From Programmatic to AI Agents and Lessons from 20 Years of Automation on Forgotten Infrastructure
- Jun 15
- 7 min read
Updated: Jun 19
Christopher Smith is an award-winning author and entrepreneur dedicated to protecting people from cybercrime. After being the target of a major cyberattack, he founded DFend, a digital safety platform, and wrote Privacy Pandemic, inspired by his real-life story.
In 2014, I was saying the same thing in every meeting, on panels, webinars, and in industry conversations: Programmatic advertising would reach linear television and become mainstream within five to ten years. Most people disagreed. Less than one percent of TV inventory had been sold programmatically. Most executives still believed television was fundamentally different, relationship-driven, premium, and resistant to automation. Why would you change how a $70 billion industry works? Automation and data.
That summer, I said it on camera during an industry interview with Beet.tv in Cannes and referenced a conversation with a senior executive at a major cable operator who had insisted that real-time television advertising would happen "over my dead body." The resistance was real, and the meeting ended quickly.
My team and I had spent the past year at Turn building one of the first buy-side programmatic TV platforms, running case studies with advertisers, data providers, new TV technology partners, and satellite operators. I was not speculating. I was pattern-matching.
More than a decade later, programmatic video, TV, CTV, FAST, and addressable advertising have become major pillars of the media ecosystem. Even many of the companies that once resisted automation now operate versions of the infrastructure they originally questioned.
Getting the prediction right required understanding something the market was missing: automation follows infrastructure, not the other way around. When automation races ahead of the infrastructure it depends on, the gap does not close on its own. It accumulates.
Twenty years later, that gap is no longer just an advertising industry challenge. It is becoming a digital safety challenge at a societal scale.
The load-bearing beam
I entered the ad-tech industry in the early formation years of programmatic first at Adap.tv, one of the first online video monetization platforms, then at Collective, where I spent three years watching programmatic scale from a niche capability into the dominant mechanism for multi-screen digital advertising.
We processed millions of data signals daily, making audience decisions across display, video, mobile, and the early edge of programmatic television. The efficiency was real. The dependency beneath it was becoming a structural challenge nobody wanted to name.
Programmatic's gains were never just about smarter bidding logic. They depended on continuity — the machine's persistent confidence that it was acting on the same user, household, or behavioral cluster over time. Every identifier framework promised more durable recognition. Cookies. Device IDs.
Cross-device graphs. Each one worked until it didn't. Each one ran into the same underlying challenge, just wrapped differently. Identity was not a feature of programmatic advertising. It was the load-bearing beam beneath the machine decisioning.
At Collective, we began addressing this directly through the TV Accelerator, launched in 2011 in partnership with Rentrak. The idea was to connect set-top-box viewing data from 19 million televisions with Collective's digital audience capabilities, allowing advertisers to reach people online who had already seen their TV advertising. It required solving for identity across two completely different data environments that were never designed to talk to each other. That experience shaped everything that followed.
The programmatic TV prediction and what it was really about
In 2013, my team and I scaled the global Emerging Media division at Turn and led the development of an early buy-side programmatic TV platform. By the time Cannes arrived in June 2014, I had watched automation move through display, video, mobile, and social media.
The pattern was consistent. Automation moves where inventory and audiences are. Television had both in abundance, and everyone in the industry was racing to unlock access to billions in traditional TV advertising dollars. The resistance was largely organizational and cultural, not 100% structural.
But the prediction was really about something more important than timing. It was about recognizing that each new wave of automation inherits the unresolved challenges of the prior one. Programmatic video inherited display's identity fragmentation.
Programmatic TV inherited everything that programmatic video had not resolved. By the early 2020s, when linear TV programmatic became mainstream, it landed in an environment where the identity layer the entire ecosystem depended on was fracturing under privacy regulation, platform fragmentation, and third-party cookie deprecation. The automation had scaled. The foundation had not stabilized.
The identity pivot
In 2018, I joined Civic Technologies following its $33 million ICO. Civic was building decentralized digital identity infrastructure, an approach that made explicit what the advertising industry had been treating as implicit for a decade.
Identity is not a data challenge. It is a permission architecture challenge. Which signals are legitimate? Which joins are consented to? Which inferences remain defensible as regulatory boundaries move?
The advertising industry spent two decades asking whether it could recognize the user. The right question was always whether it was allowed to. Those are not the same question. The difference between them is the infrastructure gap that accumulated over twenty years of prioritizing optimization speed over foundation.
Building a global partner ecosystem of organizations across financial services, healthcare, government, and consumer applications made one thing unmistakable. Every sector that depends on automated decision-making faces the same foundational challenge.
The advertising industry happened to encounter it first, at scale, in a commercially visible environment. The consequences were measured in campaign performance. In other sectors, they are measured differently.
The pattern completes itself
A direct experience with cybercrime documented in Privacy Pandemic shifted my focus permanently from building on top of the identity layer to building the identity layer itself. That work led to DFend, an identity-centric digital safety platform.
