hours are we working to sell things that barely make us money?
MuyPymes: Raw material prices can change constantly. How can AI help quickly detect those variations and prevent a restaurant from continuing to sell a dish that has stopped being profitable?
Daniel Fontanet: This is where technology starts to have very practical value. You don’t need to wait to close the month to realize something went wrong. The problem is that traditionally many management processes are designed precisely like this: you collect information, someone reviews it when they can, and by the time you have time to analyze it, the problem has already occurred. AI can help us accelerate that process dramatically, detecting changes and connecting them to the business.
If a supplier changes terms without notice, if certain ingredients rise in price, or if the true cost of a preparation changes, the system can detect that change affecting certain dishes and alert the person responsible so they can act as soon as possible.
And that changes the question. It’s no longer just: “What did I sell today?” , but “What changed today in my business and what should I do about it?”. In addition, predictive models already exist capable of anticipating certain purchasing needs using historical data and consumption patterns. The next step will be to cross much more information: demand, weather, calendar, events, historical behavior and price evolution.
External factors can even end up impacting management. A geopolitical situation can affect a raw material weeks or months later. Technology can help us anticipate it. But it isn’t about an AI automatically deciding to raise a dish’s price. It’s about the restaurant having enough information and lead time to decide. The competitive advantage won’t be just knowing what’s happening. It will be knowing in advance what could happen before others.
MuyPymes: MindChef uses AI to automate tasks such as costings, updating costs, or business analysis. Where is the line between automating management and preserving the chef’s or owner’s judgment and experience?
Daniel Fontanet: For us, the line is very clear: technology should not substitute the restaurateur; it should give them better tools to decide. But before talking about AI, there’s a preliminary step that is hardly discussed: getting the house in order. We see many restaurants wanting to adopt software, automate processes, or start using AI, but they still have trouble organizing their own information and internal processes.
In hospitality it often seems you can never stop. There’s always something urgent, a problem to solve, a supplier to call, or a fire to put out. We live in a constant stress cycle, as if running around all day like a headless chicken were part of the job. If you don’t have order, technology can help you do things faster, but not necessarily better.
First you need to understand what’s happening in your business, have at least minimally organized processes, and know what information you need to make decisions. From there, automation and AI can multiply your management capacity.
And here I would add another note: we like to talk about AI because it’s an extraordinarily powerful technology and we’re living in a moment of enormous innovation, but not everything has to be AI. Sometimes what we need is to connect information, automate a process, or accelerate a task that used to take hours.
Because there’s something that AI, by itself, doesn’t know: the context. The chef knows their kitchen. The owner knows their customers, their neighborhood, their team, their suppliers, and the identity they want to build. They have accumulated experience that a system would hardly understand unless we give it that context.
Will we someday reach systems capable of understanding and managing more and more aspects of the business? Surely. But our goal isn’t to replace the restaurateur’s judgment, but to help them use their time and experience better.
First, get organized. Then, automate. And always, decide with information. That’s our approach: so the restaurateur can continue to be a restaurateur, with their experience and judgment, but without having to spend every day putting out fires or making major decisions blindly.
MuyPymes: One of the big arguments for digitization is reclaiming time. Which administrative tasks should disappear or be dramatically reduced in a restaurant thanks to AI in the coming years?
Daniel Fontanet: Here we first need to understand how many restaurant owners live their days. They’re basically 24 hours a day focused on the business. They’re constantly putting out fires: an employee missing, a supplier not arriving, a customer, a breakdown, a kitchen incident, an invoice, a purchase…
The problem is that they end up extinguishing fires they shouldn’t be extinguishing. Small administrative tasks that seem insignificant but that, when accumulated, can consume hours each week: reviewing information, comparing costs, checking invoices, updating data, analyzing why a margin has changed, or looking for where a deviation lies. That time carries a huge cost. That’s why, for us, the goal isn’t for the owner to spend more time in front of a computer. It’s exactly the opposite.
We’d like them to devote 10 to 20 minutes a day to review the truly important information and make quick decisions. Something like: “This has changed.” “This isn’t profitable.” “This supplier has changed their terms.” “Here you’re losing margin.” “This deserves your attention.” Technology should make the owner extinguish the fires they truly need to put out and stop wasting time on the small administrative fires that a machine can detect, analyze, and prepare much better.
