Agentic AI is Changing Where Ad Dollars Flow: PubMatic’s Bill McLaughlin
Advertisers have spent years adding more technology to media buying, presumably because the existing technology hadn’t yet produced enough dashboards, acronyms or opportunities to stare thoughtfully at a screen.
Agentic AI promises something different. Rather than simply giving buyers another tool to operate, autonomous systems can help decide where media dollars should go in the first place.
That’s beginning to change how advertisers approach campaigns, according to Bill McLaughlin, senior vice president of advertiser solutions at adtech firm PubMatic, who spoke with Beet.TV contributor David Kaplan about how agencies and brands are experimenting with agentic media buying.
Start with the outcome
“At PubMatic, we’re seeing agentic really change the starting point of how buyers think about media,” McLaughlin said.
Instead of beginning with a channel, platform or vendor that already has a claim on the budget, buyers can start by defining the business result they want. An agent can then look across supply sources for combinations of performance, inventory and efficiency.
That seemingly simple reversal could have big consequences for an advertising business historically organized around channels, partners and budget silos. McLaughlin said agentic workflows are making it easier for advertisers to explore performance-driven connected TV, direct supply and outcome-based buying that might previously have been too cumbersome to activate.
“The real impact isn’t just automation,” he said. “It’s understanding where and how buyers change their media decisions and ultimately where money can flow.”
In other words, AI may do more than save somebody from clicking through 47 screens before lunch. It could challenge some of the assumptions about which platforms and suppliers get considered at all.
Guardrails that teach the machine
Handing an autonomous system real advertising dollars does, however, raise a small matter traditionally known as “What could possibly go wrong?”
That’s where governance comes in.
“Guardrails are not just about brand safety,” McLaughlin said. “They’re giving buyers control and creating trust in a new way of buying media.”
Buyers need visibility into how agents make decisions and the ability to establish limits around those decisions. But McLaughlin sees another purpose for those restrictions: They can teach the agent.
An agent might identify a premium placement that exceeds a buyer’s CPM ceiling or discover useful inventory outside an approved list. Instead of merely hitting a digital brick wall, the system can treat that boundary as information. Traders can then see the opportunity and decide whether the rule itself deserves reconsideration.
That turns governance from a mechanism designed merely to prevent AI misbehavior into part of the learning process.
McLaughlin said guardrails can create “a framework that allows the agent to learn while keeping the buyer in control.” Done correctly, he added, governance could become a competitive advantage rather than simply a safety mechanism.
Give the humans something better to do
Agentic systems also could change the division of labor inside agencies.
Agents can perform continuous tactical discovery at a scale that would be difficult for human teams to replicate. The resulting learning can become specific to an individual client or campaign, effectively giving advertisers specialized analytical capabilities without having to staff an army of specialists.
The human strategist doesn’t disappear in McLaughlin’s scenario. Instead, people spend more time applying expertise, understanding client relationships and deciding what to do with the information the machines uncover.
Agencies can also combine agents with their first-party data, proprietary algorithms and established strategies. The goal is to scale those assets without turning every campaign into the same AI-generated bowl of oatmeal.
Independents get more muscle
That possibility could be particularly significant for independent agencies and mid-market advertisers, which don’t have the staffing or technology resources of the largest holding companies.
McLaughlin described agentic technology as “a powerful unlock” for independents because it lets them scale the characteristics that already differentiate them, including vertical expertise, proprietary data and specialized algorithms.
PubMatic’s role, he said, is to provide more of the data, technology and execution infrastructure while agencies supply the strategy and expertise.
The result could be more time for higher-value work and an opportunity for independent agencies “to punch above their weight” against much larger media organizations.
So, the machines aren’t only coming for people’s jobs. They may also be coming for some organizational charts.
CTV and AI channels move past the science project
The shift arrives as advertisers face another familiar problem: an ever-growing assortment of places to spend money.
Connected TV, live sports and AI-native environments such as chat interfaces are creating new opportunities for advertisers. Many emerging channels initially land in the experimental-budget bucket, where interesting ideas sometimes live happily for years without ever meeting serious money.
McLaughlin thinks that’s changing. Technology is making CTV and live sports easier to access while improving measurement and performance capabilities. CTV campaigns increasingly can be connected with sales outcomes. AI-native environments, meanwhile, could reach consumers while they’re actively discovering information and products.
“The difference with these channels today isn’t that they’re just new and exciting,” McLaughlin said. “It’s that they’re becoming scalable, measurable performance opportunities.”
That gives agencies a path from experimentation to proving impact and eventually incorporating those channels into their core media strategies.
From testing agents to shaping them
Simply announcing an AI pilot may soon stop qualifying as an AI strategy.
McLaughlin said buyers need to move beyond testing agentic systems and become active participants in developing them. That includes choosing technology partners, co-innovating with them and being clear about what the systems need to accomplish and what obstacles remain.
At the same time, agencies need to protect and strengthen the assets that distinguish them, including their data, models, algorithms and strategic expertise.
“Agents are only as good and powerful as the data and strategy behind them,” McLaughlin said.
That may be the less glamorous side of the agentic advertising revolution. Buying an AI system is relatively easy. Giving it proprietary knowledge, useful data, sensible rules and a strategy worth automating is considerably harder. And unfortunately for anyone hoping the robots would handle absolutely everything, that part still requires humans.
