AI Takes Over Performance Marketing’s Busywork, With Humans Still Holding the Map
CANNES, France – Artificial intelligence has spent the past few years promising to remake advertising. Now it’s being asked to do something considerably harder: produce results.
Executives from Chalice.AI, Horizon Media and OMD US discussed how AI is changing campaign planning, buying and measurement during the “AI, Automation & the New Performance Playbook” panel at the Cannes Lions International Festival of Creativity.
The industry has moved beyond philosophical debates about what AI might eventually accomplish, according to Bradley Rogers, chief executive of OMD US.
“Fast forward a year, we’re working with AI every single day,” Rogers said.
He organized the opportunity into three areas: operations, discoverability and execution. Much of the initial work has focused on automating workflows and improving efficiency. Clients are now paying closer attention to discoverability as consumers increasingly use AI tools to research products and make decisions.
That change could disrupt the customer journey that marketers spent decades mapping, color-coding and occasionally pretending was linear.
Search starts with a problem
Michael Cohen, executive vice president of performance media at Horizon Media, said consumer searches are becoming more focused on tasks and problems instead of specific products.
That shift requires agencies to improve the intelligence governing decisions across multiple platforms. The challenge is coordinating those choices without forcing individual platforms to relearn campaigns and waste money in the process.
Cohen said AI’s contribution comes in “two flavors that are not mutually exclusive.” One is efficiency. The other is “bona fide insight” that can inform planning, optimization and budget allocation.
He compared their combination to the center lever on a soft-serve machine. This may be the first description of AI architecture that also makes people want ice cream.
The math needs to math
Ali Manning, chief operating officer of Chalice.AI, said the biggest change is that some brands can now point to business outcomes from AI instead of simply discussing its potential.
“What I’ve found most interesting this year is there are brands out there talking about what AI has done for them,” Manning said.
She cited results including incremental sales for Hershey’s during Halloween and increased allergy-market share for Bayer’s Claritin. These are the kinds of numbers that chief executives discuss on earnings calls, unlike a heroic click-through rate admired mainly by the people who created the report.
Manning said AI can help advertisers optimize toward meaningful business goals instead of relying on whatever metrics platforms make available. That distinction becomes especially important when every platform claims credit for the same sale.
“All the platforms come in and they’re like, ‘We drove 30% more sales year over year.’ And then the CFO is like, ‘But sales only grew 3%,’” she said. “The math doesn’t math.”
Agents need supervision
The discussion also attempted to distinguish agentic AI from conventional automation, a useful exercise now that attaching “agentic” to a product name has become the technology industry’s equivalent of adding truffle oil to french fries.
Manning said today’s bidding agents can receive instructions and determine the right price for reaching a particular consumer in a particular context. More advanced systems will learn and begin taking actions on their own.
“We’re moving into a world of agentic where agents don’t just take instructions,” she said. “They learn and then start to self-act.”
That future remains nascent, particularly in media execution. Manning said marketers can use the current generation of agents without immediately leaping to autonomous systems that raise questions about liability and suitability.
Her own head of marketing had left an AI agent running at home to continue working on marketing plans and a website. The panelists joked that someone needed to babysit it in case it decided to acquire another company. Finally, a childcare problem with potentially significant regulatory filings.
Humans choose the outcome
Moderator Debra Aho Williamson, founder and chief analyst of Sonata Insights, asked which marketing decisions humans should retain as AI assumes more planning duties.
Manning’s answer was immediate: “The outcome.”
She warned that AI could become exceptionally proficient at optimizing gameable measurements. Marketers must ensure that models are trained with the right data, measurement systems and business goals.
Rogers described the division of labor as an 80-20 model. AI can handle repeatable tasks, analyze data and reach answers faster. People still need to determine whether those answers make sense.
“You still need that spark of human ingenuity to create what’s next,” he said.
The technology can free employees from manipulating spreadsheets and give them more time for strategic work. It can’t rely entirely on historical data to invent the future, especially if every brand wants ideas its competitors haven’t already used.
Cohen added that human judgment remains critical for connecting AI-generated insights to a brand’s long-term vision and purpose.
Marketing to people and their agents
Looking ahead, Rogers predicted the rise of agent-to-agent and agent-to-consumer marketing. Brands will need to communicate not only with shoppers but also with the digital assistants making recommendations on their behalf.
Manning expects evidence of genuine performance to separate useful products from “AI washing.”
“It’s like slap the word agentic on it, like lipstick on a pig,” she said.
She also predicted increasingly personalized consumer experiences using images and video. As an example, she described uploading a photo of her face to a chatbot for skincare and makeup recommendations. It advised her on microcurrent treatments, blush placement and a specific product.
The next performance-marketing playbook, then, may involve advertising to people, their agents and the agent critiquing their blush before a panel. Humans will still set the objective, assuming they can get the machines to stop taking credit for 30% sales growth in a 3% world.
You’re watching Beet.TV coverage from Cannes Lions 2026. For more videos from this series, please visit this page.
