By Alex Vasekin, CEO at AdTech Holding

For the past two years, most conversations about AI in adtech have focused on what it can produce: from copy to reports and analytics, you name it. That was a low-hanging fruit. But today we all know that it was just scratching the surface.

AI is moving from generating individual outputs to running entire processes: launching campaigns, adjusting bids, generating creatives, and deciding what happens next.

And that decision-making part changes everything. An AI tool waits for a person to use it. An AI operating system continuously observes, decides, acts and learns.

We see this across the ecosystem: development, campaign management, analytics, traffic quality, fraud prevention, internal operations. What’s emerging isn’t a random collection of AI features but rather a whole new model for how advertising infrastructure works.

Automation is changing the unit of competition

New app releases on the App Store and Google Play grew 60% year over year in Q1 2026, and 104% in April – a pace Appfigures links to AI coding tools that let people with little to no tech background vibecode and ship real products, all competing for the same audience we see today.

One of the clearest signals is campaign volume: at AdTech, campaigns created per advertiser rose 42% over four months in 2026, with the sharpest increase following February’s new generation of AI models. We expect that number to grow five to ten times over the next 12–18 months.

Humans have not suddenly become five or ten times faster at filling in campaign forms. Campaign creation is increasingly delegated to scripts, APIs and AI agents, which changes the economics of media buying. eMarketer has called 2026 “the beginning of the end” for manual programmatic buying, as the industry rushes from generative AI that helps people create to agentic AI that acts on its own.

Historically, scale was constrained by operations. A team could only research, launch and monitor so many campaigns. Automation weakens that: one buyer can now test more offers, geographies and traffic segments without growing the team.

Our media-buying team built a production tracker comparable to established platforms in about a week, using AI for development and MCP connectors to link it to internal systems, and since then stress-tested it on more than 200 million daily impressions.

A media buyer with no development background used an AI agent to build a hybrid CPM optimisation system: AI handled the decision logic, scripts executed and re-checked actions on schedule. Within a week it was generating over $500 daily at roughly 60% ROS, largely hands-off.

The strategic consequence is bigger than productivity: when execution becomes cheap, it stops being an advantage. The scarce skills move elsewhere. Now choosing what to test, feeding the system good signals, knowing when the model is wrong becomes the human domain.

Data is becoming part of the bid

Automation also changes what it means to buy traffic well. Bidding has long been treated as a pricing problem – how much to pay for an impression, click or user. Machine learning turns it into an information problem too.

An algorithm deciding which traffic to buy needs feedback. If the only signal is a click or registration, it optimises toward clicks or registrations. If it also sees deposits, repeat purchases or rebills, it can distinguish users who merely enter the funnel from those who create value.

AdTechinternal Q2 data illustrates how large that difference can become. Using advertisers that send no conversion events as the baseline, advertiser lifetime value was 2.9 times higher among accounts sending registration or install data, 6.3 times higher when a paid action was also shared, and 21.6 times higher among advertisers sending three or more types of deeper funnel events, including repeat monetary actions.

This should change how advertisers think about conversion data. Sending it back is usually treated as an attribution task, increasingly, it’s part of media buying itself: it trains the system deciding what traffic you get next, and withholding it while expecting better optimisation is a contradiction.

The same logic extends beyond conversion events: every conversation with an AI agent or account team becomes another signal. Among more than 2,000 new advertisers, deeper conversations correlated with activation rates three times higher than minimal contact. The system optimises on what you tell it, not just what you buy.

The interface is starting to disappear

Another sign of the change is at the product layer. Traditional ad platforms are built around UIs, like a dashboard, targeting settings, campaign creation, reports. AI agents make that workflow optional.

Advertisers already feel this shift: more than half report that AI has reshaped how consumers discover and consider what to buy. According to McKinsey research, success increasingly means being surfaced and selected by the systems shaping consumer decisions, not just securing impressions. Gartner projects that autonomous, agentic AI systems acting on a user’s behalf will increasingly replace channel-based marketing altogether. 

Our advertisers can describe an offer to our own AI agent conversationally and have it turn that brief into a live campaign, then keep optimising it. The numbers show the shift: brief to live in minutes rather than days, first launch about five times faster than manual, and 52% of users come back. Companies with their own automation stack can connect via MCP instead. It is a pattern already forming industry-wide, with the IAB Tech Lab announcing plans for a reference MCP server for advertising in 2026. 

A marketer may still use a dashboard for manual control. But an AI agent works through APIs, manages hundreds of campaigns at once, and acts continuously, not just when someone logs in on Monday morning.

That means APIs are becoming core product surfaces, quite a step up from integration features. Platforms will increasingly serve two kinds of users – humans and agents – accessing the same capabilities very differently.

AI also compresses the window for abuse

Automation cuts both ways. The same tools that let legitimate advertisers launch and optimise faster also make it faster to spin up a fraudulent campaign, rotate infrastructure, or tweak a deceptive flow. Manual review just can’t keep up anymore.

That’s why AI and automated detection are already baked into how AdTech Holding handles traffic quality and policy enforcement. ADEX, for example, analyses abuse patterns across traffic to identify invalid activity, malware and other risks.

The important shift is speed: when campaign creation happens in minutes, moderation can’t run on a multi-day cycle. Detection must move closer to real time and correlate more than the creative – account behavior, infrastructure, traffic patterns, destinations.

Attackers automate. Buyers automate. Platforms automate. Defenders automate. The ecosystem as a whole becomes machine-speed.

The real advantage is the system around the model

This is why we try to approach it as a decision on which parts of the operating model AI can actually run instead of which AI tools to add.

These are two fundamentally different approaches. In the first model, AI is yet another tool. You still design the workflow and make the decisions and AI is just a faster and more effective version of Excel, Google search or StackOverflow. It improves productivity, but the way the business operates stays largely the same.

The second model goes further. AI becomes part of the operating system. People set the direction and the boundaries, while AI handles the day-to-day flow, making routine decisions and adjusting on the spot. The human is still in control, but no longer needs to be in every loop.

This model is less tidy: agents still make mistakes and need guardrails, deterministic checks and clear handoff points to a human. But they adapt far faster than a manually designed workflow.

The winners will not be the companies with the best prompt library or the most AI features. They will be the companies that build an environment in which machines can safely make useful decisions: reliable data, accessible APIs, clear objectives, guardrails and human accountability.

The advertising companies that build that environment first will not simply perform the same work faster.

They will operate differently.

About the author

Alex Vasekin is the CEO of AdTech Holding, an EU-based advertising technology and cybersecurity group whose brands include PropellerAds, ADEX, Notix, Zeydoo, Remoby, AppLabz, and ProPush. With more than 20 years of experience in the industry, he has helped grow the group into a global business serving audiences at scale across advertising, monetization, and cybersecurity. His current focus is on how AI is changing advertising operations and decision-making: not by replacing human judgment, but by making it faster and better informed.