AI Is Only as Smart as the Data Behind It: Audience Path’s Sean Black
Artificial intelligence can process billions of signals, deploy autonomous agents and tell advertisers what consumers want before consumers have finished wondering why their refrigerator needs a software update.
But Sean Black, general manager and senior vice president of Audience Path by Dailymotion, says marketers should probably ask one basic question before surrendering the advertising industry to the machines: What exactly were they trained on?
Speaking with Beet.TV contributor David Kaplan at the EMARKETER Future of Digital conference in New York, Black said Audience Path is using Dailymotion’s enormous volume of video-viewing data to help advertisers understand audiences based on what people are actually doing online.
Dailymotion sees roughly 5 billion video streams a month, providing Audience Path with a large pool of behavioral information. “We’re a video technology company at the end of the day,” Black said. “So AudiencePath is really the audience part of the business.”
The consumer is doing something right now
The advertising industry has spent years accumulating enough data to know practically everything about consumers except why they keep clicking “reject all” on cookie banners.
Black said Audience Path is trying to move beyond audience building that starts with cookies or seed audiences. Instead, it examines real-time behavior and combines video consumption with search, social activity, geographic and demographic information.
“Ours is real-time behavior based on what a consumer is looking at,” Black said. The goal is to understand “who that person is today and what their intent is now.”
That distinction matters as brands struggle to extend their reach beyond existing datasets. Audience Path is designed to identify relevant consumers across connected TV, online video and desktop rather than treating each screen as its own small bureaucratic republic.
“We’re looking at it as, how do I reach the right audience based on their behavior, regardless of the environment at the end of the day?” Black said.
Electricity wasn’t exciting enough
No discussion of advertising technology in 2026 can proceed for more than several minutes without somebody invoking AI, presumably under penalty of being reassigned to the fax machine division.
Audience Path introduced an agentic platform called Ray about a year ago. It uses Dailymotion’s billions of streaming signals and custom AI agents that communicate with one another to analyze consumer behavior.
Black said AI is necessary simply because of the scale involved.
“AI is important because there’s no other way to consume and analyze that kind of information,” he said.
But Black said the more important question is what powers the AI in the first place.
“What I’m challenging the marketplace to start thinking about, though, is what are the agentic platforms actually trained on?” he said. “That’s the biggest deal.”
Ray, he said, is trained on data Dailymotion owns rather than rents. That gives Audience Path a defined foundation for its models instead of assembling an AI system from whatever happens to be floating around the internet, humanity’s magnificent digital archive of scientific discovery, restaurant complaints and arguments between strangers about whether the moon landing happened.
Brand versus performance refuses to die
AI may be rewriting civilization, but advertising still has unfinished business from approximately 1997: deciding whether brand advertising and performance marketing belong together.
Black thinks the distinction is increasingly artificial.
“There’s never just a brand campaign,” he said. “It’s never just a performance campaign. They all now still work together.”
Audience Path is developing products that can bring more attribution to channels such as connected TV while avoiding the temptation to judge every CTV impression like a banner ad.
“We don’t want CTV all of a sudden to become another banner campaign and have all of that accountability,” Black said. Instead, he said marketers should consider CTV’s contribution to other activity across a campaign.
Pharma has a particularly difficult data problem
Pharmaceutical marketers face another problem: reaching relevant audiences without casually converting personal health information into something resembling a supermarket loyalty program.
“Pharmaceutical ultimately has a privacy problem,” Black said. “It’s very difficult.”
Pharma brands also can struggle to achieve scale because treatments may apply to relatively small patient populations. Black said behavioral data can help marketers identify patients, caregivers and other relevant audiences while maintaining audience quality.
The broader approach starts with Dailymotion’s first-party streaming information, including what people watch, how they watch it and where. Audience Path then combines those signals with search, social and geo-demographic information.
Five billion streams and one important question
For Black, the next stage of AI in advertising is less about attaching the word “agentic” to another product announcement and more about determining whether the technology is working from information that actually matters.
“There’s too much AI, agentic, all this crazy stuff they’re talking about,” Black said.
The important question, he added, is: “What are you really training these tools on to be valuable for your business?”
It is an almost suspiciously practical question for an era in which artificial intelligence is simultaneously expected to eliminate drudgery, cure disease, replace half the workforce and perhaps usher humanity toward an extinction event that makes the Atomic Age look like a moderately tense episode of “Leave It to Beaver.”
In the meantime, advertisers still have quarterly targets. And 5 billion video streams to analyze.
You’re watching Beet.TV’s coverage of EMARKETER’s 2026 The Future of Digital summit. For more videos from this summit, please visit this page.
