This is the first part of our three-part series on AI in the food industry. Check out Part II: Operations and Part III: Consumer, along with our latest Future of Food Series: Food x AI video that ties the three parts together.
Bluestein’s Take
Bluestein invests in technologies that enable a better, healthier, and more sustainable food system. One such technology, artificial intelligence, is poised to transform the food industry. In a three-part series, we’ll delve into AI’s potential throughout the industry starting with food formulation then moving to operations and the consumer – identifying the most compelling startups across each segment.
While there has been a lot of hype surrounding AI over the past decade, recent changes in consumer behavior, regulation, and technology make it ripe to invest in AI for the food industry.
Shifting consumer behavior is driving investment in AI for food, as modern consumers demand better food products and are influenced by ever-evolving social media trends, requiring faster innovation.
New regulations around data privacy and food safety are pushing food companies to adapt quickly, accelerating the need for AI solutions.
From a tech perspective, to-date, AI within food has been constrained by limited data availability. However, access to better datasets and breakthrough technologies such as Generative AI and AlphaFold are enabling sophisticated decision-making and rapid innovation for the food industry.
Conventional food R&D methods, relying on trial and error, limit innovation by consuming resources and time, often with high failure rates and waste. We see an exciting opportunity for AI within food formulation and believe the best solutions deliver on the triumvirate of health, taste, and price. These solutions should leverage high-quality data, be scalable, and easily integrate into existing systems.
AI’s Potential to Transform the Food Industry
Artificial Intelligence (AI) holds immense promise to fundamentally alter the way we live, work, play – and eat. The food industry is facing pressures from all sides, and AI is poised to improve the supply chain from farm to fork; it has the potential to bring unprecedented levels of efficiency, precision, and innovation to create a more resilient, responsive, and responsible food system that can better meet the demands of consumers. The market opportunity is ripe across the value chain due to convergence of technology advancement, regulation, and consumer behavior.
Technology
Generative AI has the potential to be a transformative technology in terms of its potential impact on global GDP and labor productivity.1 AI is already reshaping the business landscape, revolutionizing how companies address complex challenges across various industries. In fact, Goldman Sachs predicts that investment in AI could reach $200B by 2025.2 What makes AI so powerful in business is its advanced pattern recognition techniques, allowing companies to more easily make sophisticated decisions based on large, multivariate datasets. These techniques span from traditional statistical models and machine learning regression methods, which excel in numerical computation, to AI systems that can emulate human judgment and reasoning.
Only in recent years has AI reached more mainstream adoption across business and consumers – in no small part due to the launch of ChatGPT in 2022. Within one month of release, ChatGPT had 100 million active users, making it the fastest growing application in history.3 The technology has continued to advance rapidly due to the combination of computing power, data availability, and improved algorithms.4 The pace of new model releases has accelerated, with OpenAI’s GPT-4 coming out only a few months after ChatGPT. GPT-4 demonstrated a significant leap in performance compared to earlier models, achieving 86% accuracy on the Massive Multitask Language Understanding (MMLU) benchmark, which approaches the 89.8% accuracy of human experts.5 There has also been an exponential increase in the computation used to train AI models over the last decade that has enabled more complex and capable AI systems.6 AI systems have surpassed human-level performance across a myriad of domains, such as handwriting, image, and language comprehension, in a few years – which makes it highly versatile and applicable across many fields.
ChatGPT and other generative AI tools are one of two catalysts accelerating adoption of AI in the food industry. The other? AlphaFold, which achieved a breakthrough by enabling AI models to "learn the language of proteins," predicting the biological and chemical functions of proteins based on their amino acid sequences. Developed by DeepMind in 2020 and later acquired by Google, AlphaFold has broad applications in drug and enzyme discovery for diverse uses like recycling plastics and discovering novel proteins with functions such as binding or sweetness for food innovation.78
The impact of AlphaFold underscores AI’s potential within the food industry. Food, which is often siloed into operations like R&D and marketing, has the power to be unified by AI, enabling rapid innovation through quick data analysis and prediction modeling. Example applications include developing new recipes, optimizing distribution logistics based on weather patterns, and scaling production processes – seamlessly handling multiple data inputs that would be impossible for human analysts alone. As companies collect more data from their operations, supply chains, and consumer interactions, AI models can process this information to generate increasingly refined insights, meaning that AI is inherently scalable, which is perhaps its greatest strength.