It validated what the advertising industry's experience had already suggested: the identity layer is not a feature you add to automation. It is the condition under which automation can safely operate.
That work now continues through The Digital Safety Brief and the Defend Foundation, examining the economics of digital fraud, the limits of enforcement, and the infrastructure required to address risks that individuals and institutions cannot manage alone.
The thread from programmatic to agentic AI to digital safety infrastructure is not a pivot. It is the same challenge, arriving at its natural scale.
2011: The machine is making thousands of audience decisions per second. Every decision assumes continuity. Nobody in the room can fully verify the assumption. We ship it anyway.
Wave three: Agentic AI on an unresolved foundation
Earlier this year, I reconnected with a longtime colleague from the early days of programmatic video. The conversation moved past the current buzzwords quickly, AI copilots, autonomous agents, machine-led planning, and back to the same unresolved dependency we had been circling for nearly two decades. Identity. Permissions. Continuity. Auditability.
In April 2026, Magnite announced autonomous buyer and seller agent capabilities, with Disney Advertising, Spectrum Reach, Kepler, and MiQ as early partners. The IAB Tech Lab is simultaneously building autonomous media protocols. The market is moving fast. The foundation is not settled.
Agentic AI is the third expression of the same ambition that drove real-time bidding (RTB), then programmatic, and then addressable TV: to remove more human friction from machine decisions. The interfaces are newer. The commercial instinct is identical.
But faster autonomy does not solve uncertain signals. It industrializes decisions made on uncertain signals. Autonomous systems require identity confidence, machine-readable permissions, reliable telemetry, interoperable execution pathways, and auditable feedback loops. Remove confidence in any one of those layers, and autonomy does not become smarter. It becomes faster at acting on unstable assumptions.
The stakes have changed
The advertising industry demonstrated this across two waves of automation. The cost was measured in wasted media spend, attribution challenges, and privacy enforcement actions. Those consequences were and remain real, contained within the advertising ecosystem, and absorbed primarily by brands, agencies, and platforms.
Agentic AI is moving into financial services, healthcare, government services, identity verification, and consumer platforms that directly shape access, trust, and economic participation. When autonomous systems operate in environments with fragmented identity layers, uncertain permissions, and unreliable signals, the issue is no longer advertising efficiency. It becomes a digital safety challenge.
In 2025, the FBI documented $20.9 billion in reported U.S. cybercrime losses, the measurable cost of what happens when the load-bearing beam breaks at the individual level. The World Economic Forum's April 2026 readiness framework makes the dependency explicit: agentic AI requires a verified identity layer, an auditable data exchange, and a transparent record of who authorized each action, what data was used, and what outcome was produced.
The United States currently lacks that coordinated foundational layer. Federal agencies use multiple non-interoperable identity systems. Payment rails cannot communicate across providers. Data exchange operates through fragmented sector-specific frameworks. The result is not a grid. It is a collection of private fire brigades, each protecting its own building, with no shared alarm system. Agentic AI is being built on top of that architecture right now.
What twenty years have taught me
The pattern I have watched repeat is consistent. Efficiency gains arrive first. The foundational dependencies on which those gains rest are treated as secondary. Infrastructure investment lags. The gap accumulates.
Each automation wave inherited the unresolved challenges of the prior wave and added new ones on a larger scale programmatic inherited fragmented cookies. Addressable TV inherited a fragmented identity. Agentic AI will inherit both, along with the privacy enforcement environment, regulatory fragmentation, and the complexity of the permission architecture that two decades of deferred investment have produced.
The companies and institutions that navigate this well will not be the ones with the most sophisticated AI interfaces or the fastest optimization loops. They will be the ones who treat identity confidence, machine-readable permissions, and auditable decisioning as prerequisites rather than features and build them into the foundational layer before racing to scale on top of it.
Programmatic spent twenty years teaching machines how to decide faster. Agentic AI will determine whether those machines can decide on a stable enough infrastructure to justify the autonomy we are giving them.
I have watched this race twice. The infrastructure gap does not close on its own. The question is whether we close it before it becomes the defining digital safety challenge of the next decade or after.
Read more from Christopher A. Smith
Christopher A. Smith, Author & Digital Safety Advocate
Christopher Smith is the award-winning author of Privacy Pandemic and the founder of DFend, a digital safety platform built to protect people from cybercrime. After being the target of a major cyberattack, he transformed his story of loss into one of purpose, turning a personal crisis into a global mission. His experience inspired him to develop technology that helps individuals safeguard their identity and privacy in the age of AI. Through his work and writing, Chris advocates for greater awareness, protection, and resilience online. He believes the future of digital safety is personal, because the threat already is.
Disclaimer:
This article reflects the author's personal observations and professional experience across two decades in advertising technology, identity infrastructure, and digital safety. It does not represent the views of any current or former employer or partner organization.
