Hershey, Kellanova, Diageo & Giant Eagle Execs Strategize on How to Prove AI Value

Getting Into AI Applications

Taylor expedited what would be a years-long process by throwing modeling at a problem around sourcing. As prices have gone up threefold, modeling has allowed the company to better structure how it manages its hedging positioning and commodities buying in cocoa. 

This approach helped the company show the business case and potential impact so stakeholders could see the value before spending time and money building out the foundation. 

Pineda is taking a start-small approach in his company’s AI investment, using the technology within demand forecasting to solve significant pain points around accuracy and productivity. The team is applying the tech in one state to see if it will save people time before deciding whether or not to expand to several states.

For Jena Anton, senior director, commercial OE, Kellanova, AI has been used as a prioritization tool, helping the company to determine the next best action for retail execution reps and what insights reps can get with a picture of the shelf. Additionally, if there are any executional issues, how can AI be used to understand the root cause? 

More: Results from Kellanova’s data clean room and AI pilot 

Determining the Value & Measuring Success 

AI and advanced analytics haven’t changed the spectrum of how the company drives the business, said Taylor. What’s fairly new for Hershey is the concept of measuring automation combined with human touches and manual interventions. The company has been trying to build out some of those metrics into its process flow. 

It’s about fully understanding the opportunity you are trying to capture and problem you are trying to solve, and adopting those measures in your use case, said Pineda. “Define them at the very beginning and make sure it’s very clear.”

Edwards reminded the audience that it’s more about getting incremental wins and building out the broader solution along the way, particularly with newer technology like generative AI.

“You have to moderate expectations by showing a relative view of where we are in the process right now,” said Edwards. “Walk them back. … We are doing this because we need to have foundational elements in place.”