This scalability makes AI in the food industry particularly compelling and ripe for adoption. While it’s still early, food retailers are already turning to AI to optimize parts of their business. In fact, according to The Food Retailing Industry Speaks 2024 report, 41% of retailers surveyed say they are using AI for parts of their business – double the number from just one year ago.9
We’ve also increasingly seen almost every major food conglomerate implement AI either through in-house capabilities or a strategic partnership. A few examples include:
PepsiCo: Using AI within its Frito-Lay unit to better understand consumer preferences and develop new flavors.10
Unilever: Utilizing AI solutions for food development, starting with consumer insights and formulating for taste, texture, and shelf life.11
McDonald’s: Acquired DynamicYield to provide a personalized ordering experience at its drive-throughs.12
Starbucks: Launched Deep Brew, its proprietary AI platform that analyzes extensive data to personalize customer experiences, offer tailored marketing messages, and customize menu recommendations.13
Kraft Heinz and Ingredion: Partnered with NotCo, a startup leveraging AI to develop plant-based food formulations.1415
Consumer Behavior
What’s fueling the corporate race to adopt more AI? The rapidly changing demands of today’s consumers and the need to keep pace. Modern consumers are increasingly discerning about their food choices and demanding more from food products – seeking higher nutrient density, more functional benefits, and minimally processed ingredients.
With more constraints on time from work, commuting, and personal commitments, it’s no surprise that convenience is also top of mind for consumers. Consumers strongly prefer food and beverage options that require little preparation, with one study citing over 60% of consumer morning and midday meals prepared in under five minutes.16 Companies are under pressure to innovate and deliver food solutions that cater to these hectic lifestyles, as well as capture limited consumer attention.
Social media has also significantly influenced expectations regarding the pace of food innovation. Modern consumers are more fickle, eager to try the latest food trends that go viral on TikTok or to incorporate new products endorsed by their favorite influencers. This has created a need for companies to respond swiftly with both products and content that resonates. Target's Chief Merchandise Officer recently echoed this: "What's changed recently is the pace of change getting faster and faster, driven by social media and platforms like TikTok…We are constantly listening to our guests and listening to trends. We have a team that’s monitoring social media, looking for what’s new, what’s trending, and try to capitalize."17
Harvard Business School estimated that 30,000 new food products are launched every year with a failure rate of 80%.18 This rapid introduction of new entrants has only reinforced the need for companies to maintain a competitive advantage. By leveraging AI solutions, companies can quickly analyze and forecast trends, target and engage consumers, optimize product development, and address the evolving needs of their customers.
Regulation
Beyond shifting consumer tastes, regulatory tailwinds are creating more urgency for companies to rapidly adapt. The Food Safety Modernization Act (FSMA) is one of the most comprehensive reforms of federal food safety laws in decades and has a compliance deadline approaching in January 2026. FSMA requires that food companies adhere to stricter safety standards, enhanced traceability, and increased testing and documentation. Companies must quickly implement changes to meet the new standards.
California recently adopted the California Food Safety Act, which will prohibit four food additives with known negative health effects starting in 2027.19 Other states have followed suit and introduced their own similar legislation, while the FDA is reviewing more sweeping regulation at a federal level for chemicals.20 In July 2024, the FDA also announced a ban on the use of brominated vegetable oil (BVO) in food, a commonly found ingredient in sports drinks and sodas such as Keurig Dr Pepper’s Sun Drop, requiring companies to reformulate any existing products with BVO to comply.
Along with new food directives, data privacy regulations have had an impact on food companies. These regulations, such as the General Data Protection Regulation (GDPR) in Europe, the California Consumer Privacy Act (CCPA) in the United States and Apple’s App Tracking Transparency (ATT), have imposed stricter controls on how companies collect, store, and use consumer data. Previously, this data was essential for understanding consumer preferences, behaviors, and trends. Companies now need to adjust their approach for targeting consumers and leveraging insights, creating an opportunity for AI solutions to reach and engage consumers while respecting user privacy.
Introducing Food.AI
Given the hype around AI, it can be hard to distinguish true potential from the noise. In a three-part blog post series, we will explore the opportunities for AI across the value chain within the food industry beginning with food formulation, followed by a deep dive into operations, then the consumer. The chart below outlines how we define each of these segments.
