Agentic Buying Turns Media Planners Into Guardrail Architects For Autonomous Campaigns

Brand marketers are about to hand real-time media decisions to autonomous systems they did not build, and the rules for what those systems can and cannot do with the brand are still being written. Standards bodies, DSPs, and publishers are all building the technical infrastructure for agentic media, and the brand-safety and KPI guardrails meant to sit atop it are still mostly blank pages.

Matteo Del Bianco manages digital consumer experience, social, paid media, and omnichannel strategy for the Ploom brand at JTI (Japan Tobacco International). His 15 years in digital media have taken him through 35 global markets and brands, including Swatch, Frédérique Constant, Diesel, and Nestlé. Working in tobacco, one of the most heavily regulated categories in advertising, has trained him to treat brand safety as something that must fit within the machine's instructions from the start. The most valuable work in media now, he says, happens before any agent touches a campaign.

"The human role will never disappear from the equation. Humans just move a little bit upstream in the process, from operational work into strategic business governance," Del Bianco says.

Agentic infrastructure is outpacing its rulebook

The IAB Tech Lab has published its agentic roadmap for digital advertising, followed by a draft framework of management protocols that enable autonomous systems to communicate and execute across the bidstream. The trade press has been all over the standards work, and media agencies are already piloting planning agents with one eye on the full buying tools behind them. "We are still in discovery. It is not yet clear how the industry will fully embrace agentic media buying," Del Bianco says.

Most planners already know where this ends up. The open question is when. Agents will eventually take over real-time bidding, duplicate path mitigation, traffic filtering, and budget reallocation across channels. What was once a fragmented set of channel-by-channel workflows becomes more fluid, with budgets automatically shifting from underperforming channels to better-performing ones based on live signals. 

The shift is already visible in programmatic DOOH, where cross-channel optimization has the shortest runway and where the complexity of coordinating inventory across formats, geographies, and media owners makes the case for agent-assisted execution strongest. Amazon's recently open-sourced dynamic traffic engine signals where filtering will go at production volume. Once shared protocols let those systems talk to each other, the operational plumbing of media buying will speed up dramatically, but the strategic layer above it—the part that decides which markets to weight, which supply partners to trust, and how to balance reach against brand safety—stays with the people who understand the medium.

Guardrail design is the planner's new craft

When agents handle execution, the marketer's job moves to design. The craft is translating a CMO's objective into a machine-readable prompt. The oversight work is defining supply standards, brand safety thresholds, and KPI hierarchies. The two together produce a layer of guardrail architecture that sets the rules the agent will play by.

"In a traditional setup, a human trader manually inputs targeting, selects age brackets, and paces budgets. In an agentic environment, the human writes the objectives and defines the rules," Del Bianco says. "Humans are the people who should set the boundaries for the agent's playground. The human will still be the one setting the core KPIs, ensuring those goals are achieved by letting the agent define how to get there."

That work mirrors a broader shift inside marketing organizations. Marketing org charts are being redrawn to bring people and AI agents together within shared accountability structures, and creative teams are already operating within brand governance systems that decide which generative outputs are cleared for distribution. Media planning is an extension of a discipline that marketing leaders practice elsewhere in the organization.

Tobacco gives Del Bianco a sharper version of the oversight problem than most marketers face. An autonomous system optimizing for visual appeal could drop a legal disclaimer and hand the brand a compliance breach before the campaign finishes its first flight. "I come from a very restricted industry. Even when I use Gen AI to multiply the number of assets, at the end of the day, there are key elements within a specific asset that need to be respected, such as the legal disclaimer, to make sure that we operate and display in a specific environment with brand safety," he says.

Vanity metrics have lost their defensive power

CMOs are asking media teams to do more with less. The post-GDPR attribution environment has made it harder to reconcile platform metrics with on-site analytics, and impressions and clicks no longer hold up under finance scrutiny. Analyst forecasts already point to generative AI absorbing programmatic in the near term, with agentic systems following close behind. The squeeze on budgets and the gaps in attribution are what give the case for AI-enabled optimization its weight in finance reviews.

"I cannot go to leadership and simply claim we generated 2 million impressions when privacy implications cut the traffic you can actually monitor. When there is no reflection of that on your website traffic, they naturally question how we have the courage to ask for more budget without showing any direct impact on the business," Del Bianco says.

Attention metrics have taken over his executive reviews. "I now go to my managers with attention metrics to justify our awareness campaigns. The critical change is shifting our focus from whether an ad was delivered to whether an ad was actually seen," he says. The shift from delivered to seen has older roots in OOH, where memory and brand-building measurement have been the proof standard for years.

