ORLANDO, Fla. — After ChatGPT’s launch created a massive generative AI wave, CIOs were tasked with carving out which use cases actually made sense for their teams.
It’s the journey of Frank Iannella, U.S. CIO and SVP for digital and technology at Heineken, who said he’s spent the last two years honing the beverage company’s AI strategy. At that point, the tech team had already been rethinking how to better turn company data and insights into action, but generative AI was quickly added to the fold.
“Part of this is being enabled by this new technology — AI,” Iannella said during a session at the Gartner IT Symposium/Xpo in October. “It’s a new part of our toolkit.”
AI systems are being used throughout various parts of Heineken’s business, Iannella said, such as its supply chain and sales. Iannella and his team have overhauled some key areas of their functions such as reporting and insights, data and information silos, that he said are replicable for tech leaders looking to get more out of the AI they’ve deployed.
The framework
Even before AI became ubiquitous in enterprises, Heineken was working to turn data into decision-making, Iannella said.
The first step was identifying a signal — a complex problem in operations that can be altered, such as flagging if a certain product often goes out of stock. Step two is identifying the lever or action that the business would take. In this case, one action could be to accelerate delivery.
Heineken was doing this manually, Iannella said, but with AI, the team was able to identify the signals and levers faster — and put action into motion. Next, the team focused on how to operationalize findings by writing these capabilities into their transactional systems.
Then comes a learning phase, he said. Iannella and his team assessed if automating the signals and levers prevented items from going out of stock and adjusted until they saw improvement.
“It took us over two years to kind of figure this out, leveraging what we want to use as well as a kind of new AI stack to actually accelerate and make this process successful,” Iannella said. “It’s the building blocks of decision intelligence.”
An ontology layer — a structural, machine-readable model which provides AI systems connection points — in their tech stack brought more context to AI use cases, the tech executive said.
“If you want your AI tool to actually make the right decisions, that ontology layer is where you have to have most of your work,” Iannella said. “That’s where you’re actually configuring a lot of the settings to make it work.”
Decision-ready data
Most CIOs are well-versed in the importance of clean data for AI use, but Iannella said his team realized it needed to go a step further to make its AI systems do the work it wanted them to do.
The team instead shoots for decision-ready data, which is aligned with the use cases a business is aiming to achieve, he said. A few factors for Heineken’s data Iannella considers are: Is the data current enough to spot signals such as low stock? Does the data actually connect to the levers around orders and deliveries? Does the team have the right quality of data to enable action?
While AI value can be hard to measure, this is an area where CIOs can capture ROI, Iannella said.
“Is it accurate enough? Is it specific and timely enough?” he said. “Can you actually link it to an outcome so you can actually start to measure it?”
Democratizing AI
Deploying AI throughout the enterprise means that it has to get into the hands of employees who may know a lot or very little about the technology, Iannella said.
The goal is to enable users who are closest to the work, but in a way that is guided — “where there’s safe places, and places where you may need to be a little bit more concerned,” Iannella said. Heineken rolled out Microsoft Copilot to its employees, and built a community of users who share how it’s being used across functions.
His team sees democratized AI use as having free reign to experiment on some functions, and a level of governance where pause is needed, such as with sensitive data, anything external, automating decisions or anything that impacts customers.
“There’s still trust to be earned,” Iannella told CIO Dive in an interview. “AI systems are providing us with an amazing scale, they’re providing the recommendations and actions, but we’re not there yet. There’s still certain places you need a human to validate.”
Iannella said the company has found its value in AI by striking the right balance of building use cases specific to their needs, taking action from the insight provided, but remaining cautious in high-level decisions.
“I think the way we started thinking about it was: Here’s a new reality,” Iannella said. “You can fight it, or you can try to guide it.”