In Part I, we focus on how AI can enhance food formulation to meet evolving consumer preferences around health and taste as well as new regulations. Our next installment on operations will address the process of bringing products to consumers, from ingredient sourcing and inventory management to manufacturing, food safety, marketing, and distribution. Part III will cover the consumer, how AI can be leveraged to better understand and target consumers including insights, marketing, education, and buying behavior.
Taking a Byte out of Food Formulation
Today's food companies face a nearly impossible dilemma: launching products that offer better nutrition and functionality while also maintaining strong profit margins. They must continually innovate to keep up with ever-changing consumer preferences and the rapid pace of new product innovation.
One of the biggest limiting factors to innovation today is that traditional product R&D relies on experiment-based trial and error methods. These experiments typically test only incremental variations and consume large quantities of ingredients. Considering the high failure rate of new products, R&D today has a substantial cost – not only capital and substantial food waste, but also time and brand reputation.
This is where the power of AI shines: rapid, cost-effective formulation could lead to breakthrough innovations. AI's capabilities in product R&D are extensive, covering everything from ideation to product finalization. For example, AI can generate novel product ideas by analyzing datasets on flavor-pairing, social media images, and sensory responses, and can accelerate new ingredient discovery by screening molecule datasets. Within manufacturing, AI can simulate processes to maximize production yields and design efficient downstream methods. Finally, AI methods are comprehensive enough to handle the complexities of creating novel foods like healthier carbohydrate products, plant-based cheeses, and alternative proteins – optimizing for intricate sensory preferences, including taste, texture, and aroma.
The key to unlocking the space? The increased availability of high-quality datasets and advancements in sophisticated algorithms for data processing. In the past, companies were constrained by low-resolution and costly data, limiting the scalability and effectiveness of AI methods.
AlphaFold is an example of high-resolution data for protein modeling, which could be replicated across additional food molecules. Another significant advancement is the publicly accessible Periodic Table of Food Initiative (PTFI) database, funded and supported by the Rockefeller Foundation and other research and federal agencies.21 This database provides detailed molecular data on thousands of foods, collected globally from diverse samples. PTFI has the potential to catalyze and incentivize emerging startups to leverage high-quality data, enabling them to make informed decisions before investing in their own costly proprietary datasets.
Headwinds
While AI is rapidly advancing, it’s still a radically new technology and there are both technical and production challenges that need to be addressed before it can be reliably integrated into the food production system.
Technical Challenges
Comprehensive Data: The analytical methods to detect and characterize molecules in food are still in development, and some molecules can be better characterized than others.
Rapid Technology Evolution: AI innovation moves quickly. As new methods of data generation and novel algorithms emerge, older data models must evolve or risk becoming obsolete. This poses a challenge for startups that don’t have the resources to continually upgrade to the latest and greatest.
User Interface: AI solutions must have a user-friendly GUI that integrates seamlessly into the product development cycle as a decision-making tool for the R&D team.
Production Challenges
Chasm from ingredient discovery to product: Solutions that focus solely on ingredient discovery encounter the challenge of formulation. The complexity arises because food molecules are not simply interchangeable, often requiring additional modifications to fit into recipes. In addition, many molecules that offer functional or health benefits may have undesirable sensory properties or unknown formulation functions.
Integration: The datasets for ingredient discovery are not yet integrated with datasets for scalable production of said novel ingredient and/or integrated with a solution to formulate a product using it. Therefore, any prediction still has practical challenges. For example, a protein that is created through generative AI might be challenging to realistically produce at scale or include in a product, as formulating food products is highly complex.
Scale-up: Companies looking to scale a novel molecule or ingredient may encounter issues along every step in the process from strain engineering to downstream processing.22
Market Segments
To better understand the landscape, we created a market map of companies across several pillars, each housed under the umbrellas of ideas & ingredients or product manufacturing.
Ideas & Ingredients
Ideas and Ingredients encompass solutions that support the foundation of a product across two main objectives: idea generation and ingredient discovery.
