Few corners of marketing attract as much breathless forecasting as media buying. Every January, the industry's conference circuit declares a new era; every December, practitioners quietly report that most of the old machinery is still running. The truth usually sits in between. Media buying is being reshaped right now by several powerful forces at once, automation and agentic AI, the rise of retail media, the convergence of TV and digital, and a long, messy adjustment to data privacy. But the pace and shape of that change are easy to overstate.
This piece tries to separate the structural shifts from the hype, and to describe where the practice of buying media is genuinely heading.
The execution layer is being automated - slowly, then suddenly
The most visible change is the steady automation of the buying task itself. Platform tools like Meta's Advantage+ and Google's Performance Max already absorb much of what used to be hands-on work: audience discovery, placement selection, bidding, and creative optimisation are increasingly handled by the platform's own models rather than by a human operating a console.
The newer story is "agentic" AI, systems that don't just optimise within a campaign but plan, negotiate, and execute across the supply chain with limited human oversight. The infrastructure is being built in plain sight. PubMatic launched its AgenticOS framework in early 2026 with agency partners including WPP Media and MiQ, positioning it as an operating system for agent-to-agent advertising in which buyers set objectives and guardrails and a set of agents executes within them. Yahoo's demand-side platform has been testing a cluster of agents for optimisation and measurement, and Microsoft has reported building thousands of agents on its Copilot platform. Emerging interoperability standards — the Model Context Protocol and agent-to-agent communication frameworks, point toward a future in which buy-side and sell-side agents transact directly.
Yet the people who actually spend the money remain notably pragmatic. Industry voices at CES 2026 described agentic buying as "inevitable but incremental" and not ready for prime time, and the IAB Tech Lab has predicted several false starts before practical adoption, warning that real progress will require years of standardisation and alignment across platforms, agencies, and publishers. There is also a structural skepticism worth taking seriously: the programmatic supply chain is famously noisy and adversarial, and training autonomous systems on that substrate risks making existing blind spots permanent rather than smarter.
The grounding fact is that programmatic still works at enormous scale. US programmatic ad spending is set to surpass $200 billion in 2026, with most automated buys still transacted through direct deals rather than open auctions. Automation is advancing, but it is layering onto a mature system, not replacing it overnight.
Retail media has become a third pillar
If agentic AI is the most discussed shift, retail media is the most consequential one already in motion. Retailers have turned their first-party shopper data and owned digital shelves into advertising businesses, and the money has followed. Global retail media spend reached roughly 15% of digital ad spend in 2025 and is on track to approach the mid-hundreds of billions of dollars in 2026, growing at double-digit rates and now ranking alongside search and social as a top-tier channel. By most forecasts it will overtake social media spending within a few years.
Two features explain the pull. The first is closed-loop attribution: because the retailer sees both the ad exposure and the purchase, advertisers can tie spend to sales with a directness that the open web rarely offers. The second is durable, consented first-party data, which matters more as third-party signals erode.
The channel is also concentrating. Amazon commands the overwhelming majority of US retail media revenue, by some estimates close to 80%, with a scaled second tier forming beneath it and a long tail of smaller networks struggling to break through. For buyers, this creates a familiar tension: retail media is highly measurable and increasingly unavoidable, but it pulls budget into closed environments where the network grades its own homework.
TV and digital are finally merging
Connected TV is the channel where the long-promised convergence of brand and performance media is actually happening. CTV ad spend is expected to exceed $40 billion in 2026 as budgets continue migrating from linear, and addressable formats let advertisers target households and audience segments rather than buying time slots on a program. With roughly 60% of US households now cord-cutters or cord-nevers, the audience has shifted decisively.
CTV's strengths come bundled with hard problems. The ecosystem is fragmented across devices, operating systems, and content owners, each with its own inventory and measurement quirks. And while CTV never depended on cookies, it is not immune to signal loss: IP addresses, the connective tissue for much of CTV targeting and attribution, are unreliable identifiers whose match rates decay quickly over time. The result is a channel that is growing fast and maturing in capability, but where measurement remains the unsolved frontier.
