Originally published on Forbes with commentary by Ted Sweetser, VP of Strategic Partnerships
Marketing has shifted from AI to agentic marketing. All major brands, including HubSpot, Adobe, Meta, Amazon, Google and others, recently rolled out new agentic marketing tools, while countless companies are choosing to develop their own agentic marketing platforms in-house.
For decision-makers, CMOs and marketing teams, this new digital transformation and the flood of new tools present them with several questions. What can these new tools do? Is a full marketing adtech overhaul needed? How are in-house agentic marketing tools built and how can brand safety risks, from fraud to ad misplacement to cybercriminal operations, be navigated safely?
In this report, executives and experts from Zappi, PurpleLab and CyberLuna answer the hard questions and offer clear signals for decision-makers to navigate this new shift.
Back in 2025, Mark Zuckerberg shook the advertising industry when he presented his vision for the future of marketing. In that future, Zuckerberg said an advertiser would simply state their business goals, point the tech tool to a website or product image and connect his credit card to the account to run the campaign. An AI agent would handle everything.
Technically, as in the technology available, that future is already here. However, adoption is slow. A recent survey from the Boston Consulting Group (BCG) found that only under a third of CMOs have moved to AI agent-led workflows. Understanding what agentic technologies can do in marketing is a way to bridge adoption gaps. So, what can agents do today?
"Agents are creating a third system of decision-making,” Steve Phillips, Founder, Chair and Chief Innovation Officer at Zappi, the Boston-based consumer insights platform, told me.
“The real power comes when different agents focus on different jobs or different sources of information and then work together,” said Philips.
Instead of looking at advertising, brand and innovation separately, agents can connect those signals and give marketers a much fuller picture of what is happening across all of them and why, Philips explained.
New agentic marketing tools can orchestrate the full campaign lifecycle, from generating creative assets to programming ad placement, managing ad inventory, budgets, targeting and optimizing omnichannel distribution, while continuously monitoring performance metrics.
“At this point we may better ask what agentic marketing tools can't do,” Ted Sweetser from PurpleLab, an analytics and technology company working in healthcare, told me.
Many publishers are using AI to reassess how they sell ad inventory, their highest-margin offering, in light of agentic tools' ability to handle so many more of these functions, said Sweetser.
“Marketers do not need to rip out the technology they already have,” Phillips from Zappi said.
The bigger opportunity is to make those tools work together in a way they largely do not today, Phillips explained. Most marketing systems are still fragmented, with different platforms holding different pieces of the customer picture, said Philips.
“Agents can sit across those systems, connect the signals and help marketers understand the whole rather than each platform in isolation,” Phillips added. The value comes from turning the stack into a more connected system that fuels fast decisions with evidence, said Phillips.
On the other hand, Sweetser from PurpleLab added that agentic tools are integrating with existing adtech tools in one of two ways. The first one includes the use of agentic tools available in the market, which all work and look nearly identical, said Sweetser.
“The second way they integrate, which is likely to be the more long-enduring, is via a framework called MCP (model context protocol), which is a means of giving an agentic tool access to the adtech vendors' goods and services,” said Sweetser.
MCP is an open-source, free framework that allows agents to connect to accounts and different data sources, accounts, workflows, tools, and even other agents. MCP is used in a wide range of industries, from cybersecurity to advertising. More importantly, using MCP, a company can limit an agent's behaviour and restrict what it accesses and shares, a vital function to meet compliance and cybersecurity demands and keep spending and costs within expected ranges; for example, when an agent can make bid for ads and make payments.
“The reason this (MCP in-house agentic marketing tech integration) is likely to endure more than individual third-party vendors’ adtech solutions is that virtually every company of significance is building its own agents,” said Sweetser.
Regarding tech overhaul, Sweetser added that given the expansiveness of agentic software applicability, virtually every company should be assessing where they can use agents to increase efficiency, but benefits will vary depending on the company and the tech.
Whether building in-house agentic marketing models using MCP or using third-party tools, brand safety issues need to be considered. These include governance and compliance, ad fraud and click-bot farms, brand impersonation and ad fraud and the cybercriminal threat environment.
“Brand safety with agents is fundamentally a knowledge problem,” Philips from Zappi said.
Phillips explained that the question is not just, ‘How capable is this AI?’ but 'What does it know about my brand, my consumers and the rules we operate by?'