Idea Generation: AI companies in this category focus on the earliest stage of product development – identifying novel flavor profiles and product concepts. Companies spur ideas based on data reflecting consumer interest on social media, as well as more specialized data such as flavor pairing hypotheses and product-specific sensory ratings. For example, Kellogg’s recently leveraged two tools for idea generation – Foodpairing to launch a new Pringles flavor and AI Palette to identify new breakfast cereal concepts that appeal to its Southeast Asian customers.2324
Ingredient Discovery: Solutions in this space help food businesses identify ingredients that provide desired sensory and/or health properties. Given the vast array of ingredient, startups typically only focus on one class of molecules. Prominent classes include proteins and bioactive secondary metabolites.25 To date, protein discovery has been more prolific due to the breakthrough success of AlphaFold in 2021. For example, Digestiva discovers enzymes (a kind of protein) that maximize the nutrient absorption in humans, and Arzeda has discovered an enzyme that facilitates mass production of Stevia.2627 Within secondary metabolites, Brightseed, has discovered a fiber made from upcycled hemp hulls that supports gut health using its proprietary AI discovery platform.28
Ingredient Scale-up: Most novel ingredients today require the use of new production methods such as precision fermentation, which poses a challenge for scale-up because it is currently so expensive. AI solutions can leverage past data, both failures and successes, to help companies understand the complexities of scale-up, including strain engineering and downstream processing. This allows them to better predict the optimal parameters and steps, considering factors like low-cost feedstock and yield, with the goal of achieving cost parity with existing ingredients. For example, MK2 offers a low-cost filtration technology for proteins produced by precision fermentation, while Teselagen provides services ranging from engineering for protein expression to final purification.2930
Product Manufacturing
AI companies in product manufacturing encompass a diverse range of solutions. Some startups act as third-party vendors to food manufacturers (e.g. Turing, Benchling) while others are vertically integrated (e.g. Climax Foods, NotCo). Companies also use varied datasets, which range from high-resolution molecular data for ingredients to low-resolution experimental data for product development.
The most significant differentiation among solutions lies in their specific focus area; this focus is critical due to the complex chemistry of food formulation, where success is more likely if a solution targets a specific aspect. Focus areas may include particular types of ingredients, food products, or specific sensory elements such as texture, aroma, and taste in the final product. For example, Aromabit optimizes quality control and ingredient selection for aroma, while Grainge focuses on enhancing the texture of grain-based products like pastas and breads by incorporating fiber ingredients for health benefits.31 While solutions are typically focused on one aspect, to deliver on consumer needs, companies would need to knit these together to have the best result.
Evaluative Criteria
There is no question that AI will be a catalyst for innovation within food formulation. We believe that the best solutions in the space are those that provide:
Health + Taste + Price: Solutions need to ultimately lead to a better overall product that delivers on the holy grail of health, taste, and price that meet consumer demands.
Integration: End-to-end food formulation is a multi-step process involving ingredient discovery, scale-up, product R&D, and large-scale manufacturing. The best AI solutions integrate these parts for greater efficiency, enhancing connectivity across teams and optimizing concerns early in the R&D process. Often parts of this solution suite may already exist in-house, so it’s also important for external vendors to easily and seamlessly integrate with legacy systems.
Scalability: The best solutions are those that provide solutions capable of generalizing across various products and customers without the added layer of a human touch.
High-Quality Data: Reliable, accurate AI solutions require high-quality, abundant datasets. Fortunately, the success of recent AI models and the publicly available PTFI dataset strongly indicate that the necessary infrastructure for high-resolution data collection and AI modeling methods could now be a reality.
What’s Next
This is an incredibly dynamic space and we’re eager to see it continue to develop as it’s fueled by more investment, regulatory changes, and consumer behavior shifts. If you’re an entrepreneur in this space or know of one, we’d love to hear from you.
This is the first part of our three-part series on AI in the food industry. Check out Part II: Operations and Part III: Consumer, along with our latest Future of Food Series: Food x AI video that ties the three parts together.
Thank you to our PhD fellow, Tarini Naravane, for leading the efforts on this series.
Sources
The four substances that California is banning are brominated vegetable oil, potassium bromate, propylparaben, Red Dye No. 3, and titanium dioxide. Each of these additives have been have each been linked to serious health problems
Secondary metabolites are compounds that only exist in foods of a plant origin, for example polyphenols. These compounds have gained significance because of potential health benefits including immunity and treating inflammation