Bytes to Bites part one

Bytes to Bites part one

Why now?

In the face of mounting challenges, the food system is calling for the integration of AI solutions to pave the way for enhanced resilience and sustainability. This three-part deep dive begins with this article centered on new food product development, followed by two articles that will explore other pivotal areas within the food system, where the urgency for change and the potential for AI is high.

Each part unveils insights into AI applications, investments, challenges and opportunities, and the startup landscape.

Image: Peakbridge

Apart from presenting an opportunity to solve complex challenges across the food system, AI is also changing the overall business landscape. Companies are responding to consumer trends and bringing products to market faster than ever (think fashion brand Shein that launches 6,000 new items on its website every day), and consumers are starting to expect this. To keep up with trends and succeed in their go to market strategies, food innovation must become fast-paced.

2. The product development cycle of food corporates

The new product development cycle in food corporates from bench to shelf traditionally suffered from limited information and fragmented data. This complexity arises from distinct departments handling various facets of the cycle, including marketing, R&D and sales. These challenges result in slow decision-making and lengthy innovation cycles characterized by:

  • Trial-and-error approaches
  • High costs
  • Reliance on time-consuming human panels
  • Limited product-market fit

It therefore comes as no surprise that around 80% of food product launches fail, primarily due to consumer rejection. AI addresses these challenges through reducing the need for extensive trials and fostering cross-departmental collaboration using robust data networks. It can streamline the entire process by optimizing product formulation, process parameters and market trend analysis.

“For me, the whole digitization agenda is so relevant and so exciting because, if done well, it should truly move things much faster. Avoid a lot of tests and trial and error approaches which a classical R&D organization still has and go into predictability at a much faster pace.” Miriam Überall, former R&D executive at Kraft Heinz and Unilever

The new product development cycle. Image : Peakbridge
The new product development cycle. Image: Peakbridge

Let’s now dive a bit deeper into how AI can accelerate the different parts of the product development cycle.

3. The role of AI in accelerating innovation cycles

Consumer insights and idea generation

AI is reshaping new product development by leveraging a multi-faceted data-driven approach. Firstly, AI interprets real-time trends from external sources capturing information such as opinions and emotions. This involves analyzing social media, tracking keywords, using GenAI-powered chatbots for surveys, and AI’s image analysis for food preferences.

AI also extends to IoT sensors, gathering consumer data on product choices and cooking preferences. Furthermore, ML conducts predictive analysis internally,
leveraging historical sales data and market trends to accurately forecast consumer demand and preferences, optimizing new food product launch timings and adapting to market changes. This continuous data analysis uncovers consumer insights, fostering deeper engagement in a competitive, fast-changing food market.

PeakBridge’s portfolio company, Tastewise, is an example of a startup that is using AI to inspire new product development. The company has developed software that collects a vast amount of data from different sources (e.g., social media, reviews, menus, recipes) to understand evolving food trends and consumer tastes. This software is a valuable tool for food corporates as it helps them create consumer-desired and preferred products.

See below an example of the company’s most recently launched tool, ‘Concept Discovery’ by ‘TasteGPT,’  that discovers AI-generated concepts and images based on top-rising consumer demand pairings. The prompts provided for these examples were “coffee variations” (top image) and “fun crazy food” (bottom image).

TasteGPT from Tastewise

TasteGPT from Tastewise
TasteGPT is a generative-AI program from Tastewise designed to help users get personalized insights faster than ever. Image: Peakbridge

Two other companies working on consumer insights are Bite.ai and Gastrograph. bite.ai is a startup focused on food image recognition to improve food logging and generate data-driven customer insights; while Gastrograph has created the “world’s largest sensory database”, helping food corporates align their products with the flavors, aromas and textures that are driving consumer demand.

Discovery of new ingredients

Moving along in the new product development cycle, AI can also accelerate the discovery of new food ingredients, improving the screening and characterization of molecules. Startups working on the intersection of ingredient discovery and AI engage food and data science experts to develop an effective algorithm that supports food discovery pipelines. Examples include Ginkgo Bioworks and Arzeda, that use a combination of computational design and AI to create novel proteins and enzymes.

Amai Proteins uses computational biophysics and AI to design novel proteins optimized for various functionality and taste traits. Finally, Brightseed supports novel = ingredient development by identifying health-promoting bioactive compounds using their AI-powered platform Forager.

