How Mondelēz is using AI to accelerate revenue management in Latin America
A practical look at Mondelēz AI revenue management in Latin America: building data confidence and governance to make AI useful at scale across LATAM RGM.
Note: Quotes have been translated from Spanish to English
In consumer goods, AI conversations often default to speed. Mondelēz AI revenue management in Latin America prioritized something else first: data confidence, governance, and user trust—so AI could accelerate revenue decisions at scale without compromising how teams work.
At the 2026 Promotion Optimization Institute LATAM Summit, XTEL’s Ben Peart, Director Sales, NA, sat down with two of the company’s Latin America technology leaders, Gabriel Romero, Digital Experience LATAM Regional Director, and Alejandro Calva, RGM IT Manager LATAM, to unpack how one of the world’s most recognizable brand portfolios is putting AI to work in revenue management without losing sight of the people, the data, and the decisions underneath it.
Key Takeaways:
Start deliberately:
map the RGM process before investing in AI tools.
Adoption is earned:
prove AI helps users faster, better daily decisions.
Data won’t be perfect:
build trust with ownership and continuous cleansing.
Move deliberately before investing in AI
For a company known for its brands rather than its corporate name, the reputational stakes shaped the AI strategy from day one. “Many people recognize Ricolino, Ritz, Toblerone, Oreo, all the brands we make,” Romero noted, “and there was clearly a very high risk when it came to the image of each of those brands. [We needed to ensure that we protected the equity and value of our brands.]” The response was deliberate restraint. “The first decision the company made was: let’s slow the ball down a bit, let’s hold the ball and look at where we want to go, what we want to do [before we purchase and implement tools without fully outlining our overall strategy],” he said.
That pause produced a guiding principle the panel returned to repeatedly. The aim was never “to turn AI into the company’s strategy, but rather how AI becomes an accelerator of the strategy,” Romero explained. For revenue growth management specifically, that meant mapping the entire process and pinpointing where AI could create value “from the standpoint of generating better decisions at scale and, above all, at the right moment.”
A late insight, he warned, is no insight at all: “It would be of no use to us to make a decision if we’re making it three weeks later in a dashboard that, with luck, someone looks at, or once part of a seasonal promotion has already passed.” His three pillars: “scalability, the quality of decisions, and timeliness in decision-making.”
ROI and Building Trust Internally Lead to Adoption
When Peart pressed on how Mondelēz avoided the classic “here’s a tool, go run with it” trap, Calva was direct. “You don’t get adoption by buying [an] AI [tool]. Adoption doesn’t come with the purchase. Adoption comes from showing the users that the tool, in this case AI, works and is going to make their life easier one way or another.”
Romero reinforced the point with a change-management lens. When AI’s findings show a gap or an underperformance, he stressed; it does not “necessarily mean you’re doing things wrong.” In RGM and TPM, the technology “shows you where the gaps are and where you have areas of opportunity to improve.” His example was a familiar one across the season: a brand launches a Valentine’s product and floods it with promotions on the assumption that spending drives volume. As a former boss once told him: “In this company, we’re here to sell, not to give things away. Have you analyzed whether that product sells on its own without putting any promotion on it?” Now, Romero said, AI tools help the team see this “in a timely way,” rather than learning the truth “three months later or two weeks later, when the promotion is already over.”
The myth of perfect data
Both leaders rejected the idea of waiting for pristine data. “If we want to wait for the data to be perfect from the start before we can launch initiatives in TPM, RGM, or AI, it will never happen,” Romero said. Data cleansing, he added, “is a constant task. It has to be a capability that a company develops.” The consequences of getting it wrong are real: “If you don’t build the user’s trust in the data, in the end you’re going to lose them, because they’ll say: I prefer to keep doing it with my Excel spreadsheet, which I have under control.”
Mondelēz earned that trust slowly, beginning its work with XTEL back in 2018, with Brazil and Mexico launching around 2022. “The first years were dedicated to building the foundations, building the base, acquiring the right footing to generate that trust,” Romero said.
For Calva, accepting imperfection is the starting point, not an excuse. “If we assume our data is perfect, we live in this fairy-tale world where everything is perfect, and it isn’t,” he said. “The first step is to accept the error.” The fix is structural: a cross-functional “four-legged table” of sales, trade marketing, finance, and commercial administration that must work “in conjunction, in communication, consistently.” Without it, he warned, “you’ll have just one of the legs working and throwing the whole process off balance.”
Humans still own the decisions
Even as the discussion turned to agents and an AI-shopping future, Mondelēz drew a firm line. “AI is going to help us make better decisions, but the decision is made right now by a human being,” Calva said. “Better, faster, higher-quality, more informed, of course. But the human is the one who has to be there for decision-making, and a human also owns the data.” That ownership, he added, is built into the governance model: “If you don’t have a process owner, a data owner, you won’t make good decisions either.”
Leadership, the panelists agreed, made that discipline possible by resisting the temptation to scale too soon. When the team prepared to roll its Excel-based template into eight or nine new markets, leadership pushed back, Romero recalled: “You’re in a hurry, I’m not. I’d rather you do it right from the start, with solid foundations.” That patience, he said, “was key, so we didn’t start solving by giving in to the temptation of AI or the temptation of technology, buying for the sake of buying.”
The early returns are encouraging. “We’ve seen very genuine reactions of surprise,” Calva said, describing trade, finance, and sales teams realizing that pouring money into a seasonal product did not double its sales after all. These are, he said, “decisions are now based on data.” It is, as Romero framed the broader journey, “a step in evolution,” one that “takes quite a bit of time to build that trust, and you have to do it with that patience.”
Mondelēz’s approach is a useful reminder that AI value is rarely unlocked by a single model or feature. It is earned through disciplined sequencing: unify the process, tighten governance, raise data confidence, and only then let AI accelerate the decisions that matter. Organizational AI tools must be fast enough to act and transparent enough for people to trust.
Closing Thoughts
Mondelēz’s approach is a useful reminder that AI value is rarely unlocked by a single model or feature. In Mondelēz AI revenue management, it’s earned through disciplined sequencing: clarify the process, strengthen data confidence, and put governance and ownership in place so teams can act on insights at the right moment.