IN Brief:
- Nestlé is using AI to analyse scientific literature, consumer behaviour, and a database of roughly 120,000 recipes.
- Development work addresses protein intake, lean-muscle loss, hydration, and other nutritional issues associated with rapid weight loss.
- Existing products provide a commercial base while AI tools are being used to accelerate formulation and consumer testing.
Nestlé is expanding its development work for consumers using GLP-1 weight-loss medicines, combining artificial intelligence with nutrition science as appetite-suppressing drugs begin to influence formulation priorities across food and beverage manufacturing.
The company is using AI systems to analyse scientific literature, model consumer behaviour, identify potential nutrient combinations, and search a database of roughly 120,000 existing recipes for reformulation opportunities. The work builds on products Nestlé has already developed around weight management and high-protein nutrition rather than starting a new category from scratch.
Chief technology officer Stefan Palzer has said the company is studying nutritional effects associated with rapid weight loss, including loss of lean muscle mass and reduced intake of essential nutrients. Nestlé is consequently investigating combinations of protein, micronutrients, hydration support, and other nutritional components that could be incorporated into foods and beverages aimed at consumers eating considerably less than before treatment.
The formulation problem is more complicated than increasing the protein number on a nutrition panel. Consumers whose appetite is strongly reduced may eat smaller portions and fewer meals, which increases the amount of nutritional work each serving is expected to perform. Protein concentration, fibre, vitamins, minerals, texture, flavour, digestive tolerance, and portion size can all become more important when less food is supplying a larger share of daily intake.
AI moves deeper into formulation
Nestlé’s AI systems are intended to reduce the amount of development work required before a promising formulation reaches a physical trial. Internal tools can compare existing recipes, published scientific work, ingredient combinations, and consumer information, allowing development teams to screen a larger number of options before committing laboratory and pilot-plant time.
The company’s scale gives those systems a substantial data base. A recipe archive containing roughly 120,000 formulations can include information accumulated across product development, ingredient substitutions, processing conditions, sensory performance, and commercial launches. Palzer’s argument is straightforward: machine-learning systems become more useful as the quantity and quality of the data available to them increases.
The output still has to survive ordinary food engineering. Raising protein can increase viscosity, alter heat stability, introduce sedimentation, or change flavour and mouthfeel. Fibre and micronutrients can create their own sensory and processing effects, while a reformulated product must continue to tolerate mixing, thermal treatment, filling, transport, and shelf life without drifting outside specification.
Nestlé has already established several products around the emerging GLP-1 market. Its US business has developed Boost Advanced Nutrition Shake with 35g of protein, while the Vital Pursuit portfolio was created around portion-controlled and protein-focused foods for consumers managing weight. Other markets have seen higher-protein formats, including Milo PRO High Protein.
Those products provide manufacturing experience as well as market data. Ready-to-drink nutritional products have to deliver high protein concentrations without excessive thickening or instability during heat treatment, while frozen foods designed around smaller portions still have to meet expectations on texture, flavour, satiety, and value.
Nestlé is also investigating more targeted nutritional combinations. Palzer told Reuters that researchers had industrialised a combination of two micronutrients intended to support muscle tissue alongside protein intake. The company is separately working on proprietary ingredient combinations associated with appetite after GLP-1 treatment ends. Those programmes remain company-led research and should not be read as proof that a particular finished food will produce a defined medical outcome.
The commercial pressure behind the work is clear. GLP-1 medicines reduce appetite, creating concern among large packaged-food producers that some consumers may buy fewer snacks, meals, and drinks. At the same time, smaller portions create demand for products that deliver more protein and nutrients per serving, giving manufacturers an opportunity to redirect portfolios rather than simply accept lower consumption.
That shift is already affecting product-development briefs beyond foods explicitly marketed to people taking the medicines. Higher protein, greater fibre content, smaller portions, and increased nutrient density fit wider health and wellness trends, meaning work prompted by GLP-1 adoption can migrate into mainstream categories without a drug-related claim appearing on the packaging.
Regulation places practical limits around that strategy. Food businesses cannot treat a conventional product as a medicine, and health claims have to meet the requirements of individual markets. Manufacturers will therefore have to separate clinically relevant nutrition research from the claims that can legally and credibly appear on consumer products.
Cost is another constraint. Protein-rich formulations and more complex micronutrient systems can raise ingredient costs and require additional sensory trials, stability testing, packaging changes, and production validation. A specialised nutrition product may support a premium, but that economics becomes harder when similar formulation ideas are moved into high-volume mainstream foods.
AI can shorten parts of the route by reducing the number of unpromising combinations that reach the pilot stage, but it does not remove the physical constraints of food manufacturing. Ingredients still have to mix, pump, heat, cool, fill, freeze, dry, or bake as required, and the finished product still has to remain acceptable after distribution and storage.
Nestlé is therefore using GLP-1 adoption as both a nutrition-development problem and a test of its digital R&D capability. The size of its formulation archive gives AI more material to work with, but success will ultimately be measured on ordinary factory terms: repeatable processing, stable products, acceptable taste, compliant claims, and economics that continue to work when development moves beyond specialist ranges.
