I knew the project was going to fail before I ever walked into the room.
About ten years ago I watched a big data implementation unfold at a compliance software company. For weeks, the core team met behind conference room glass: the BI team lead, an expensive consultant, and a whiteboard that kept getting more complicated.
The rest of the BI team sat along the wall. Eyes wide. Shifting in their seats.
Not one subject matter expert was ever invited in to talk about what the data actually meant — or how it should be structured to be useful.
The people who knew the most were on the other side of the glass.
That project did not go well.
Fast forward ten years. S&P Global just reported that 42% of companies abandoned most of their AI initiatives in 2025. That's up from 17% the year before.
The coverage blames data quality. Costs. Technical complexity.
Those things are true. But they're not the whole story.
The data wasn't ready because the people who understood it were never in the room. The AI didn't know what it didn't know — because the people closest to the customers, the workflows, the edge cases, and the exceptions were watching from the hallway.
This isn't a technology problem. It's the same organizational problem we've been solving badly for decades, now running at AI speed and AI scale.
The conference room glass is still there. The question is who you're leaving on the other side of it.
Who in your organization should be in the room for your AI work — and isn't?