MuyPymes: You’ve worked with restaurants that discovered some of their best-selling dishes didn’t have the margins you thought. To what extent can better data change a menu, a pricing policy, or even the business model of a restaurant?
Daniel Fontanet: It can change a menu, a pricing policy, and, in some cases, it can change the most important decision of all: whether it’s worth continuing with that business. When you start to see the real numbers, you sometimes discover the problem isn’t only how the restaurant is run, but that the business model itself isn’t viable with its current costs, prices, and structure.
You may have a menu that’s too complex, overly high structural costs, certain dishes that require too much work for the margin they generate, or a pricing policy that no longer makes sense with current costs. And that can force you to rethink everything. We’ve encountered businesses that, after analyzing their situation, decided to close because the model wasn’t viable.
And while it may seem paradoxical, for us it’s also a positive outcome. Because our goal isn’t for a restaurant to remain MindChef’s client at any price. Our goal is for it to have information to make the best possible decision.
That restaurant stopped being our client, but it recommended us to peers who were going through similar problems. To me, that says much more about a tool than any sales figure. Because profitability isn’t about selling more. It’s about building a business that makes economic and financial sense.
MuyPymes: There’s a perception that tools like this are mainly aimed at large chains. How can AI be truly useful, simple, and profitable for independent restaurants as well?
Daniel Fontanet: Precisely, the independent restaurant is perhaps the one that can benefit the most from this technology. A large chain may have financial departments, analysts, buyers, or people dedicated to studying the data. In an independent restaurant, it’s often the owner who has to assume all those roles, and they don’t have time to learn about technology.
They want to run their business with peace of mind. They want to know their numbers are in check, that if something changes they’ll hear about it, and that they can make decisions based on reliable information.
The technology has to adapt to their way of working and disappear from their day-to-day life as much as possible. We don’t want the restaurateur to have to learn how an algorithm works. We want them to receive a clear answer to a simple question.
And here I think lies one of AI’s great opportunities: democratizing capabilities that until now were reserved for companies with much larger teams and resources. The best technology for an independent restaurant won’t be the one with the most features. It will be the one that makes the owner think: “At last I know what’s going on in my business without having to devote half my life to figuring it out.”
MuyPymes: Looking five years ahead, how do you imagine restaurant management? Will we reach systems capable of forecasting purchases, demand, costs, waste, and profitability and recommending to the restaurateur what decisions to make each day?
Daniel Fontanet: I think we’ll go much further than we can imagine today. In fact, some of those predictive capabilities already exist. We can use sales history and consumption patterns to anticipate certain purchases and needs. But I think that’s only the beginning.
The real leap will come when we can predict more precisely what demand a restaurant will have and what factors can influence it. What the weather will be like. If it will rain. If there’s an event nearby. How demand has behaved historically in similar situations. Which products will be needed in greater quantities and how their prices are evolving.
All that information can help the restaurateur anticipate: buy smarter, adjust needs, reduce waste, and avoid both excess product and shortages during peak demand. In short, tailor costs to what the business will truly need, rather than reacting after the problem has already occurred.
There’s a lot of talk about dynamic pricing, as in aviation or hotels. Personally, I don’t think directly translating that model to hospitality is the solution. A restaurant customer is especially price-sensitive, and constantly changing prices can breed distrust and hurt loyalty.
I think technology has a much more interesting opportunity: helping the restaurateur know their customers better, build loyalty, and generate more volume, all while keeping costs and margins in check. Knowing what customers want, when they return, which products perform best, and how to tailor the offering can help sell more, but in a financially sustainable way.
I’d like software to truly become the restaurant’s back office. Not another program the owner has to open and consult, but a system where the entire business is connected and capable of anticipating what may happen and explaining what should be reviewed. A system you can ask: “How are we doing?” And it answers: “The projected demand for this weekend is higher than usual. These are the purchases you should review”; “This product is rising in price and these are the recipes on your menu that could be affected”; “You’re buying more of this ingredient than you need”; “This dish is selling well, but its margin is shrinking”; “If you continue this trend, you could face profitability problems in a few weeks.” And, later on, that it can also suggest concrete actions to anticipate those problems.
No, I don’t imagine a restaurant run by robots, nor a business where prices shift constantly based on an algorithm. I imagine something much more interesting: a restaurateur who understands their demand better, controls costs, and has behind them a system capable of predicting what’s coming and helping them decide.
