Silicon Valley-based food & beverage innovation firm Mattson has announced a chief AI officer and unveiled a suite of AI-enhanced services it claims can both enhance and dramatically speed up the product development process.
While anyone with a ChatGPT account can generate a concept and packaging design for a new high-protein snack or caffeine-laden energy drink, Mattson has developed customized AI services via its ProtoThink* cloud-based innovation platform and Food Studio Ideation service, which are tailored towards food industry professionals, says chairman and new chief AI officer Steve Gundrum.
“A lot of our clients come to us and say they’ve tried using off the shelf AI and in many cases, don’t feel it’s very powerful or helpful,” says Gundrum, who has led the development of new tech-and-trend-driven innovation services at Mattson for over three decades, working with everyone from high-profile foodtech startups to multinationals. “And then they see what we’re able to do with our customized models and it completely changes their perspective.
“We don’t look at every problem as a nail and AI is the hammer, but it really changes everything. AI will be used to enhance our holistic innovation approach, from strategy and insights to ideation, concept development, product development and commercialization.”
He adds: “A huge percentage of what’s on the internet is food related, either directly, or indirectly,” he notes. “A blog about long distance running and marathon prep is going to talk a lot about nutrition, while half of the material on travel blogs is about eating out. Type ‘intermittent fasting’ or ‘butter chicken’ or ‘BlueZone Diets’ into Google and you’ll get millions of hits. AI can synthesize all of this information in seconds.”
Customized AI models, AI-powered assistant Leo, and AI-personas
From AI-powered assistant ‘Leo’ to highly customized AI personas to use as sounding boards, Mattson has a series of tools designed to enhance and speed up the innovation process by developing concepts with a fraction of the cost of legacy techniques, says Gundrum.
“If you’re using off the shelf generic ChatGPT to develop product ideas, your prompt engineering skills need to be really high, really refined, and multi-conditional. With ProtoThink AI, a lot of the heavy lifting is being done in the background by the model, so you can interact with the AI models in a natural language way and still get incredible results.”
Inexperienced gen AI users “often use AI like they use a search engine,” and as a result, the “responses aren’t very good,” he adds. “When we create our custom models, they become far more focused and curate the responses in a way that makes sense to a food industry professional because they speak the language of food. Our models are optimized for responses based upon subject matter experts in the food industry.
“So for example, I’m working on something in the dairy category right now where one of the parameters is we want [AI-powered] food scientists who are PhDs, who’ve studied the dairy category for 20 years. When you start putting those kinds of lenses and filters on things, you get infinitely better responses.”
He adds: “We optimize the models to be an instructor at the CIA [Culinary Institute of America] with 20 years’ experience or maybe a chef who specializes in preparing seafood, or a consumer insights person who’s got 20 years of doing research in a certain category. So when you input your prompts and query those personas, you get something you won’t get from off-the-shelf AI.
“We call them [AI personas] The Magnificent Seven. Who are the seven experienced people you would love to be on your team to create new products, and we essentially program them into our AI models. Can you imagine how much more focused and how much deeper the insights are when you’re talking to the right people?”
Hour-long therapy sessions…
In other scenarios, Mattson creates AI-generated target consumers that clients can interact with, or in turn watch them interact with other AI-generated peers, to discuss everything from thin crust pizza preferences in a group text chat to the challenges of maintaining a keto diet.
In some cases, interactions between AI-generated personas can unlock intimate insights that go beyond what you might get from a focus group, for a fraction of the cost and time, he claims.
As one example, Gundrum created two personas: a 42-year-old woman with two kids that is losing weight but reluctant to tell anybody she’s using Ozempic; and a therapist helping her work through her issues: “We simulated hour long therapy sessions where she’s talking to the therapist to understand why she doesn’t want to tell anybody.”
TURF analyses in seconds
AI-powered assistant Leo, who went live in December 2023, helps lead clients through a structured ideation process asking targeted questions, analyzing market trends, consumer behaviors and sustainability considerations from proprietary and publicly available data, to ensure new concepts are both innovative and aligned with market demands, says Gundrum.
ProtoThink AI can also help clients develop a TURF analysis (Total Unduplicated Reach and Frequency) for a given concept in order to optimize the offering to achieve the highest market penetration and effectiveness, he explains.
“Every client since December who’s engaged us on ProtoThink is now using ProtoThink AI, and they’re finding it pretty eye-opening, as it’s dramatically speeding up the creative process and saving time and money compared to legacy innovation processes. There’s just a huge difference.”
Where are CPG companies on AI for product development?
In general, says Gundrum, CPG companies are “all over the map” when it comes to using AI, but the larger the company, he says, the slower they tend to be to react. “When we hear that someone has formed a committee to look at AI, that’s usually a pretty good indication that things are going to move slowly.”
Speaking to AgFunderNews in April, Mattson CEO Dr. Justin Shimek noted that large CPG companies “are trying to become more nimble and agile, more scrappy, so they can get things into market more quickly and behave more like startups versus the old days, where they would do a lot of really rigorous, expensive, long-term testing, and it would take forever to launch anything. And everyone is trying to figure out how to integrate AI into their process.”
The COVID-19 pandemic temporarily stymied the innovation process at many food companies as attention shifted to handling supply chain disruptions and getting products on shelves, while many large food companies have also been able to ride on an inflationary wave for the past couple of years by taking pricing, said Shimek. But the pressure is now on to innovate again, he said.
“Companies have realized that the only way they can continue to fuel their business is through innovation, so they’re asking, how do we grow categories and not just market share?”
*Mattson will host sessions at its HQ in San Mateo on Sept 10, and at its Chicago offices on October 1, in which guests will be able to create a new food or beverage concept using its ProtoThink AI tool.
Further reading:
Mattson expands to Chicago with acquisition of Hyde Park Group, explores the Ozempic effect
Bytes to Bites part one: Digitizing consumption insights. Leveraging AI in food product development
Guest article: The real impact of tariffs on US produce prices