R&D and optimization

Further along in the product development cycle is the optimization of food matrix functionality —covering sensory and nutritional attributes. AI takes center stage in predicting and enhancing these attributes for diverse food products. It suggests ingredient ratios to match taste profiles and offers healthier substitutions while preserving flavor. Additionally, AI, in conjunction with IoT sensors and imaging systems, assists in assessing food product texture, ensuring characteristics like chip crispiness or meat tenderness meet expectations.

On the nutritional front, AI optimizes recipes to achieve specific goals, whether it’s reducing sugar content or increasing protein levels, and predicts the nutritional composition to align with labelling requirements.

Recently, food corporates have embraced AI in R&D cycles, reducing product development and processing time from months to mere days. Unilever employed AI to craft a low-salt bouillon, expediting flavor analysis from months to days. Kraft Heinz piloted an AI algorithm to optimize cost, sugar, and salt levels, achieving remarkable results. QDA (Quantitative Descriptive Analysis) factored in sensory characteristics like taste and mouthfeel, achieving a 94% accuracy in replicating the original tomato- based product.

Precision fermentation, which involves the manipulation of genetic material to produce proteins, is a good example of a recent food technology that is converging with data science. Gene-expression modelling, 3D protein configuration modelling, and simulations assist scientists in identifying genome changes that enhance the likelihood of generating specific proteins.

Without AI, conducting these experiments could take centuries. Emerging in the late 2010s, startups like Shiru and Imagindairy are examples of this convergence of innovation.

“We can look at cells as small factories for generating proteins: cells evolved for billions of years to replicate themselves very efficiently; thus, using gene expression is the most efficient way to generate protein. This process includes the introduction of the gene encoding the target protein into the genome of a host cell such that it will be translated to the target proteins by the intracellular machinery of the cell. To make this process efficient and compete with the conventional food industry, we need to re-design the ‘factory’ (i.e., the host cell). This is done via introducing ‘smart’-model based mutations into the genomes of the host and cannot be done without sophisticated algorithms, models, and AI.” Prof. Tamir Tuller, cofounder, Imagindairy

Scaleup: Optimization of yield and cost

After developing food products at a laboratory scale, the subsequent phase in the cycle involves establishing machinery and production lines for large-scale manufacturing. The goal is to ensure cost competitiveness while maintaining the same nutritional and sensory attributes observed during lab-scale production. This can be achieved through the utilization of ML algorithms that predict a food’s physical and chemical properties (e.g., pH during cheese and cocoa fermentation) and model processing conditions (e.g., antioxidant extraction or potato peel conditions).

Innovations in the cultivation of alternative proteins, such as lab-grown meat, exemplify AI’s role. Precise control over temperature, pH levels, nutrient supply, and growth factors is imperative in laboratory settings. Entrepreneurs and scientists face the challenge of upscaling manufacturing to meet the increasing demand for animal protein while minimizing the need to clear forests and wild vegetation for feed crops. AI offers a solution by analyzing extensive data to identify optimal conditions for scaling up production.

Pioneering startups such as Animal Alternative Technologies and Umami Bioworks are leading this field, developing intellectual property and scalable technologies with data science optimization. Another notable startup in this space is Eternal, which leverages AI and robotics to automate experimentation, analysis, and biomass fermentation optimization. These advances are also set to benefit large producers of animal protein seeking viable and sustainable pathways for transitioning to large-scale alternative protein production.

Startup-corporate partnerships

Startups, as compared to food corporates, benefit from their smaller size, and often experience swifter information dissemination due to reduced compartmentalization. Nevertheless, constructing suitable data infrastructure remains a challenge, given their limited financial and human resources. This can impede the commercialization of novel products in an appropriate time span.

Recognizing the value of startup advancements in fulfilling the demand for healthy, delicious, and sustainable foods, numerous food corporates are forging partnerships with startups. Examples include:

4. Startup landscape: New product development ecosystem

Since the late 2010s, we have witnessed a surge in startups dedicated to AI-driven food product development. Their expertise lies in providing AI-driven solutions for tasks such as market analysis, consumer insight forecasting, and predictive modeling for product and process parameters. In the figure below, various startups are shown, working on different parts of the new product development cycle.

We can distinguish between two types of startups here. This differentiation highlights two distinct trajectories within the food industry’s adoption of AI.

