Future of Marketing Briefing: AI has created a new talent paradox in programmatic agencies
This Future of Marketing Briefing covers the latest in marketing for Digiday+ members and is distributed over email every Friday at 10 a.m. ET. More from the series →
It’s been another busy week for big news but I wanted to use this week’s briefing to write about something that’s been sitting with me since the Digiday Programmatic Marketing Summit last week: AI is flattening the base of the talent pyramid before agencies have figured out how to rebuild it.
Session after session, leaders admitted that the work that once served as training wheels for juniors is evaporating. Reporting is auto-generated. Performance summaries write themselves. Pieces of planning now run in the background. The muscle memory that used to form by doing the repetitive stuff is no longer being built as intensely.
One moment captured that shift. During a closed-door town hall, an agency exec walked through an internal experiment meant to test how AI is rewiring media planning. The findings weren’t really about speed or efficiency. It exposed a deeper question of how value is distributed inside teams — and how unsteady that foundation has become.
The agency had three teams build the same media plan for a client. One team consisted of AI specialists who don’t work on that client but are deeply trained on the agency’s proprietary tools. Another team worked on a different brand for the same client. They understood the category and the client’s expectations but had only light exposure to the agency’s AI tools, leaning mostly on ChatGPT to speed things up. A third team built the plan without using AI at all.
The result was telling: the AI specialists, despite having no brand context, produced a plan that closely resembled what the agency would normally deliver through its traditional process. The old-school team did too. The weakest work came from the middle — the group that had client familiarity but hadn’t fully internalized how AI now fits into the workflow.
It felt like a small but revealing snapshot of a broader talent paradox inside programmatic ad teams. AI isn’t eliminating the need for people. It’s demanding that teams make a clean commitment to one operating model or another. The AI specialists succeeded because they understood how to work inside the new system. The traditional team succeeded because they followed a logic they knew. The group in the middle struggled not because they lacked planning skills but because they tried to graft old instincts onto AI tools they didn’t full understand. That’s the risk agencies are staring at — a layer of talent caught between two modes of working, where partial adoption becomes more destabilizing than no adoption at all.
“They [the execs in the middle team] requested more training on AI because they said it would have been really helpful had they known what kind of buttons to press,” said the agency exec during the town hall.
They’re far from the only ones asking.
Programmatic leaders kept circling the same worry throughout the summit: the next generation of managers may never develop the instincts that come from doing the laborious work AI is now doing for them.
“We’ve gotten so good at automating a lot of the entry-level work,” said PMG’s head of programmatic Doug Paladino. “It isn’t impacting jobs right now, but I worry that over time, if we truly need fewer entry-level people, then three to five years from now we won’t have the people needed to be future managers.”
Like many of his peers, he doesn’t yet have the solution. For now, it’s something he and his team are actively working through. They’re giving junior talent earlier exposure to the diagnostic backbone of programmatic — signal interpretation, strategic reasoning and platform mechanics — not through grunt work but by supervising and interrogating richer data. Instead of learning through manual QA and repetitive reporting, traders are being handed access to the core of the operation: log-level signals, wrapped in AI interfaces they can query intuitively.
“We’ve layered in an LLM to log-level data so now you can just ask it: ‘what shows am I delivering on’, and the they get the answers back,” said Paladino. “It’s really about getting the most valuable, complex data set in the hands of the folks who can do the most with it — our day-to-day campaign operators — without them needing advanced training.”
What’s emerging is a different kind of junior altogether — one expected to reason across channels, interpret patterns, question machine outputs and move comfortably across creative, data and supply. The job isn’t execution anymore. AI handles that. The job is judgement.
“Those agency roles won’t disappear,” said Matt Barash, chief commercial officer at Nova, a creative tech platform that partners with agencies globally. “The skillset will evolve into harnessing better AI produced outputs that feed core automation and infrastructure platform investments across agencies. Talent will still be foundational to manage those operating systems.”
At independent advertising agency BarkleyOKRP, for example, teams are increasingly cross-trained — planners, traders, analysts, even creatives — to ensure they understand not just their own lane but how AI-driven decisions reverberate across the entire media plan.
“We essentially want all rounders who don’t just look at activation, media planning, strategy or search,” said Reshma Karnik, the agency’s chief media officer. “We want to have a holistic thought process.”
Which comes back to the core issue. It’s hard to think holistically about anything if the very experiences that once built that perspective are slipping away.
At the same time, the work that remains is more technical and interconnected. Junior staff are expected to interpret impressions-level signals, understand how supply choices influence attribution and move across functions that used to sit in distinct lanes. In other words, the baseline for competence is rising while the traditional ways of building it are eroding. The ladder is still there but the rungs that teach people how to climb it are disappearing. And harder still, there are fewer junior roles available to train in the first place. Agencies need broader, more capable talent earlier — just as the foundation that used to produce that talent is being squeezed from both ends.
“My concern isn’t the learning curve, it’s the shrinking number of entry-level roles,” said Rui de Freitas co-founder and CEO at ad tech vendor C Wire. “Before hiring, most companies will try to automate the process. So agencies must reinvent how juniors learn: give them real problems earlier and let them supervise AI, not sit behind outdated workflows.”
Numbers to know
58%: Combined global market share (excluding China) that Alphabet, Amazon and Meta are expected to take in 2026.
61%: Percentage of Gen Z shoppers that used AI tools to help with them with a purchase last year.
41%: Percentage of TikTok’s global ad revenue that is generated by its U.S. users.
35%: Percentage by which Reels reach has dropped, possibly as a result of new friction in Instagram’s algorithm.
What we’ve covered
Agencies push curation upstream, reclaiming control of the programmatic bidstream
Agencies told Digiday that they plan to tighten curation in 2026, reclaiming some control over what the term actually means — and how it’s used in practice.
In Graphic Detail: Here’s what the creator economy is expected to look like in 2026
Digiday used five charts to explain how the industry is expected to expand next year, factoring in revenue, the types of creators brands are planning to work with, and which platforms will be key for that content.
‘A trader won’t need to leave our platform’: PMG builds its own CTV buying platform
The independent ad agency has built its own CTV buying platform called Alli Buyer Cloud, which sits inside PMG’s broader operating system Alli.
Why 2026 could be Snap’s biggest year yet – according to one exec
Snap’s senior director of product marketing, Abby Laursen, had a chat with Digiday about its plans for automated advertising next year, and how its partnership with Perplexity lays the foundations for it to create an ecosystem of AI-powered intelligent agents in the platform.
What we’re reading
Disney has signed a three-year deal to become the first major licensing partner on OpenAI’s AI-video app Sora, according to Variety.
Pinterest Saving 90% on AI Models Through Open Source, Ready Says
While other companies are paying tons of cash to use AI, Pinterest, it seems, has found a more cost efficient approach. The platform’s CEO Bill Ready said the team has saved 90% when using brand-name AI by relying on open source models, according to The Information.
Where Meta’s biggest experiment in governance went wrong
Meta’s Oversight Board was designed to act as the platform’s own version of a Supreme Court — but it has mostly failed to meet expectations, according to Platformer.
Meta’s new A.I. superstars are chafing against the rest of the company
The platform’s TBD AI Lab has created an us-versus-them mentality against its longtime execs, according to The New York Times.