The data ground keeps shifting - even after the cookie reprieve
For half a decade, the industry organised its planning around the death of the third-party cookie. Then, after repeated delays, Google reversed course: by April 2025 it had confirmed it would not roll out a dedicated deprecation prompt in Chrome and would instead maintain a user-choice model, leaving third-party cookies in place rather than removing them.
The reprieve changed the timeline, not the trajectory. Cookies continue to erode through browser defaults, blocking, and deletion, and they were never available at scale on Safari, in apps, or on CTV. The investments made during the "cookieless" scramble, first-party data strategies, data clean rooms, identity frameworks like Unified ID 2.0, and contextual targeting, remain relevant precisely because addressability is degrading regardless of any single browser decision. A parallel pressure is mounting from generative search: as more queries are answered without a click, some publishers have reported steep declines in referral traffic, squeezing the open-web inventory that independent buyers have long relied on.
Consolidation and the migration of value
Underneath these channel shifts, the economics of the intermediary are moving. The major agency holding companies have seen their share of US ad spending fall sharply over recent years, from the mid-40% range in 2019 to roughly 30% by 2024, on the figures cited by eMarketer, even as overall ad spend has grown. Part of the response has been consolidation, driven in no small part by the capital required to build enterprise-grade AI and to assemble proprietary data assets.
The deeper question is where value accrues. As platforms build the agents that buy their own inventory, and as retailers and streamers monetise their own first-party data, more of the buying intelligence, and more of the margin, risks migrating to the sell side and the walled gardens. An independent buyer's value in that world is less about operating the tools and more about keeping the platforms honest: auditing performance across closed environments, negotiating terms, and guarding against the conflicts of interest that arise when a network optimises and measures its own results. Long-running transparency debates, including industry scrutiny of principal-based media buying, become sharper, not softer, in an agent-mediated market.
What this means for the role
Put together, these forces don't abolish the media buyer so much as pull the role apart. The commoditised execution layer i.e. manual bidding, placement, and optimization, is genuinely being automated, and headcount tied purely to operating platforms will contract. But the work that sits before and around the buy is more durable: strategy and channel-mix decisions, creative and measurement design, governance of automated systems, and the accountable relationship a client needs when something goes wrong. The likely trajectory is a shift from execution toward orchestration and oversight, designing the objectives and guardrails that agents run within, then auditing what they do.
The honest forecast, then, is neither disruption nor business as usual. Media buying is becoming more automated, more concentrated in a handful of large platforms and retailers, and more measurable in closed loops, while becoming harder to evaluate transparently across the open ecosystem. The buyers who thrive will be the ones who move up the value chain faster than the routine parts of their job are absorbed by the machine. The transition is real. It is also slower, messier, and more contested than the keynote stages suggest.
Sources
- Digiday, CES 2026: Agentic AI hype vs. media buyers' pragmatism (Jan 2026)
- Digiday, Media agencies test AI planning agents, while edging toward buying tools (Dec 2025)
- TensorOps AdTech Lab, Agentic AI in Advertising: A 2026 Field Guide (2026)
- eMarketer, FAQ on ad agencies: Consolidation, AI disruption (Jan 2026) and Programmatic Advertising Forecast and Trends H1 2026
- eMarketer / Fugo, Retail Media Growth, Statistics, and Trends for 2026; eMarketer Retail Media Ad Spending Forecast H1 2026; Osmos, Future of Retail Media 2026
- Power Digital, What Is CTV? How Connected TV Advertising Works (Feb 2026); eMarketer, CTV is vulnerable to disruption from signal loss
- Google Privacy Sandbox announcements (April 2025) as reported by ExchangeWire, Didomi, and Usercentrics
- AdExchanger, Why CTV is Positioned to Thrive as Third-Party Cookies Fade Away and related 2026 coverage on zero-click search
Figures are drawn from third-party forecasts and trade reporting current as of early-mid 2026; market estimates vary by source and methodology.