Advertisers have spent years adding more technology to media buying, presumably because the existing technology hadn’t yet produced enough dashboards, acronyms or opportunities to stare thoughtfully at a screen.
Agentic AI promises something different. Rather than simply giving buyers another tool to operate, autonomous systems can help decide where media dollars should go in the first place.
That’s beginning to change how advertisers approach campaigns, according to Bill McLaughlin, senior vice president of advertiser solutions at adtech firm PubMatic, who spoke with Beet.TV contributor David Kaplan about how agencies and brands are experimenting with agentic media buying.
Start with the outcome
“At PubMatic, we’re seeing agentic really change the starting point of how buyers think about media,” McLaughlin said.
Instead of beginning with a channel, platform or vendor that already has a claim on the budget, buyers can start by defining the business result they want. An agent can then look across supply sources for combinations of performance, inventory and efficiency.
That seemingly simple reversal could have big consequences for an advertising business historically organized around channels, partners and budget silos. McLaughlin said agentic workflows are making it easier for advertisers to explore performance-driven connected TV, direct supply and outcome-based buying that might previously have been too cumbersome to activate.
“The real impact isn’t just automation,” he said. “It’s understanding where and how buyers change their media decisions and ultimately where money can flow.”
In other words, AI may do more than save somebody from clicking through 47 screens before lunch. It could challenge some of the assumptions about which platforms and suppliers get considered at all.
Guardrails that teach the machine
Handing an autonomous system real advertising dollars does, however, raise a small matter traditionally known as “What could possibly go wrong?”
That’s where governance comes in.
“Guardrails are not just about brand safety,” McLaughlin said. “They’re giving buyers control and creating trust in a new way of buying media.”
Buyers need visibility into how agents make decisions and the ability to establish limits around those decisions. But McLaughlin sees another purpose for those restrictions: They can teach the agent.
An agent might identify a premium placement that exceeds a buyer’s CPM ceiling or discover useful inventory outside an approved list. Instead of merely hitting a digital brick wall, the system can treat that boundary as information. Traders can then see the opportunity and decide whether the rule itself deserves reconsideration.
That turns governance from a mechanism designed merely to prevent AI misbehavior into part of the learning process.
McLaughlin said guardrails can create “a framework that allows the agent to learn while keeping the buyer in control.” Done correctly, he added, governance could become a competitive advantage rather than simply a safety mechanism.
Give the humans something better to do
Agentic systems also could change the division of labor inside agencies.
Agents can perform continuous tactical discovery at a scale that would be difficult for human teams to replicate. The resulting learning can become specific to an individual client or campaign, effectively giving advertisers specialized analytical capabilities without having to staff an army of specialists.
The human strategist doesn’t disappear in McLaughlin’s scenario. Instead, people spend more time applying expertise, understanding client relationships and deciding what to do with the information the machines uncover.
Agencies can also combine agents with their first-party data, proprietary algorithms and established strategies. The goal is to scale those assets without turning every campaign into the same AI-generated bowl of oatmeal.
Independents get more muscle
That possibility could be particularly significant for independent agencies and mid-market advertisers, which don’t have the staffing or technology resources of the largest holding companies.
McLaughlin described agentic technology as “a powerful unlock” for independents because it lets them scale the characteristics that already differentiate them, including vertical expertise, proprietary data and specialized algorithms.
PubMatic’s role, he said, is to provide more of the data, technology and execution infrastructure while agencies supply the strategy and expertise.
The result could be more time for higher-value work and an opportunity for independent agencies “to punch above their weight” against much larger media organizations.
So, the machines aren’t only coming for people’s jobs. They may also be coming for some organizational charts.
CTV and AI channels move past the science project
The shift arrives as advertisers face another familiar problem: an ever-growing assortment of places to spend money.
Connected TV, live sports and AI-native environments such as chat interfaces are creating new opportunities for advertisers. Many emerging channels initially land in the experimental-budget bucket, where interesting ideas sometimes live happily for years without ever meeting serious money.
McLaughlin thinks that’s changing. Technology is making CTV and live sports easier to access while improving measurement and performance capabilities. CTV campaigns increasingly can be connected with sales outcomes. AI-native environments, meanwhile, could reach consumers while they’re actively discovering information and products.
“The difference with these channels today isn’t that they’re just new and exciting,” McLaughlin said. “It’s that they’re becoming scalable, measurable performance opportunities.”
That gives agencies a path from experimentation to proving impact and eventually incorporating those channels into their core media strategies.
From testing agents to shaping them
Simply announcing an AI pilot may soon stop qualifying as an AI strategy.
McLaughlin said buyers need to move beyond testing agentic systems and become active participants in developing them. That includes choosing technology partners, co-innovating with them and being clear about what the systems need to accomplish and what obstacles remain.
At the same time, agencies need to protect and strengthen the assets that distinguish them, including their data, models, algorithms and strategic expertise.
“Agents are only as good and powerful as the data and strategy behind them,” McLaughlin said.
That may be the less glamorous side of the agentic advertising revolution. Buying an AI system is relatively easy. Giving it proprietary knowledge, useful data, sensible rules and a strategy worth automating is considerably harder. And unfortunately for anyone hoping the robots would handle absolutely everything, that part still requires humans.