“You would not put one person in charge of creative, legal, consumer insight and brand governance and expect them to be an expert in all four,” said Philips. “AI should not work that way either.”
“Different agents should take responsibility for different areas, with clear rules, strong data and human judgment where the stakes are high,” said Philips.
“Good infrastructure gives agents better context, connects what the organization already knows and helps the system get smarter, not just faster,” said Philips.
One of the key areas of brand safety is compliance. In the United States, companies need to meet several regulations at a federal and state level. The Federal Trade Commission (FTC) enforces advertising laws at the federal level that range from truthfulness and evidence-based requirements to the Children's Online Privacy Protection Act, and laws for endorsements, influencers and reviews, environmental marketing, health claims and other laws.
At the state level, where compliance varies depending on the state, new AI laws are emerging. California, for example, is moving ahead with a new AI-advertisement-specific law. With brand impersonation on the rise, deepfakes and synthetic media, which are illegal in almost every U.S. state, are also central for compliance. Other regions like Europe have even stronger regulatory demands for advertising, while in Asia new and old laws present advertisers with a fragmented landscape which they must comply with.
“In the U.S., fake reviews, including AI-generated fake reviews, are already covered by FTC rules and brand impersonation is also an enforcement issue,” Ran Geva, CEO of CyberLuna, a free, enterprise-grade exposure monitoring platform used by security teams, enterprises and public sector organizations worldwide, told me.
“The (agentic) technology may be new, but the company is still responsible for what is published, purchased, or promoted on its behalf,” said Geva.
Geva added that the biggest mistake marketers can make with agentic AI is assuming that automation also includes good judgment. “An agent told to maximize reach, engagement and conversions will optimize for these goals, but it won't know what not to do,” said Geva.
“It may end up buying low-quality inventory, rewarding fake engagement, interacting with bot farms, or generating content that crosses compliance or brand-safety lines,” he added.
Another hotspot is ad fraud. 2026 reports reveal a massive surge in ad fraud, fueled heavily by sophisticated, AI-driven bots and click farms. Juniper Research projects $170 billion in ad spend lost to fraud by 2028, and said that in 2023, 22% of ad spend was lost to ad fraud. TrafficGuard says that every minute in the U.S., where the cost of a click is among the highest globally, $180,000 is lost to click fraud. Based on TrafficGuard’s customer data, global ad fraud losses are currently set at $250 billion.
Besides ad fraud, companies running marketing campaigns are being targeted by industrial-level, highly automated cybercriminal operations that impersonate known and trusted brands, their new products, and their ads.
Brand impersonation is one of the most popular techniques in the cybercriminal threat landscape. For example, the recent cybercriminal operation PasteSwitch hijacked HBO Max's Reddit account to run hundreds of ads. That threat campaign was part of a broader scheme where OpenAI Codex, Claude AI, Alfred, Homebrew, GitHub, and crypto wallet applications were also impersonated.
Detecting impersonation campaigns and taking action to shut down the fake websites and malicious online resources is fundamental to avoid reputational damage and secure brand safety.
“The marketing executive role needs to change,” Geva from CyberLuna told me.
“They can delegate execution to an agent, but they cannot delegate accountability.”
Companies need to define where agents can advertise, what sources of traffic they can use, what they are allowed to create or say about the brand and which actions require human approval, Geva explained. They also need independent measurement of traffic quality rather than trusting the same platforms or agents that are optimizing the campaigns, he added.
“My rule would be simple: give AI agents freedom to optimize inside a controlled environment, not freedom to decide what constitutes acceptable marketing,” said Geva. “The faster and more autonomous the agent becomes, the stronger the verification layer around it needs to be.”
Marketing teams, CMOs and decision-makers today face multiple challenges on different fronts. From an explosion of agentic marketing tools, to understanding how agentic AI works and integrating the tech into their operations to remain competitive, while safely navigating brand safety high-risk issues.
As agentic marketing tools evolve, keeping up to date with their capabilities and limiting what an agent can do, while having full visibility into its operations, is what experts recommend.
Despite the compliance, cybersecurity and technical complexities that agentic marketing brings forward, a full tech stack is not required, and opportunities across the board exist to drive engagement and maximize budgets safely and compliantly.