The Meeting That Should Have Been a Decision

I was in a meeting last month — two hours, eight people, three teams — to decide which AI model to use for an internal tool. The conversation covered benchmarks, pricing, compliance implications, integration complexity, and vendor relationships.

At the end of the meeting, someone said: “Let’s schedule a follow-up to align on next steps.”

Nobody made a decision. Nobody was authorized to. The meeting existed to discuss, not to decide. And everyone knew it walking in, but nobody said it.

This happens constantly in enterprise AI projects, and it’s not unique to AI. But AI makes it worse for a specific reason: the technology is moving fast enough that the information discussed in the meeting is stale by the follow-up. The model comparison from last Tuesday? A new version dropped Thursday. The pricing analysis? The vendor changed tiers. The compliance guidance? Legal is still reviewing.

The meeting cadence can’t keep up with the decision pace. So teams do one of two things:

They delay until the decision makes itself. The project stalls long enough that the urgency passes, the budget gets reallocated, or a senior leader just picks something in frustration. The “decision” is actually a default.

They over-index on reversibility. “Let’s just try it and we can always switch.” This is sometimes fine. But for AI systems that touch customer data, require compliance review, or integrate with production infrastructure, switching is never as cheap as it sounds. The decision to try something becomes the decision to use it permanently.

What I’ve learned to do: ask at the start of every meeting, “Who in this room can make this decision today?” If the answer is nobody, the meeting needs a different guest list, not a longer agenda.

One authorized decision-maker and thirty minutes will accomplish more than eight stakeholders and two hours. Every time.