Gen AI Speeds Content Production for Our Brands: Mondelez International’s Michael Lampert
AMENIA, NY – After spending up to two years building and testing its generative artificial intelligence platform, Mondelez International is ready to serve it to more brands and markets. Unlike a cookie, however, the technology comes with training sessions, compliance checks and a lawyer.
“For us, it’s really about activation now,” Michael Lampert, senior director of global gen AI at Mondelez International, said in an interview with Beet.TV contributor David Kaplan at the Beet Retreat Berkshires.
Mondelez piloted its ADA content-generation platform in seven global markets while evaluating whether the technology could deliver what the company needed. It is now moving from development into worldwide adoption.
“We’ve done all the change management, we’ve done all the training, we know the product does what it’s supposed to do,” Lampert said. The next step is getting the platform into wider use across Mondelez brands and business units.
Teaching AI the rules of the cookie jar
Producing content at scale creates an obvious concern: Artificial intelligence can make things quickly, but it can also make things that no brand manager wants to explain on a Monday morning.
Mondelez created knowledge bases containing brand guidelines, previous campaigns, older assets and other standards. That information helps guide the platform after a user enters a prompt.
The company also built a proprietary content checker. Before material can go live, it must pass about 16 compliance tests. These cover issues such as overconsumption, the depiction of people or celebrities and requirements unique to individual brands.
“The machine generates the content, the machine checks the content against the governance rules and then the human has the final say as to whether that ad goes live or not,” Lampert said.
In other words, AI may be doing more of the production work, but it still has to answer to an adult.
Agencies get their hands on the keyboard
Mondelez designed ADA primarily for its creative agency partners rather than its own employees. Only about 5% of the company’s content is produced internally, Lampert said.
The platform is intended to strengthen the creative process within agency content-development teams. That means much of the organizational challenge involves teaching agencies how to incorporate ADA throughout their workflows.
“The change management comes from training the agencies on how to use the system and how to embed the system into their process from beginning to end,” Lampert said.
Mondelez employees may brief projects, review work or approve it, but they generally are not the people operating the platform. Lampert said employees are eager to try it anyway, perhaps proving that nothing makes a new tool more appealing than being told it is meant for somebody else.
Four people walk into an AI governance meeting
Mondelez oversees the platform through what Lampert called a “management quartet.” The group represents business strategy, the connection between technology and brands, information technology and the legal department.
The quartet manages business needs, technical development, the product itself and governance. It also helps ensure that the excitement around generative AI does not outrun the company’s ability to control it.
Anyone using ADA must complete formal training and become an authorized user. The instruction covers the capabilities of generative AI and the fine art of writing a useful prompt.
“If your prompt isn’t very good, we created a tool called prompt enhancers, so they will help you,” Lampert said.
The platform can detect when a prompt is unnecessarily specific. Users do not need to tell it that an Oreo is a chocolate cookie because the system already possesses that important piece of human knowledge.
Mondelez wants users to spend their time taking advantage of the technology instead of wrestling with instructions. That requires a significant upfront investment in education, Lampert said.
“So far the users get in there and they love it,” he said.
Video remains the bigger prize
Mondelez began with text and image generation because those technologies were mature enough for commercial use. They have worked particularly well for commerce content, where speed is often more valuable than producing the next great conceptual advertising campaign.
“If we’re honest, it’s not super conceptual,” Lampert said. “The speed to market is very fast.”
The company added video generation about six months ago, starting with clips lasting six to eight seconds. Its longer-term goal is to produce complete broadcast-quality and digital video spots from beginning to end.
That ambition matters because about 80% of Mondelez’s advertising impressions are video-based. The company is therefore working to expand adoption as quickly as the technology allows.
“The first results look great all the way from storyboard through execution in the system,” Lampert said. Still, he added that more work is needed before video generation can be fully adopted across all of Mondelez’s global brands.
For now, ADA can help create the ad, check the ad and flag potential problems. A human still decides whether the public should see it. That may be reassuring to anyone who has ever watched a machine invent fingers, faces or an Oreo filling that appears to be plotting something.
You’re watching coverage from Beet Retreat Berkshires 2026. For more videos from this event, please visit this page.
