Agentic AI is reshaping CTV ad buying, but the playbook is unwritten. Here's what SEA marketing teams need to know before the infrastructure locks in.
Autonomous AI agents making real-time media buys without a human in the loop — it sounds like a 2028 problem. According to AdExchanger, it’s already being piloted in connected TV today.
What Agentic Buying Actually Means in CTV
Agentic AI, in the ad buying context, refers to systems that don’t just recommend actions — they execute them. In CTV, that means an AI agent that can independently query inventory, evaluate audience signals, negotiate programmatic deals, and place buys across multiple SSPs, all within parameters set by the advertiser. It’s the logical endpoint of automation: you define the goals, the guardrails, and the budget, and the agent runs.
But AdExchanger’s reporting is careful to distinguish ambition from reality. Most current implementations are closer to assisted automation than true agency — AI that can handle a narrow slice of the buying workflow, like frequency management or daypart optimization, rather than end-to-end campaign orchestration. The infrastructure for full agentic buying in CTV is still being assembled, and crucially, the trust frameworks — how advertisers audit what agents did and why — don’t yet exist in any standardized form.
For teams used to having a human make the final call before a dollar moves, that’s not a small gap.
Why CTV Is Both the Right and Wrong Place to Start
CTV is an appealing testbed for agentic buying for structural reasons: inventory is fragmented across dozens of streaming platforms, audience data is relatively rich compared to linear TV, and deals increasingly flow through programmatic pipes that AI can interface with. The efficiency case is real — a well-configured agent could theoretically outperform a human buyer on pure optimization speed across that fragmented landscape.
But CTV also has characteristics that make autonomous buying genuinely risky. Brand safety controls are less mature than in display or social. Measurement is still being standardized across platforms — in Southeast Asia, where streaming adoption on platforms like Vidio (Indonesia) and Disney+ Hotstar is growing fast but fragmented, there’s no single clean signal for reach or frequency. An agent optimizing against incomplete data doesn’t just underperform; it can confidently drive in the wrong direction.
The other wrinkle is inventory quality. CTV’s premium positioning is partly maintained by the friction in its buying process. Remove that friction via full automation and you risk agents arbitraging their way into lower-quality placements that technically hit the brief but don’t deliver the brand environment the buyer assumed they were getting.
The Identity Problem Underneath the Automation Promise
Agentic buying doesn’t solve the identity problem — it inherits it. Any AI agent making CTV buys is only as good as the audience signal it’s optimizing against. In a cookieless environment, that signal increasingly comes from first-party data clean rooms, contextual inference, and platform-specific IDs that don’t travel cleanly across walled gardens.
For Southeast Asian brands, this is a compounding challenge. Mobile-first markets mean users move fluidly between app-based streaming, mobile web, and smart TV environments — often without persistent IDs that link those touchpoints. An agentic buyer optimizing for a target segment across, say, Netflix, YouTube, and a local FAST channel in Thailand is working with three different identity frameworks that may not reconcile. The agent can be smart; the underlying data architecture has to be smarter.
The brands that will extract value from agentic CTV buying earliest are those that have already done the unsexy work: first-party data consolidation, clean room partnerships with key publishers, and audience taxonomy that holds up across platforms. Agentic AI accelerates execution — it doesn’t fix a broken data foundation.
What a Sensible Pilot Looks Like Right Now
Given the nascency of the technology and the identity infrastructure gaps, a full agentic CTV deployment in 2026 is a research project, not a media strategy. But there are structured ways to learn.
A constrained pilot might look like this: isolate a single streaming platform with strong first-party data (think a platform like iQIYI, where logged-in users provide consistent identity signals), define a narrow optimization objective (cost-per-completed-view, not brand lift), set hard guardrails on placement and spend velocity, and run the agent alongside a human buyer for the same brief. The comparison isn’t about which performs better on day one — it’s about understanding where the agent’s logic diverges from the buyer’s judgment, and whether that divergence is a bug or a feature.
Three things to monitor closely: how the agent handles inventory shortages (does it hold or compromise on quality?), how it responds to mid-flight signal changes, and whether its audit trail is legible enough for post-campaign analysis. If you can’t reconstruct why the agent made a specific buy, you can’t improve the system — and you can’t defend it to a CMO asking why the brand appeared next to something it shouldn’t have.
Agentic AI in CTV buying is not a question of whether — it’s a question of when the infrastructure catches up to the ambition, and which brands have built the data foundations to extract value when it does. The more interesting strategic question is this: as buying becomes increasingly automated, where does human expertise actually live in the media process — and are marketing teams investing in the right capabilities to stay relevant when the agents take over the execution layer?
At grzzly, we work with marketing teams across Southeast Asia navigating exactly this shift — from identity architecture to programmatic strategy to understanding which automation plays are ready for prime time and which are still vaporware. If you’re trying to figure out where agentic AI fits in your media infrastructure, we’d rather have that conversation before the pilots go sideways. Let’s talk
Sources
Written by
Rogue GrizzlyOperating at the contested frontier of cookieless targeting, clean rooms, and identity resolution. Comfortable where the infrastructure is shifting and the playbooks have not yet been written.