Before Mondelēz International could trust AI, it had to trust its data
Mondelēz International, the maker of Oreo, Cadbury, and Chips Ahoy, rebuilt how it governs data before expanding AI use. Using Ataccama ONE, the company connected data quality, cataloging, and lineage into a single environment, then shifted ownership of that data from IT to the business teams who create and use it. That foundation now lets Mondelēz extend data trust to AI and agentic systems, not just to the people making decisions.
Mondelēz International spent years building a data-savvy organization, yet meetings could still get bogged down in debates over whether a number was right. By putting trusted data and clear accountability in place, the company created the foundation it needed to expand into AI.
At the same time, Mondelēz is simplifying how it manages data across the enterprise. Ataccama ONE, which spans data quality, data catalog, data observability, and data lineage, provides a common environment for defining data, setting quality expectations, and understanding how it moves through the company’s systems.
For the company, the shift was as much about people as it was about technology. By establishing who is responsible for data, agreeing on what it means, and setting expectations for its quality, Mondelēz strengthened data trust across the business as new AI use cases emerged.
Outcomes
- Faster, more confident decisions: Leadership teams spend less time questioning whether a number is right and more time acting on what the data tells them.
- Clearer accountability for data: Responsibility has shifted from IT alone to the people who use and understand data across the business, with defined owners and stewards responsible for its meaning and quality.
- A stronger foundation for AI: Trusted, well-defined data gave Mondelēz the confidence to expand into AI use cases that Bjorkqvist says the company would not have pursued without this groundwork.
- Governed data supporting core business processes: Beyond dashboards and KPI tracking, the same data foundation underpins demand and supply planning, financial reporting, regulatory compliance, and internal processes such as employee data. These are critical operations, and governing the underlying data improves them across the board.
- Groundwork for agentic commerce: As AI systems increasingly retrieve product information on consumers’ behalf, governed product data, or brand as data, positions Mondelēz to make sure those systems find accurate, trustworthy information.
When checking the data slows down decisions
Mondelēz International was already a data-savvy organization, with dashboards and analytics supporting decisions across the business. But having data available did not always mean people could trust it immediately.
Bjorkqvist described a familiar scenario: a number appears on a dashboard during a leadership meeting, someone questions whether it is right, and the conversation stalls. Teams then have to trace where the number came from and find the person responsible for the underlying data before they can get back to the decision at hand.
Mondelēz has changed that process by strengthening its approach to data trust.
“We spend far less time debating whether a number is right. That layer of discussion is largely gone,” Bjorkqvist said.
Instead, teams can spend more time acting on the information in front of them. The challenge was never a lack of data or analytical capability. “We’ve always been a very data-savvy organization,” Bjorkqvist said. The gap made it easy for people across the business to understand what a number meant, where it came from, and whether they could rely on it.
When every team managed data differently
For years, different parts of Mondelēz International had their own ways of managing data, supported by different tools and processes. That made it difficult to see how information created in one part of the business was used elsewhere. Marketing data, for example, could flow into finance systems without a clear connection between the people who produce the data and those who rely on it.
“There wasn’t always a clear connection between the people creating data and the people using it,” Bjorkqvist said.
The complexity also made data management & analytics difficult for employees outside specialist teams. People with deep technical knowledge could work across different systems, but for a sales representative in Brazil or a marketing employee in Europe, managing and understanding data was not a natural part of the job.
“Data governance used to sit in the back room, far away from where the data was actually being used,” Bjorkqvist said.
Mondelēz wanted to change that by bringing data responsibility closer to the business.
Moving accountability into the business
For Mondelēz International, improving trust in data required more than introducing new technology. It meant changing who was responsible for the data in the first place.
Bjorkqvist spends much of her time working with leaders across the company on that shift. “I spend a lot of my time working with leaders on the cultural shift this requires and what it means to take ownership of data,” she said.
Mondelēz has defined clear roles for managing data across the business. Data owners are typically senior leaders who are accountable for particular areas of data. Data stewards are closer to the day-to-day work, such as employees working in SAP, managing external data from providers like Nielsen, or entering information into operational systems. IT teams serve as custodians, responsible for keeping the technology and systems behind the data running.
The biggest change is that responsibility for what the data means and whether it is reliable now sits with the business.
“Data ownership now sits with the business. The people closest to the data are accountable for what it means and the quality we expect from it,” Bjorkqvist said.
Stewards define datasets and KPIs in Ataccama ONE’s data catalog and establish the level of data quality required based on how the information will be used. A finance dataset, for example, may need to be virtually error-free, while data used to forecast sales may allow for more variation.
Ataccama’s data quality capabilities monitor the data against those expectations and flag potential quality issues for review. Stewards can confirm or reject what the system identifies, helping improve how those issues are detected over time.
Getting there also requires people to agree on what their data actually means. Bjorkqvist gave the example of something as seemingly simple as a “red apple”: people in different countries might interpret that description differently. Mondelēz brings those perspectives together to agree on a common definition before recording it in Ataccama and using it more widely.
In many cases, the people taking on these responsibilities were already looking after data as part of their day-to-day jobs; they simply weren’t formally recognized as data stewards. Mondelēz has focused on identifying those employees, giving them defined responsibilities and helping them manage data as part of their roles.
Bringing data management into one place
Mondelēz International is also simplifying the technology behind its new approach. Instead of relying on different solutions across the business, the company is using Ataccama ONE as a common place for employees to define data, establish expectations for its quality, and understand how it is used across the organization.
As part of that effort, Mondelēz is consolidating governance and data quality processes previously spread across multiple legacy tools into Ataccama ONE.
Mondelēz does not need to move all of its underlying data into Ataccama. Instead, Ataccama connects directly to source systems and processes data where it already lives rather than copying it into a separate repository. That lets teams understand where information comes from and how it moves between systems.
“Rather than moving data into Ataccama, we connect to our existing data sources to understand where the data comes from and how it flows across the organization,” Bjorkqvist said.
At Mondelēz’s scale, that means data doesn’t have to be duplicated, migrated, or reconciled into yet another system before it can be trusted. Data lineage then makes visible how information moves between systems and downstream assets, so when something changes upstream, Mondelēz can see what it affects before it shows up as a wrong number on a dashboard.
For employees responsible for that data, the result is a more consistent way to understand what it means, who is responsible for it, and whether its quality is sufficient for its intended use.
From data trust to faster decisions
Mondelēz International has removed some of the friction that once slowed decision-making. Instead of spending time tracking down who owns a number or checking where it came from, teams can focus more quickly on what the data is telling them and what to do next.
“We’ve removed a layer of uncertainty. We can make decisions faster because people have greater confidence in the data they’re using,” Bjorkqvist said.
Mondelēz has also brought more people into data. Employees who work with data every day, from sales teams to people working in operational systems, now have a defined role in making sure the information they use is understood and reliable.
AI raises the stakes for that data trust.
Why data trust matters even more for AI
Mondelēz International’s AI ambitions extend well beyond dashboards. The company has reported investing more than $40 million in AIDA, a generative AI advertising platform built with Publicis Groupe and Accenture that Mondelēz says can cut marketing production costs by 30 to 50%, according to reporting from Food Dive. Separately, Mondelēz digital leadership has described its AI strategy as split between everyday productivity tools for employees and deliberate, enterprise-scale initiatives across manufacturing, supply chain and sales, in comments to CIO.
As Mondelēz International expands its use of AI, the company needs to know what its data means, where it came from, and whether it can be trusted.
“Our move into AI and agentic use cases depended on getting the data foundation right first,” Bjorkqvist said.
For Bjorkqvist, AI creates a new requirement for context. She describes this as “context governance”: making sure Mondelēz understands where data originates and how it moves through the organization before it reaches an AI model.
“As AI models begin using data autonomously, context becomes critical. We need to understand where that data came from and how it moved through our ecosystem before a model uses it,” she said.
That makes the governance work Mondelēz began years ago increasingly relevant. The company needs clear definitions, established responsibility for the data, and quality expectations that reflect how the information will be used. It also needs visibility into how that information moves between systems.
Together, those controls help Mondelēz extend data trust beyond the people making decisions to the AI systems increasingly using enterprise data.
Preparing product data for an AI-driven world
Mondelēz International is already introducing AI agents to help employees work more efficiently, including tools that can assist with tasks such as meeting notes.
Looking further ahead, the company is also preparing for what Bjorkqvist calls “agentic commerce”: a future in which consumers increasingly find product information through AI-powered search and other AI systems.
Bjorkqvist used Oreo as an example. Today, a consumer can read product information directly from packaging or look it up online. As AI becomes another route to that information, Mondelēz needs to make sure the product data available to those systems is accurate and reliable.
If a consumer asks an AI-powered service about an Oreo product, the goal is for the system to find and use the right product information.
That changes the audience for governed data. It needs to be understandable and trustworthy not only for employees making decisions but also for AI systems that retrieve information on behalf of consumers.
Data quality is never finished
Mondelēz International isn’t trying to govern every piece of data at once. Instead, it has focused first on a smaller set of critical data elements: the information the business has determined is most important to manage and trust.
Mondelēz can expand that scope as its needs change. A new AI application, for example, may create a need to manage data that was not previously considered a priority.
“Data quality is an ongoing journey. We’re never going to be 100% finished, because the data we need to govern will continue to evolve,” Bjorkqvist said.
For Mondelēz, that makes data governance an ongoing business capability rather than a project with a fixed end date. As employees, consumers and AI systems find new ways to use data, the company will need to continue defining, managing and building trust in the information behind them.
The work started with a practical business problem: helping people spend less time questioning the numbers in front of them and more time acting on them. Now, as Mondelēz moves further into AI, data trust is taking on a broader role: giving both people and AI systems the confidence to use the company’s data.
Mondelēz International, the maker of Oreo, Cadbury, and Chips Ahoy, rebuilt how it governs data before expanding AI use. Using Ataccama ONE, the company connected data quality, cataloging, and lineage into a single environment, then shifted ownership of that data from IT to the business teams who create and use it. That foundation now lets Mondelēz extend data trust to AI and agentic systems, not just to the people making decisions.
Mondelēz International spent years building a data-savvy organization, yet meetings could still get bogged down in debates over whether a number was right. By putting trusted data and clear accountability in place, the company created the foundation it needed to expand into AI.
At the same time, Mondelēz is simplifying how it manages data across the enterprise. Ataccama ONE, which spans data quality, data catalog, data observability, and data lineage, provides a common environment for defining data, setting quality expectations, and understanding how it moves through the company’s systems.
For the company, the shift was as much about people as it was about technology. By establishing who is responsible for data, agreeing on what it means, and setting expectations for its quality, Mondelēz strengthened data trust across the business as new AI use cases emerged.
Outcomes
- Faster, more confident decisions: Leadership teams spend less time questioning whether a number is right and more time acting on what the data tells them.
- Clearer accountability for data: Responsibility has shifted from IT alone to the people who use and understand data across the business, with defined owners and stewards responsible for its meaning and quality.
- A stronger foundation for AI: Trusted, well-defined data gave Mondelēz the confidence to expand into AI use cases that Bjorkqvist says the company would not have pursued without this groundwork.
- Governed data supporting core business processes: Beyond dashboards and KPI tracking, the same data foundation underpins demand and supply planning, financial reporting, regulatory compliance, and internal processes such as employee data. These are critical operations, and governing the underlying data improves them across the board.
- Groundwork for agentic commerce: As AI systems increasingly retrieve product information on consumers’ behalf, governed product data, or brand as data, positions Mondelēz to make sure those systems find accurate, trustworthy information.
When checking the data slows down decisions
Mondelēz International was already a data-savvy organization, with dashboards and analytics supporting decisions across the business. But having data available did not always mean people could trust it immediately.
Bjorkqvist described a familiar scenario: a number appears on a dashboard during a leadership meeting, someone questions whether it is right, and the conversation stalls. Teams then have to trace where the number came from and find the person responsible for the underlying data before they can get back to the decision at hand.
Mondelēz has changed that process by strengthening its approach to data trust.
“We spend far less time debating whether a number is right. That layer of discussion is largely gone,” Bjorkqvist said.
Instead, teams can spend more time acting on the information in front of them. The challenge was never a lack of data or analytical capability. “We’ve always been a very data-savvy organization,” Bjorkqvist said. The gap made it easy for people across the business to understand what a number meant, where it came from, and whether they could rely on it.
When every team managed data differently
For years, different parts of Mondelēz International had their own ways of managing data, supported by different tools and processes. That made it difficult to see how information created in one part of the business was used elsewhere. Marketing data, for example, could flow into finance systems without a clear connection between the people who produce the data and those who rely on it.
“There wasn’t always a clear connection between the people creating data and the people using it,” Bjorkqvist said.
The complexity also made data management & analytics difficult for employees outside specialist teams. People with deep technical knowledge could work across different systems, but for a sales representative in Brazil or a marketing employee in Europe, managing and understanding data was not a natural part of the job.
“Data governance used to sit in the back room, far away from where the data was actually being used,” Bjorkqvist said.
Mondelēz wanted to change that by bringing data responsibility closer to the business.
Moving accountability into the business
For Mondelēz International, improving trust in data required more than introducing new technology. It meant changing who was responsible for the data in the first place.
Bjorkqvist spends much of her time working with leaders across the company on that shift. “I spend a lot of my time working with leaders on the cultural shift this requires and what it means to take ownership of data,” she said.
Mondelēz has defined clear roles for managing data across the business. Data owners are typically senior leaders who are accountable for particular areas of data. Data stewards are closer to the day-to-day work, such as employees working in SAP, managing external data from providers like Nielsen, or entering information into operational systems. IT teams serve as custodians, responsible for keeping the technology and systems behind the data running.
The biggest change is that responsibility for what the data means and whether it is reliable now sits with the business.
“Data ownership now sits with the business. The people closest to the data are accountable for what it means and the quality we expect from it,” Bjorkqvist said.
Stewards define datasets and KPIs in Ataccama ONE’s data catalog and establish the level of data quality required based on how the information will be used. A finance dataset, for example, may need to be virtually error-free, while data used to forecast sales may allow for more variation.
Ataccama’s data quality capabilities monitor the data against those expectations and flag potential quality issues for review. Stewards can confirm or reject what the system identifies, helping improve how those issues are detected over time.
Getting there also requires people to agree on what their data actually means. Bjorkqvist gave the example of something as seemingly simple as a “red apple”: people in different countries might interpret that description differently. Mondelēz brings those perspectives together to agree on a common definition before recording it in Ataccama and using it more widely.
In many cases, the people taking on these responsibilities were already looking after data as part of their day-to-day jobs; they simply weren’t formally recognized as data stewards. Mondelēz has focused on identifying those employees, giving them defined responsibilities and helping them manage data as part of their roles.
Bringing data management into one place
Mondelēz International is also simplifying the technology behind its new approach. Instead of relying on different solutions across the business, the company is using Ataccama ONE as a common place for employees to define data, establish expectations for its quality, and understand how it is used across the organization.
As part of that effort, Mondelēz is consolidating governance and data quality processes previously spread across multiple legacy tools into Ataccama ONE.
Mondelēz does not need to move all of its underlying data into Ataccama. Instead, Ataccama connects directly to source systems and processes data where it already lives rather than copying it into a separate repository. That lets teams understand where information comes from and how it moves between systems.
“Rather than moving data into Ataccama, we connect to our existing data sources to understand where the data comes from and how it flows across the organization,” Bjorkqvist said.
At Mondelēz’s scale, that means data doesn’t have to be duplicated, migrated, or reconciled into yet another system before it can be trusted. Data lineage then makes visible how information moves between systems and downstream assets, so when something changes upstream, Mondelēz can see what it affects before it shows up as a wrong number on a dashboard.
For employees responsible for that data, the result is a more consistent way to understand what it means, who is responsible for it, and whether its quality is sufficient for its intended use.
From data trust to faster decisions
Mondelēz International has removed some of the friction that once slowed decision-making. Instead of spending time tracking down who owns a number or checking where it came from, teams can focus more quickly on what the data is telling them and what to do next.
“We’ve removed a layer of uncertainty. We can make decisions faster because people have greater confidence in the data they’re using,” Bjorkqvist said.
Mondelēz has also brought more people into data. Employees who work with data every day, from sales teams to people working in operational systems, now have a defined role in making sure the information they use is understood and reliable.
AI raises the stakes for that data trust.
Why data trust matters even more for AI
Mondelēz International’s AI ambitions extend well beyond dashboards. The company has reported investing more than $40 million in AIDA, a generative AI advertising platform built with Publicis Groupe and Accenture that Mondelēz says can cut marketing production costs by 30 to 50%, according to reporting from Food Dive. Separately, Mondelēz digital leadership has described its AI strategy as split between everyday productivity tools for employees and deliberate, enterprise-scale initiatives across manufacturing, supply chain and sales, in comments to CIO.
As Mondelēz International expands its use of AI, the company needs to know what its data means, where it came from, and whether it can be trusted.
“Our move into AI and agentic use cases depended on getting the data foundation right first,” Bjorkqvist said.
For Bjorkqvist, AI creates a new requirement for context. She describes this as “context governance”: making sure Mondelēz understands where data originates and how it moves through the organization before it reaches an AI model.
“As AI models begin using data autonomously, context becomes critical. We need to understand where that data came from and how it moved through our ecosystem before a model uses it,” she said.
That makes the governance work Mondelēz began years ago increasingly relevant. The company needs clear definitions, established responsibility for the data, and quality expectations that reflect how the information will be used. It also needs visibility into how that information moves between systems.
Together, those controls help Mondelēz extend data trust beyond the people making decisions to the AI systems increasingly using enterprise data.
Preparing product data for an AI-driven world
Mondelēz International is already introducing AI agents to help employees work more efficiently, including tools that can assist with tasks such as meeting notes.
Looking further ahead, the company is also preparing for what Bjorkqvist calls “agentic commerce”: a future in which consumers increasingly find product information through AI-powered search and other AI systems.
Bjorkqvist used Oreo as an example. Today, a consumer can read product information directly from packaging or look it up online. As AI becomes another route to that information, Mondelēz needs to make sure the product data available to those systems is accurate and reliable.
If a consumer asks an AI-powered service about an Oreo product, the goal is for the system to find and use the right product information.
That changes the audience for governed data. It needs to be understandable and trustworthy not only for employees making decisions but also for AI systems that retrieve information on behalf of consumers.
Data quality is never finished
Mondelēz International isn’t trying to govern every piece of data at once. Instead, it has focused first on a smaller set of critical data elements: the information the business has determined is most important to manage and trust.
Mondelēz can expand that scope as its needs change. A new AI application, for example, may create a need to manage data that was not previously considered a priority.
“Data quality is an ongoing journey. We’re never going to be 100% finished, because the data we need to govern will continue to evolve,” Bjorkqvist said.
For Mondelēz, that makes data governance an ongoing business capability rather than a project with a fixed end date. As employees, consumers and AI systems find new ways to use data, the company will need to continue defining, managing and building trust in the information behind them.
The work started with a practical business problem: helping people spend less time questioning the numbers in front of them and more time acting on them. Now, as Mondelēz moves further into AI, data trust is taking on a broader role: giving both people and AI systems the confidence to use the company’s data.