AMENIA, NY – After spending up to two years building and testing its generative artificial intelligence platform, Mondelez International is ready to serve it to more brands and markets. Unlike a cookie, however, the technology comes with training sessions, compliance checks and a lawyer.
“For us, it’s really about activation now,” Michael Lampert, senior director of global gen AI at Mondelez International, said in an interview with Beet.TV contributor David Kaplan at the Beet Retreat Berkshires.
Mondelez piloted its ADA content-generation platform in seven global markets while evaluating whether the technology could deliver what the company needed. It is now moving from development into worldwide adoption.
“We’ve done all the change management, we’ve done all the training, we know the product does what it’s supposed to do,” Lampert said. The next step is getting the platform into wider use across Mondelez brands and business units.
Teaching AI the rules of the cookie jar
Producing content at scale creates an obvious concern: Artificial intelligence can make things quickly, but it can also make things that no brand manager wants to explain on a Monday morning.
Mondelez created knowledge bases containing brand guidelines, previous campaigns, older assets and other standards. That information helps guide the platform after a user enters a prompt.
The company also built a proprietary content checker. Before material can go live, it must pass about 16 compliance tests. These cover issues such as overconsumption, the depiction of people or celebrities and requirements unique to individual brands.
“The machine generates the content, the machine checks the content against the governance rules and then the human has the final say as to whether that ad goes live or not,” Lampert said.
In other words, AI may be doing more of the production work, but it still has to answer to an adult.
Agencies get their hands on the keyboard
Mondelez designed ADA primarily for its creative agency partners rather than its own employees. Only about 5% of the company’s content is produced internally, Lampert said.
The platform is intended to strengthen the creative process within agency content-development teams. That means much of the organizational challenge involves teaching agencies how to incorporate ADA throughout their workflows.
“The change management comes from training the agencies on how to use the system and how to embed the system into their process from beginning to end,” Lampert said.
Mondelez employees may brief projects, review work or approve it, but they generally are not the people operating the platform. Lampert said employees are eager to try it anyway, perhaps proving that nothing makes a new tool more appealing than being told it is meant for somebody else.
Four people walk into an AI governance meeting
Mondelez oversees the platform through what Lampert called a “management quartet.” The group represents business strategy, the connection between technology and brands, information technology and the legal department.
The quartet manages business needs, technical development, the product itself and governance. It also helps ensure that the excitement around generative AI does not outrun the company’s ability to control it.
Anyone using ADA must complete formal training and become an authorized user. The instruction covers the capabilities of generative AI and the fine art of writing a useful prompt.
“If your prompt isn’t very good, we created a tool called prompt enhancers, so they will help you,” Lampert said.
The platform can detect when a prompt is unnecessarily specific. Users do not need to tell it that an Oreo is a chocolate cookie because the system already possesses that important piece of human knowledge.
Mondelez wants users to spend their time taking advantage of the technology instead of wrestling with instructions. That requires a significant upfront investment in education, Lampert said.
“So far the users get in there and they love it,” he said.
Video remains the bigger prize
Mondelez began with text and image generation because those technologies were mature enough for commercial use. They have worked particularly well for commerce content, where speed is often more valuable than producing the next great conceptual advertising campaign.
“If we’re honest, it’s not super conceptual,” Lampert said. “The speed to market is very fast.”
The company added video generation about six months ago, starting with clips lasting six to eight seconds. Its longer-term goal is to produce complete broadcast-quality and digital video spots from beginning to end.
That ambition matters because about 80% of Mondelez’s advertising impressions are video-based. The company is therefore working to expand adoption as quickly as the technology allows.
“The first results look great all the way from storyboard through execution in the system,” Lampert said. Still, he added that more work is needed before video generation can be fully adopted across all of Mondelez’s global brands.
For now, ADA can help create the ad, check the ad and flag potential problems. A human still decides whether the public should see it. That may be reassuring to anyone who has ever watched a machine invent fingers, faces or an Oreo filling that appears to be plotting something.
You’re watching coverage from Beet Retreat Berkshires 2026. For more videos from this event, please visit this page.
