This food giant's India AI teams are helping R&D, client experience

Just before Jennifer Hartsock sat down to talk with us, Cargill’s chief information & digital officer had had a discussion with the head of the company’s global AI centre of excellence based in Bengaluru. “I loved it. I think it was awesome. And the passion is very real. Of course, that’s a very fast moving space, so it’s also one we want to stay in front of,” she tells us.
And as it turned out, much of our discussion too was around AI – because it’s work that’s become critical to the $160-billion US agribusiness and food company, and it’s work that the India centre is playing a huge role in.

“I’ll give a really fun (example) that is probably less scientific, but we think it’s really cool,” Hartsock says, as she goes into details of a new GenAI product that they’ve developed. It allows Cargill’s customers – the likes of McDonald’s and KFC – to do personabased product innovation. The customer could ask a question like, ₹For a millennial who is social media s av v y, what would be potential menu items that utilises existing or future Cargill products?’ The system will come up with an answer, and then the customer could ask how that menu should be in India or the US. “It generates some really, really cool, innovative, and sometimes crazy suggestions. But I have to tell you, the customer interaction with that was incredible,” Hartsock says.

It also provides Cargill with insights into where additional product innovation might be required, and enables it to work with customers’ recipes to create unique flavours for their markets.

Even though Cargill teams work on predictive analytics, generative and agentic AI, Hartsock says GenAI is where most projects are happening right now. It’s being used to improve customer experience, to increase the efficiency of Cargill’s teams, in decision support (ensuring knowledge in one part of the company is available easily to all), and in R&D.

One ChatGPT client the company developed allows sales executives to use natural language to suggest the most appropriate ingredients to bakeries. Cargill has a hugely complex ingredient portfolio and it’s not easy for sales folks to know what’s best in all contexts. Hartsock says she recently played around with the system, asking what types of baking goods zucchini could go into, and it came back saying it could go into muffins, into breads, into cakes. But since zucchini has a very high moisture content, it recommended Cargill ingredients that could address that high humidity and still maintain shelf stability for the product.

Developing new sweeteners
Hartsock’s colleague Florian Schattenmann (CTO and VP for innovation and R&D) has been innovating on ingredients using predictive models around molecules and narrowing the universe of experiments they need to perform. One of these has been around low-calorie or no-calorie sweeteners. The stevia leaf has been one of the original sources of such sweeteners. But it’s possible to create the exact same molecule through fermentation of other commodities too, like corn. Hartsock says they helped Schattenmann’s team with AI models to fine tune those experiments to transition from a stevia leaf to other commodities. “It allows those scientists to work better and faster,” she says.

A number of AI projects, Hartsock says, are also going on to improve efficiency. In coding, it’s already easily giving productivity improvements of 10-15%. Work is on to use AI to find ways to do core ERP projects faster and less expensively.

The legal team uses AI to analyse contracts. AI is being used to quickly understand new govt policies in order to help Cargill’s commodity traders. “When one of our govts releases a new policy, it could be 800 pages. And if we can figure out a way to summarise that with accuracy for people, that really facilitates faster and more accurate decision making. With AI, what took days can now be given to traders in minutes,” Hartsock says.