1. AI-centric service startups: facilitate corporate product development by infusing AI across different stages of their innovation lifecycle.
2. AI-integrated startups: use AI to enhance their own product development, introducing novel market offerings.

The startup landscape and product development ecosystem.
The startup landscape and product development ecosystem. Disclaimer: this overview is not exhaustive and only represents a selection of startups. Furthermore, the categorization of companies is non-exclusive, meaning startups can fall under multiple categories. Image: Peakbridge.

5. Challenges for AI in food

As previously explored, utilizing AI for new food product development offers a spectrum of advantages, including cost-efficiency, speed, customization, predictive capabilities, and data-derived insights.

However, several challenges emerge in this landscape:

Data quality, quantity, and connectivity (‘garbage in, garbage out’): Achieving optimal algorithm performance relies heavily on abundant, high-quality data. The output is only as valuable as the input data. This requires investments in data generation for R&D teams and in proper data infrastructures and connected data systems between departments in food corporates. Once data infrastructure is established, effectively utilizing it furthermore involves understanding existing data, integrating it into an overarching data architecture, designing data strategies, and aligning with objectives. Here, data infrastructure providers such as Google, AWS, and Microsoft play an enabling role.

Insufficient processing power: AI’s ability to handle extensive datasets is limited by processing power. Quantum computing holds the potential to address processing power limitations, with investments in infrastructure and research. Infrastructure companies like IBM, Google, and D- Wave are key players in advancing quantum computing for AI.

Limited historical data: An emerging field such as foodtech lacks historical data to feed algorithms, making it harder to generate meaningful results. If available at all, it is often found in many different and unstructured data formats. Developments are needed to make the input of relevant data more format agnostic.

Costs of implementation: Setting up and maintaining AI systems can be expensive, especially for small companies. Large corporates, on the other hand, have existing systems which might not be future-fit and therefore require significant investments to adapt.

Liability and ethical considerations: The increasing complexity of AI systems, especially in predictive and prescriptive applications, raises questions of liability and responsibility, challenging the legal and ethical frameworks needed to address potential AI errors and consequences. Additionally, assessing AI’s impact on jobs and traditional food cultures is crucial to understanding its overall effects.

Data protection and security: Protecting proprietary data, such as secret recipes, while promoting data sharing to optimize AI applications is a complex challenge that requires effective governance mechanisms (“you cannot untrain an algorithm”). Also, protection against digital attacks is crucial.

Regulatory adaptation: Food laws change often. AI systems must keep up. Plus, regulations often require interpretation – AI might not be fit for this (for now).

Multidisciplinary collaboration and skill sharing: Combining AI and food expertise requires effective communication among experts from different fields (think of food scientists, engineers, and data scientists). There is a need for accelerated skill sharing and building across departments to make integrated data-driven decisions.

Consumer acceptance: Building trust in AI-produced food requires education to address concerns and fears of consumers.

Environmental impact: Beyond efficiency, AI’s effects on the environment need to be considered and compared to the environmental mitigation benefits. Navigating these challenges is pivotal as the industry embraces AI’s potential while proactively addressing its limitations and societal implications.

6. Bottom Line – Paving the path ahead

In this article, we looked at the potent synergy between AI and foodtech, a transformative nexus addressing escalating food demand and sustainability imperatives. From inspiration for new food product design driven by data on consumer demand, to newly suggested process parameters that can improve sensory attributes, yield and reduce costs, AI optimizes every step of a food corporate’s new product development cycle.

In doing so, it has the opportunity to unite siloed departments and offer a new, data-driven directive that surpasses the traditional trial-and-error approach. Through AI, a novel paradigm emerges in new ingredient and product discovery, lab-scale innovation, large-scale manufacturing and in consumer engagement, enabling effective decision-making.

Agile startups merge with food corporates to accelerate this transition and thereby bridge the innovation gap – a trend that is expected to gain more momentum in the near future. Challenges in data quality, processing power and ethics emerge, yet AI’s promise of faster iterations and data-informed decisions penetrates the food industry. As we savor this harmonious interplay, AI guides foodtech toward an appetizing tomorrow.

Stay tuned for part two of this series, where we will dive deep into the potential of AI to address food safety, transparency and quality challenges (and more) across the food supply chain…

A brief history of key developments in food production and AI
A brief history of key developments in food production and AI. Image: Peakbridge

Further reading: