Why AI Pilots Stall — and What Media Organizations Are Actually Getting Wrong

You've probably heard the stat: 95% of enterprise AI pilots fail. Even if the real number is closer to 80%, the pattern is consistent enough to take seriously. Companies launch initiatives with urgency, spin up copilots, run pilots across teams and then, somewhere between the kickoff deck and the quarterly review, momentum quietly dies.

The most common explanation is change management, or insufficient training, or unclear ROI. Those aren't wrong. But there's a more fundamental issue hiding underneath all of them: a lot of organizations are using AI to solve problems they don't actually have.

What Media Leaders Actually Said

We asked media professionals across roles — CEOs, CMOs, programmatic leads, analysts, strategists, client managers — a deliberately simple set of questions: What's your primary role? What takes the most time in your day-to-day work? What tools are you using, and why is it hard?

The answers pointed somewhere unexpected. The most time-consuming work in media isn't strategic thinking. It's operational stitching — the connective tissue work that happens between systems, between teams, and between data sources that were never designed to talk to each other.

Three themes dominated across every role we surveyed: reporting across multiple platforms, billing and reconciliation, and data aggregation and validation. The friction, in other words, isn't at the edge of the workflow. It's squarely in the middle of it.

The Reporting Grind

Reporting was the single most frequently cited time sink. Teams described pulling data from multiple sources — often simultaneously — then manually re-entering metrics, reconciling discrepancies, and eventually transforming raw numbers into something a client would actually want to read. When your data lives across five systems, reporting becomes an exercise in aggregation long before it becomes an exercise in insight.

Data Abundance, Coherence Deficit

Media leaders and strategists consistently cited complexity as their primary barrier, and the specifics were telling. One respondent described the challenge of finding the right data point "in a haystack." Others pointed to variance across partners and inconsistent definitions of the same metrics. Teams are running sophisticated toolsets and still manually connecting outputs across all of them. The problem isn't access to information. It's that no one has agreed on what the information means.

The Deployment Gap

To be clear: most of the people we surveyed are already using AI tools. ChatGPT, Copilot, Gemini — these have become standard fixtures. And teams are getting real value from them, particularly for writing communications, reviewing transcripts, drafting presentations, and supporting research. 

But there's a significant gap between where AI is being deployed and where the actual drag lives. AI is helping individuals move faster on discrete tasks. It is not resolving the structural fragmentation underneath the workflows those individuals operate inside. If your core problems are cross-platform reporting, inconsistent data normalization, billing reconciliation, and governance coordination, then a better writing assistant isn't going to move the needle in any meaningful way.

Aiming at the Right Problem

This is why AI pilots struggle to scale. Organizations that treat AI as a productivity enhancer at the surface layer — drafting copy faster, summarizing notes more quickly — while leaving underlying workflow fragmentation untouched will always find the return feels underwhelming. 

The real opportunity for AI in media is reducing operational drag: automatically aggregating and normalizing reporting data, creating consistent definitions across partners, streamlining billing workflows, reducing manual validation loops, turning governance from spreadsheet management into something resembling structured intelligence. These aren't glamorous applications. They're also not incremental ones.

Infrastructure Is the Strategy

Media organizations aren't short on experimentation, and they're not short on tools. They're short on integration. The companies that win the next phase of AI adoption won't simply add more copilots to more workflows. They'll fundamentally redesign how data, reporting, billing, and strategy connect across the organization — and they'll deploy AI at the center of that architecture, not the periphery.

Want to see how Akkio is building the infrastructure that makes this possible? Check out our recent webinar on the AI Connection Layer, featuring leaders from IPG, Mediaplus, and Digitas.

‍