By Craig Aron, Regional Director, Head of AdTech GTM
Advertising Week 2026 brought together more than 1,300 speakers, 500 sessions, and an estimated 80,000-plus connections over four days in New York. The agenda covered just about everything shaping the future of marketing, from AI and commerce to creators and culture.
I listened to a few panels, but spent most of my time in scheduled meetings and impromptu run-ins. And after a few days, the same themes surfaced. Publisher traffic is drying up. Apple keeps making life harder for AdTech. The open web is dying. Platforms are consolidating. More decisions are moving to the supply side. And, of course, AI agents.
There’s a common thread here, and it’s a fight for control. Who owns the audience? Who has access to the data? And what happens when technology starts making decisions that used to belong to people? But some of my most interesting conversations weren’t about the industry’s biggest challenges. They were about how people themselves are using AI in their personal and professional lives.
Some occasionally open ChatGPT, others practically live in Claude Code or Codex, and a few have gone all in on personal agents like Meta’s Muse. The Muse users were particularly interesting. They weren’t just experimenting with it. They were all in, far more than with traditional chat tools. Personally, I’ve had early access to Instinct, a Muse competitor, and I’ve seen something similar in my own habits. My AI usage is up roughly 10x, and I was already a heavy user.
Clearly something big is happening with consumer use of agentic platforms and personal agents. According to Sensor Tower estimates reported by TechCrunch, Muse reached 10 million US downloads in 40 days. Inside businesses, adoption is moving more slowly. McKinsey’s 2026 State of AI report found that only about 20% of organizations have scaled AI agents across their operations.
Consumers aren’t waiting around. I see a clear parallel to the “prosumer” movement of the SaaS era, where enterprise adoption of software spread through companies long before IT departments had formal strategies for managing them. I suspect AI agents will follow a similar path. If the adoption of agents in the enterprise mirrors this pattern, we’re about to see it accelerate at a pace that few organizations are prepared for.
Control by another name
Thinking about what this means for AdTech brought me back to the theme of control. Not about the control of data or audiences, but the control over the AI agents that are being deployed into production. There’s plenty of enthusiasm about what AI agents will do for advertising. Planning campaigns, building audiences, making bid decisions, optimizing spend, measuring performance. The list keeps growing.
What I heard less about at Advertising Week were the underlying systems on which these agents are deployed in production. Every new agentic use case places real demands on the data architecture underneath. Having the data foundation that enables a complete understanding of how agents are deployed, what actions they have taken, and the reasoning and inference behind those actions is no longer optional.
Think about a buyer agent making thousands of decisions against live inventory. It needs access to enormous amounts of data, and it needs that access immediately. If it’s working from sampled data, incomplete records or information that’s already out of date, those limitations don’t disappear because an AI agent is making the decisions. Try explaining to your board why your buyer agent spent millions of dollars on fraudulent inventory without a proper audit trail.
That’s not a hypothetical technical concern. It’s the kind of operational risk companies need to think about before they put agents in charge of real budgets.
The data problem underneath the AI conversation
For years, AdTech companies have made compromises in how they store and query data. Sampling, sketches, batch processing and complicated pipelines have all helped manage the cost and complexity of working with enormous datasets.
Those compromises become a lot harder to live with when agents are making decisions continuously.
I think there are four things that become especially important:
- Access to complete data. Agents need to work from the actual bids, impressions and targeting signals from billions of ad auctions, not sampled approximations.
- Speed at scale. Waiting minutes or hours for a query isn’t practical when AdTech decisions are happening in real time.
- A unified view of the data. Bidder logs, audience information and other structured or semi-structured data need to be accessible without constantly stitching together different systems and pipelines.
- A record of what happened. Companies need to know which agents took which actions, what information they used and why those decisions were made.
This is a big part of what we’re building at Ocient.
OcientAIQ™ brings AI agents and analytics directly to full-resolution, petabyte-scale AdTech data, giving programmatic platforms an AI-ready data foundation. The goal is to make it possible to work with enormous datasets quickly, without giving up the detail or traceability that matters when real money is involved.
You can get a brief overview from our Principal Product Manager Neil Kumar in this video.
I came away from Advertising Week invigorated by the explosion of creativity and innovation we are already seeing as AI agents move from demos to scaled production. I’m more convinced that the agentic wave is going to come through bigger and sooner than many companies expect.
Companies that figure out the underlying data problem first will have a significant advantage. If you’d like to see how OcientAIQ can help your company, get in touch today.