Supply path optimization is the practical on-ramp

Before any team turns execution over to a fully agentic system, supply path optimization gives marketers a way to see where their dollars actually land and to cut the intermediary fees that have hollowed out programmatic for years. Buyers apply DSP filters to block low-quality inventory and partner with curation platforms such as Jounce Media to access vetted private marketplaces. Some go further by entering into direct supply contracts and negotiating quality KPIs outside the open exchange. The technical lift on SPO is a fraction of what full agentic buying demands, which is why adoption is moving faster here.

"Supply path optimization in reality is something that can already be implemented, and it is much faster in terms of adoption, much easier in terms of adoption than moving from regular buying to agentic buying," Del Bianco says. The work doubles as preparation for what comes next. Buyers who get fluent in defining supply standards, vetting partners, and enforcing quality thresholds today are practicing the judgment that will define their roles once agents take over execution.

New KPIs belong inside the prompt

The KPIs that will guide agentic systems differ from those on the legacy ad tech dashboard. Impressions, clicks, and CPM give way to business outcomes, attention metrics, ROAS, qualified leads, supply efficiency, and brand safety constraints. Those metrics need to be included in the prompt that drives the agent. Hand the system a single signal, and it will chase that signal to the floor, usually with the brand picking up the tab.

"If you are only focusing on generating ROAS, then maybe you are focusing on lower-funnel activities. There is a set of KPIs that need to be entered into your prompt to guide the agent in delivering sales while also protecting your brand and bringing value to the other activities you are doing," Del Bianco says. The same logic is reshaping brand safety in programmatic, where AI-driven context analysis is replacing static keyword block lists, and constraints are being applied within the bid evaluation itself.

Del Bianco says the burden of proof has moved with that infrastructure, from the brand onto the providers building it. "From a brand perspective, you need to update your KPIs, make sure they are modern, fresh, aligned with new capabilities, and ingest them in your prompt. The critical challenge and where the pressure really lies is on the other side, with the media agency and the publisher. Now we are getting the tools that allow us to deliver on the business requirements that have been there for ages, but we just did not have the tech to achieve or report on them," he says.

Between the blueprint and the buildout

The vision assumes conditions the industry has not yet met. The first is talent. Del Bianco estimates that roughly 70 percent of today's media planners lack the strategic fluency to operate as guardrail architects. "Agencies need to stop training people to push buttons on walled-garden consoles and start molding them into business consultants," he says. "Otherwise, they risk falling behind."

The second is data. In an agentic ecosystem, a brand's first-party data infrastructure becomes either its core competitive advantage or its single point of failure. Feed the agent fragmented or non-compliant signals, and it will optimize for nothingness. "The real job of tomorrow's media planner is not just writing the prompt. It is ensuring the cleanliness, compliance, and exclusivity of the data plumbing that powers the agent."

And the third is interoperability. A truly fluid cross-channel agentic environment assumes open protocols between platforms, and Meta, Google, and Amazon have historically resisted that kind of access. Before an agent can reallocate budgets across detached ecosystems in real time, the industry must resolve what Del Bianco calls a "protocol war." Universal adoption of shared frameworks like those proposed by the IAB Tech Lab cannot remain limited to independent DSPs. It has to become the baseline for the entire digital ecosystem.

The skills that matter sit further upstream

Almost none of the practitioner's skill set retires here. Marketers who trace how pixels fire, how line items get built, and where DSP filters collide with publisher direct deals are the same marketers who catch the autonomous system the second it starts running its own agenda. Same toolkit, new target.

"One of the main skills is understanding AI overall and the pros and cons of using it. When you let the machine make decisions and optimize the campaign on your behalf, you need to make sure the machine is working with reliable data. You need to make sure that the ad tech and the infrastructure still consider everything that is protecting your brand," Del Bianco says.

The new talent is fluent in AI literacy, scenario planning, strategic prompting, KPI modernization, supply chain judgment, and the instinct to keep the brand intact while the systems chase the numbers. For Del Bianco, that shift is in the process of becoming a reality within the team. The traders who used to live inside bidding consoles every afternoon now spend that time writing constraints and pulling the agent back when it goes off-brief—same glowing dashboards, different instructions getting typed. "We're all working and moving in this direction through POCs," says Del Bianco. "We're not 100% there yet, but it's becoming a reality."

The views and opinions expressed are those of Matteo Del Bianco and do not represent the official policy or position of Japan Tobacco International or any other organization.