CANNES, France – Artificial intelligence has spent the past few years promising to remake advertising. Now it’s being asked to do something considerably harder: produce results.
Executives from Chalice.AI, Horizon Media and OMD US discussed how AI is changing campaign planning, buying and measurement during the “AI, Automation & the New Performance Playbook” panel at the Cannes Lions International Festival of Creativity.
The industry has moved beyond philosophical debates about what AI might eventually accomplish, according to Bradley Rogers, chief executive of OMD US.
“Fast forward a year, we’re working with AI every single day,” Rogers said.
He organized the opportunity into three areas: operations, discoverability and execution. Much of the initial work has focused on automating workflows and improving efficiency. Clients are now paying closer attention to discoverability as consumers increasingly use AI tools to research products and make decisions.
That change could disrupt the customer journey that marketers spent decades mapping, color-coding and occasionally pretending was linear.
Search starts with a problem
Michael Cohen, executive vice president of performance media at Horizon Media, said consumer searches are becoming more focused on tasks and problems instead of specific products.
That shift requires agencies to improve the intelligence governing decisions across multiple platforms. The challenge is coordinating those choices without forcing individual platforms to relearn campaigns and waste money in the process.
Cohen said AI’s contribution comes in “two flavors that are not mutually exclusive.” One is efficiency. The other is “bona fide insight” that can inform planning, optimization and budget allocation.
He compared their combination to the center lever on a soft-serve machine. This may be the first description of AI architecture that also makes people want ice cream.
The math needs to math
Ali Manning, chief operating officer of Chalice.AI, said the biggest change is that some brands can now point to business outcomes from AI instead of simply discussing its potential.
“What I’ve found most interesting this year is there are brands out there talking about what AI has done for them,” Manning said.
She cited results including incremental sales for Hershey’s during Halloween and increased allergy-market share for Bayer’s Claritin. These are the kinds of numbers that chief executives discuss on earnings calls, unlike a heroic click-through rate admired mainly by the people who created the report.
Manning said AI can help advertisers optimize toward meaningful business goals instead of relying on whatever metrics platforms make available. That distinction becomes especially important when every platform claims credit for the same sale.
“All the platforms come in and they’re like, ‘We drove 30% more sales year over year.’ And then the CFO is like, ‘But sales only grew 3%,’” she said. “The math doesn’t math.”
Agents need supervision
The discussion also attempted to distinguish agentic AI from conventional automation, a useful exercise now that attaching “agentic” to a product name has become the technology industry’s equivalent of adding truffle oil to french fries.
Manning said today’s bidding agents can receive instructions and determine the right price for reaching a particular consumer in a particular context. More advanced systems will learn and begin taking actions on their own.
“We’re moving into a world of agentic where agents don’t just take instructions,” she said. “They learn and then start to self-act.”
That future remains nascent, particularly in media execution. Manning said marketers can use the current generation of agents without immediately leaping to autonomous systems that raise questions about liability and suitability.
Her own head of marketing had left an AI agent running at home to continue working on marketing plans and a website. The panelists joked that someone needed to babysit it in case it decided to acquire another company. Finally, a childcare problem with potentially significant regulatory filings.
Humans choose the outcome
Moderator Debra Aho Williamson, founder and chief analyst of Sonata Insights, asked which marketing decisions humans should retain as AI assumes more planning duties.
Manning’s answer was immediate: “The outcome.”
She warned that AI could become exceptionally proficient at optimizing gameable measurements. Marketers must ensure that models are trained with the right data, measurement systems and business goals.
Rogers described the division of labor as an 80-20 model. AI can handle repeatable tasks, analyze data and reach answers faster. People still need to determine whether those answers make sense.
“You still need that spark of human ingenuity to create what’s next,” he said.
The technology can free employees from manipulating spreadsheets and give them more time for strategic work. It can’t rely entirely on historical data to invent the future, especially if every brand wants ideas its competitors haven’t already used.
Cohen added that human judgment remains critical for connecting AI-generated insights to a brand’s long-term vision and purpose.
Marketing to people and their agents
Looking ahead, Rogers predicted the rise of agent-to-agent and agent-to-consumer marketing. Brands will need to communicate not only with shoppers but also with the digital assistants making recommendations on their behalf.
Manning expects evidence of genuine performance to separate useful products from “AI washing.”
“It’s like slap the word agentic on it, like lipstick on a pig,” she said.
She also predicted increasingly personalized consumer experiences using images and video. As an example, she described uploading a photo of her face to a chatbot for skincare and makeup recommendations. It advised her on microcurrent treatments, blush placement and a specific product.
The next performance-marketing playbook, then, may involve advertising to people, their agents and the agent critiquing their blush before a panel. Humans will still set the objective, assuming they can get the machines to stop taking credit for 30% sales growth in a 3% world.
You’re watching Beet.TV coverage from Cannes Lions 2026. For more videos from this series, please visit this page.
