This year, we found ourselves in more conversations than ever: across offices and Zoom calls, in Slack channels and AI chat windows, and on stage at CES, Cannes, and AdWeek NYC.
The through-line? AI’s collision with the media industry. Not just as a tool, but as a fundamental shift in how the industry creates, buys, and measures advertising.
Some of these conversations happened in public, on panels, in studios, at roundtables with agency leaders.
Others happened internally, as we stress-tested AI across Infillion. All of them challenged assumptions we didn’t know we were making.
Here are five conversations that changed how we think about media in 2025.
1. Why Composability Wins In The AI Era
Advertisers are hemorrhaging budget on bloated tech stacks. They license tools they don’t use, stack platforms that don’t integrate, and pay overhead for features that sit dormant.
Infillion CTO Simon Asselin reframed the solution to Beet.TV: “Think Formula 1 race car, not factory sedan. You assemble the best parts from multiple vendors to create a custom machine optimized for performance. There’s no one-size-fits-all model that works. You’re not stuck in a rigid system—you can mix and match. That’s what composability means, and that’s the platform we’re building for advertisers.”
Traditional stacks lock you into rigid systems. Composable platforms let you mix and match—combining third-party solutions with proprietary tech under one roof, activating only what drives performance.
The most forward-thinking players aren’t trying to own every piece of the ecosystem. That’s why for 2026, we are building systems that connect to the right ones, faster.
2. The Infrastructure Gap Between Programmatic and Agentic Buying
AI has been reshaping digital advertising through predictive algorithms and machine learning for years. But the next wave of AI tools like agentic buyers and generative AI will test infrastructure in new ways.
Unlike programmatic systems that execute predefined rules, agentic systems make real-time decisions and adapt strategies mid-campaign. That requires infrastructure built for dynamic decisioning at scale.
Many retail media networks launched before their infrastructure was truly ready, racing to capture budgets. Now they’re backfilling foundations. AI agents will expose those cracks faster.
The shift from programmatic to agentic buying is operational. Retail media networks built for batch processing and static rules will struggle.
Success won’t hinge on who adopts AI first. It will belong to those who build systems robust enough to make it work.
3. What We Learned Stress-Testing AI Across Every Department
The AI implementation crisis is real: MIT Media Lab reported that 95% of organizations see no measurable ROI on AI investments, and a Harvard Business Review survey reported that 40% of employees receive AI “workslop” that lacked substance and created additional headaches.
Rather than adopt AI indiscriminately, Infillion launched an internal Tiger Team to systematically evaluate where AI adds genuine value versus where it creates noise.
The scope: sales automation, creative workflows, operational efficiencies, productivity tools.
The mandate: prove utility before scale.
“Most of the time when an AI project struggles it’s not actually because the model isn’t capable of doing the task at hand, but rather the task does not have strong documentation, clean data, or a clear scope,” said Mason Matchinski, Senior Researcher and Tiger Team lead at Infillion.
“Once connected to company data sources and documentation you can access an unprecedented amount of information at near instant speeds. This has shaved a significant amount of time off of countless project timelines.”
That infrastructure-first approach enabled the Tiger Team to deliver practical results across the entire organization.
“Using powerful programming tools like Claude Code we have been able to design, iterate, and productionalize the automation of entire operational processes throughout the company,” Matchinski said.
The AI infrastructure Infillion is building delivers measurable results for its employees and clients – automating processes, shortening creative production timelines, and eliminating operational bottlenecks.
And it’s only just getting started.
“The rate of change in the AI world continues to shock me after each release,” Matchinski said. “When provided with the proper business context, documentation, and infrastructure, flagship models can do an amazing amount of real, valuable work.”
4. How AI Levels the Playing Field for Independent Agencies
In July, ten independent agency leaders came to our headquarters for a wide-ranging discussion. Within minutes, one topic dominated: AI’s role in reshaping how agencies work, pitch, and profit.
The timing matters. As holding companies consolidate—Omnicom’s acquisition of IPG being the latest example—independents face a critical moment. They can’t compete on scale or bundled services. But they can compete on agility, specialization, and intelligent use of technology.
What we discovered AI unlocks for smaller agencies:
- Teams of five execute campaigns that previously required agency-of-record scale
- Custom models replace expensive proprietary tools locked behind holdco walls
- Real-time optimization happens without armies of analysts
Infillion’s platform is built for orchestration, not obligation—API-first, transparent, designed to integrate into existing workflows. We pair explainable automation with unified measurement that links attention, engagement, and conversion.
Independence and sophistication should never be a tradeoff.
5. The Attention Economy Has a Bot Problem
According to new research from DoubleVerify, AI bots accounted for up to 15% of all ad clicks in tests of “unprotected media.” In some cases, an ad click was 10X more likely to come from an AI bot than a human.
The measurement problem masks the real issue: who can you trust?
We’re asking clients a different question now: What if the only engagement that counted was engagement that required being human?
TrueX’s opt-in value exchange model creates a verification layer for brands and users, keeping the bots away . As of August 2025, HUMAN Security found a 0.13% bot rate across that keeps the bots at bay even in 2025.
Viewers actively choose to interact with an ad in exchange for premium content—uninterrupted streaming, exclusive access, ad-free viewing. That choice provides one of the best advertising experiences available for both the client and the consumer.
The attention economy we’ve built—where clicks, impressions, and engagement drive billions in ad spend—is increasingly contaminated. Not crude bots, but sophisticated AI agents scraping and interacting with content in ways that look human.
Advertising Week 2025 confirmed: attention is the next currency. But only if we can verify it’s human.
What These Conversations Showed Us
Each confronted the same question: How do we build systems—technical, operational, creative—intelligent enough to leverage AI’s power but grounded enough to maintain trust?
The answer we have come to? Composability over Frankenstacks. Verifiable attention over bot-inflated metrics. Infrastructure that enables agility rather than constraining it.
The conversations that changed how we think about media in 2025 weren’t about AI replacing human judgment. They were about building frameworks where AI amplifies it.
