More than 10 years ago you have already been working with IBM to help build the culinary generative AI system “Chef Watson”. Back then, what learning were you most excited about?
What excited me most was the realization that AI could expand creativity rather than replace it. When we were developing Chef Watson with IBM, the real breakthrough we were after wasn’t about getting a computer to write a recipe (which they are still not great at, by the way). We were focused on finding out if a computer could help humans to be more creative. On my first day in the kitchen with Chef Watson I realized that it could help us to see patterns in flavor, to reveal relationships between ingredients that many had never considered. It was analyzing thousands of ingredients, hundreds of thousands of compounds and instantly identifying hidden connections that would otherwise take hours of research. This meant we were churning out combinations that were scientifically sound but culinarily unexpected.
I remember the first time Watson suggested a pairing that made me stop mid-sentence — not because it was weird, but because it was right in a way I hadn’t anticipated. That’s when it clicked for me: AI shouldn’t be a replacement for creativity. It should be a catalyst. The best moments weren’t when Watson made a recipe. They were when it helped me see my own craft from a fresh angle.
Live at the GDI food conference
James Briscione is the co-founder of CulinAi, a best-selling author and leading voice at the intersection of food, technology, and artificial intelligence. On 18. June he will highlight the use case of AI in culinary at GDI.
Cooking is very sensory. How is it possible to quantify flavour and texture to train an AI model?
Cooking is sensory. It’s heat, aroma, sound, texture, timing. Over years in the kitchen it becomes intuitive. But it’s also surprisingly structured once you zoom out. Flavor can be partially quantified because aroma molecules are measurable. When two ingredients share volatile compounds, they often share perceptual similarities. That gives us a chemical map. Texture, too, can be described in terms of fat content, water activity, protein structure, starch behavior, temperature transformation. Many parts of the food we eat are quantifiable.
But here’s the key: quantifying isn’t the same as fully capturing. And AI systems can only be as good as the data they have access to. AI can’t taste. It can model relationships. It identifies probability. It predicts compatibility. What it cannot experience is context — the memory of a grandmother’s kitchen, the emotional resonance of a dish served at the right moment.
So we use quantification as scaffolding. The human chef still provides meaning.
With what kind of flavours or textures does AI struggle the most when it comes to recipe development? What remains irreducibly human?
AI is great at generating options and modeling patterns. It’s less great at knowing when to stop. The hardest thing to model is restraint, which is also, unironically, one of the hardest things to teach young cooks. Experience tells you when a dish is finished, when you’ve crossed the line where “more” is simply more, not better.
Flavors like bitterness, fermentation, funk, smoke, and heat live in that same gray area. They’re experienced, not prescribed. They aren’t universally “good” or “bad.” They’re contextual. You can add them to almost anything, but that doesn’t mean they improve everything.
Texture is even more human. The difference between tender and mushy, crispy and burnt, silky and greasy, those are judgments you make with your hands and your memory. It’s also the instinct to coax cohesion out of a set of ingredients. A system can tell me pumpkin, chile, and cocoa are a strong combination, and it might be right on flavor. But if the dish eats like a pile of mush, no one comes back for a second bite.
What remains irreducibly human is intent. Who are we cooking for? What do we want them to feel? What story is the dish telling? AI can suggest. Humans decide what matters.
You are about to launch an AI-powered culinary platform turning precision nutrition into personalized meals. What are the biggest advances AI made over the years that benefit the food industry most?
The biggest leap has been context. Early systems were mostly about rules and static outputs. Today, AI can hold multiple realities at once — goals, preferences, medical constraints, time, budget, what you already ate today, what’s actually realistic on a Tuesday night.
That’s a massive shift for nutrition and for food businesses. It lets us move from generic advice to specific, timely, and personalized guidance. It’s not “eat more protein,” but “you’re about 30 grams short today, and here are two dinners that get you there without feeling like punishment.”
For the industry, the benefits are both operational and creative: better forecasting, less waste, smarter product development and faster decision that feels chef-driven but also practical.
The thing I’m most excited about now is that AI has become genuinely collaborative. It’s not just analyzing, it’s helping us make better decisions in real time. But only if we know how to ask the right questions or work within systems that are purpose built to solve specific problems. GPTs have an incredible introduction and really helping to drive adoption but true, sustainable solutions will come from the purpose built models.
Where do you see the biggest potential for AI in the kitchen? What can AI truly improve for individuals but also for companies in the food industry?
I think that AI’s potential in the kitchen is to bring us better food, faster. It won’t help roast your chicken in 15 minutes instead of 45, but it might guide you to roast 2 chickens on Sunday and develop a plan for you to utilize all of that chicken throughout the week for quick, healthful meals that are custom built for your flavor preferences and dietary needs.
Most people don’t fail at healthy eating because they lack information. They know exactly what they should be eating. They fail because decision fatigue is real. AI can simplify choices at the moment they matter. What should I have for dinner tonight, how to adjust after a heavy lunch, how to adapt a childhood favorite for a new health reality.
For companies, AI can unlock intelligent personalization at scale. Imagine a restaurant group that understands regional flavor preferences in real time. Or a retailer that reduces food waste by predicting demand down to the store level.
In professional kitchens, I see AI becoming a behind-the-scenes partner: helping chefs explore new flavor pathways, pressure-test menu concepts, and accelerate R&D — while leaving the actual cooking, judgment, and hospitality where it belongs: with people. The best kitchens won’t become less human. They’ll become more focused on what matters.
What will the future hold? Might Michelin-starred kitchens use AI agents as sous chefs? Will supermarkets personalise your shopping list in real time based on your microbiome? Or do you picture something else entirely?
Both of those are very likely, just not in the way most people imagine. In Michelin-starred kitchens, AI agents will not be searing your steak or stirring your risotto. I see them more like an always-on research assistant. They can streamline daily workflows, help optimize prep and staffing, spot seasonal opportunities, track guest preferences, model menu flow, and help teams iterate faster without sacrificing the human judgment that great cooking depends on.
On the retail side, we are already moving toward personalization, although today it often shows up in the least inspiring form, like ads that follow you around the internet. The more interesting future is when biometrics, microbiome insights, purchase history, and dietary goals can genuinely inform better daily decisions. The real breakthrough is not just information, it is trust. When the system earns enough trust to make a few decisions for you, it gives you time back. That is where this becomes meaningful for real life.
The future I’m most interested in is quieter. We will know we have succeeded when AI becomes invisible and genuinely helpful, sitting between our intentions and our daily choices, nudging us toward better meals without turning food into a math problem.
Because food has never just been fuel. It is memory, culture, identity, pleasure and meant to be shared. If AI has a role here, it is not to optimize the joy out of eating. It is to protect it by making good food easier to choose, easier to cook, and more personal than ever.
«Recoding Food»: The food conference at GDI
On 18 June, James Briscione will connect technology and cooking at the International Food Innovation Conference at the Gottlieb Duttweiler Institute, demonstrating how AI makes delicious food more personalised and healthier. Experience inspiring keynotes from internationally renowned experts and join the discussion on how AI is becoming the new operating system of the food industry.