Artificial intelligence can process billions of signals, deploy autonomous agents and tell advertisers what consumers want before consumers have finished wondering why their refrigerator needs a software update.
But Sean Black, general manager and senior vice president of Audience Path by Dailymotion, says marketers should probably ask one basic question before surrendering the advertising industry to the machines: What exactly were they trained on?
Speaking with Beet.TV contributor David Kaplan at the EMARKETER Future of Digital conference in New York, Black said Audience Path is using Dailymotion’s enormous volume of video-viewing data to help advertisers understand audiences based on what people are actually doing online.
Dailymotion sees roughly 5 billion video streams a month, providing Audience Path with a large pool of behavioral information. “We’re a video technology company at the end of the day,” Black said. “So AudiencePath is really the audience part of the business.”
The consumer is doing something right now
The advertising industry has spent years accumulating enough data to know practically everything about consumers except why they keep clicking “reject all” on cookie banners.
Black said Audience Path is trying to move beyond audience building that starts with cookies or seed audiences. Instead, it examines real-time behavior and combines video consumption with search, social activity, geographic and demographic information.
“Ours is real-time behavior based on what a consumer is looking at,” Black said. The goal is to understand “who that person is today and what their intent is now.”
That distinction matters as brands struggle to extend their reach beyond existing datasets. Audience Path is designed to identify relevant consumers across connected TV, online video and desktop rather than treating each screen as its own small bureaucratic republic.
“We’re looking at it as, how do I reach the right audience based on their behavior, regardless of the environment at the end of the day?” Black said.
Electricity wasn’t exciting enough
No discussion of advertising technology in 2026 can proceed for more than several minutes without somebody invoking AI, presumably under penalty of being reassigned to the fax machine division.
Audience Path introduced an agentic platform called Ray about a year ago. It uses Dailymotion’s billions of streaming signals and custom AI agents that communicate with one another to analyze consumer behavior.
Black said AI is necessary simply because of the scale involved.
“AI is important because there’s no other way to consume and analyze that kind of information,” he said.
But Black said the more important question is what powers the AI in the first place.
“What I’m challenging the marketplace to start thinking about, though, is what are the agentic platforms actually trained on?” he said. “That’s the biggest deal.”
Ray, he said, is trained on data Dailymotion owns rather than rents. That gives Audience Path a defined foundation for its models instead of assembling an AI system from whatever happens to be floating around the internet, humanity’s magnificent digital archive of scientific discovery, restaurant complaints and arguments between strangers about whether the moon landing happened.
Brand versus performance refuses to die
AI may be rewriting civilization, but advertising still has unfinished business from approximately 1997: deciding whether brand advertising and performance marketing belong together.
Black thinks the distinction is increasingly artificial.
“There’s never just a brand campaign,” he said. “It’s never just a performance campaign. They all now still work together.”
Audience Path is developing products that can bring more attribution to channels such as connected TV while avoiding the temptation to judge every CTV impression like a banner ad.
“We don’t want CTV all of a sudden to become another banner campaign and have all of that accountability,” Black said. Instead, he said marketers should consider CTV’s contribution to other activity across a campaign.
Pharma has a particularly difficult data problem
Pharmaceutical marketers face another problem: reaching relevant audiences without casually converting personal health information into something resembling a supermarket loyalty program.
“Pharmaceutical ultimately has a privacy problem,” Black said. “It’s very difficult.”
Pharma brands also can struggle to achieve scale because treatments may apply to relatively small patient populations. Black said behavioral data can help marketers identify patients, caregivers and other relevant audiences while maintaining audience quality.
The broader approach starts with Dailymotion’s first-party streaming information, including what people watch, how they watch it and where. Audience Path then combines those signals with search, social and geo-demographic information.
Five billion streams and one important question
For Black, the next stage of AI in advertising is less about attaching the word “agentic” to another product announcement and more about determining whether the technology is working from information that actually matters.
“There’s too much AI, agentic, all this crazy stuff they’re talking about,” Black said.
The important question, he added, is: “What are you really training these tools on to be valuable for your business?”
It is an almost suspiciously practical question for an era in which artificial intelligence is simultaneously expected to eliminate drudgery, cure disease, replace half the workforce and perhaps usher humanity toward an extinction event that makes the Atomic Age look like a moderately tense episode of “Leave It to Beaver.”
In the meantime, advertisers still have quarterly targets. And 5 billion video streams to analyze.
You’re watching Beet.TV’s coverage of EMARKETER’s 2026 The Future of Digital summit. For more videos from this summit, please visit this page.
Series Videos
