Over the last three years, ad agencies of all kinds rushed to embrace the use of artificial intelligence. Debates about how much to outsource to AI in terms of production and creativity have remained fraught, but resistance has generally abated.
Results are extremely mixed: Some media agencies placing ads via programmatic came out way ahead of their creative peers while others dragged farther behind.
For one thing, machine learning has served as the very backbone of “big data-driven” automation in programmatic. But as the rest of the world has been employing generative AI into their processes regularly for some time, it’s taken a little longer for media agencies to find unique uses for the technology.
But one major infrastructure shift accelerated that change in the last year: the emergence of Model Context Protocols (MCPs). These function as wrappers around the individual APIs that advertising platforms run on. Where traditional integrations required media buyers to manage a fragmented tangle of platform-specific connections, MCPs give AI agents a common way to understand what each system can do and can translate natural-language instructions from a human into the specific API calls needed to execute them. In other words, instead of logging into a platform and pulling levers themselves, humans can give instructions to the AI agent via chat, leaving the grunt work and intricacies of navigating the platform to the AI.
Told to “activate a client’s high-value customer segment across programmatic with Q4 creative and optimize for ROAS above 4x” an agent can, in theory, handle that entire workflow without a buyer touching each platform individually.
The work of the media buyer doesn’t disappear; but where it happens, and at what level of abstraction, has shifted to specific areas of quality control and oversight
“The question stops being ‘are you using AI?’ and starts being ‘has it made your decisions better, and can you prove it?’“ says Nick Walden, VP of innovation and strategy at True Media.
The shift he describes is less about technology adoption than about where AI has settled in the workflow. “It’s less a headline and more a habit built into the middle of the work,” Walden says, such as pressure-testing a recommendation before it reaches a client, catching a gap that might otherwise slip through.
What’s changed structurally, according to practitioners, is where the heaviest work now happens. Alicia Gehring, SVP of media strategy at full-service agency WHITE64, provides an example: “Media work used to be front-loaded. Most of the heavy lifting happened in planning, before a campaign ever went live,” Gehring says. “This year, more than ever before, the weight shifted downstream: platform evaluation, CPM negotiation, and real-time optimization now eat up as much time as the planning did.”
AI’s role in that shift, she says, is less about writing the plan and more about accelerating the thinking that used to take days. It functions as a fast thought partner for evaluating media partners, surfacing their strengths and weaknesses against a specific campaign brief — a task that used to require real digging, particularly in B2B and niche categories where the right partner isn’t obvious.
Walden traces the same pattern from a different angle. Finding information, he says, is no longer the challenge. “Now it’s sorting through what’s there with tools and workflows to align on clear strategies and recommendations, faster than ever before.” The volume of available data and signals has outpaced any team’s ability to process it manually, so AI has become a crucial mechanism for filtering all of that information, rather than generating it.
To Stephen Lee, head of Data and Analytics at Mediaplus North America, the fundamentals of great media agency work haven’t changed. Clients still need sharp thinking, honest collaboration, and creative solutions that move their business. What has changed in the last year is the speed at which we can deliver it, he says.
“AI also hasn’t replaced the principles of media planning, but it has unlocked an accelerated workflow to apply those principles,” Lee adds. “The planner who once spent days (or weeks) building out a channel and tactical plan now can spend more time stress-testing it, knowing the research, audience analysis, RFP comparisons and scenario planning can be done in a fraction of the time. The analyst buried in dashboards is now using (or building) agents that surface insights before anyone thinks to ask the question. This is the shift with AI; allowing time for more curiosity and better judgment.”
Anne DiNapoli Block, chief data and technology officer at 22squared and Trade School, calls what agencies are living through “the velocity revolution.”
“AI has fundamentally changed how much we can produce, how quickly we can move, and how intelligently we can operate,” Block says. “We’re automating workflows that once consumed significant time, accelerating analysis and optimization, and giving our teams more capacity to focus on outcomes.”
What hasn’t changed, she argues, is the underlying requirement: “Strong strategy, creativity, judgment and a deep understanding of the consumer. AI doesn’t replace those things. It amplifies them.”
Block’s point speaks to something neither Walden nor his peers are ready to cede to automation: media buyers’ and planners’ judgment. “AI is outstanding at generating a dozen plausible answers, but it can’t tell you which one is right for this client, this market, this quarter,” Walden says.
That distinction — between generating options and choosing among them — runs through nearly every account of how AI has integrated into agency work.
Rachel Hipschman, senior director of growth at digital marketing and performance media agency Collective Measures, describes a workforce that has been genuinely transformed at every level, but whose core expertise requirements haven’t changed.
“What we need to know as media experts has remained the same, but how we come to know it has deeply and fundamentally changed,” Hipschman says.
Media strategists who once spent hours combing through Excel data now glean audience insights in minutes, redirecting their attention toward deeper thinking about consumer behavior. Channel experts set up campaign quality assurance (QA) faster and spend more time on optimization. Data scientists surface actionable insights daily rather than weekly.
Mike Baranowski, VP of analytics and data engineering at Collective Measures, frames the firm’s benchmark for AI success in terms that go beyond efficiency. “We have not focused on efficiency alone as a measure of AI success,” Baranowski says. “While the speed at which we can respond to the market is critical, we focus on outcomes as the true measure of adoption and effectiveness.”
Media agency work continues to evolve rapidly, with pressure to do more with less while driving growth, notes Darshan Sampathu, EVP, Media & Analytics at Allen & Gerritsen.
“AI has accelerated that evolution, but our approach is pragmatic,” he says. “We focus on where AI makes us smarter, faster, and more effective, recognizing that every new AI capability isn’t automatically better for clients.”
For instance, Allen & Gerritsen is deploying Google Ads AI Max for clients with “enhanced organic content,” Sampathu says. In essence, they’re using AI to dynamically deliver ads that convey the brand’s core message more precisely. It could also allow for more personalization.
“We expect bolder experimentation across AI and more creator content to blend, how paid amplifies organic content more effectively, with agencies playing an even greater role in determining not just what’s possible, but what’s valuable,” Sampathu says
The practitioners closest to this work share a consistent expectation for the year ahead: AI gets quieter, more embedded, and more accountable.
Walden anticipates it disappearing into the background of work — what one team learns in research carrying directly into what planning builds and activation runs, rather than sitting in separate documents. Hipschman expects agencies to further automate lengthy processes around creative versioning and local campaign management, freeing practitioners to step more fully into the role of trusted client advisor.
Gehring arrives at a different, perhaps more forward-looking question, one that cuts against the current orthodoxy of hyper-personalization. “How much longer does 1-to-1 personalization actually win?” she asks. “I have a growing hunch that 1-to-many — loud, well-targeted, and unmistakably there — is going to out-perform hyper-personalized messaging that nobody notices in the scroll.”
It’s not just the role of AI that is changing, the role of the individual is, too, Mediaplus’s Lee says. “We empower people to be entrepreneurial: to experiment and build AI solutions that improve their own work and sharpen what we bring to clients. The people closest to the work know where the friction is, AI gives them the tools to help solve it and we actively encourage ‘hackathon’ mentalities to bring new ideas forward.”
Over the next few years, Lee expects AI to become less visible but more embedded into solutions and workflows. “We will still position solutions to client challenges but focus more on the effectiveness, output and recommended way forward vs. any AI powering it.”
“What comes next?” is the kind of question that no AI tool is going to definitively, uniquely answer. Which is precisely the point.
