Yesterday's PP&M seminar left a whiteboard covered in two columns, and the second column was the uncomfortable one. The first listed what should decide who gets assigned to a project — personality fit, availability, capability, relationships, aptitude, soft skills, motivation, diversity. The second, filled in by the room admitting the truth out loud, listed what actually decides it: whoever happens to be free, whoever is loudest, whoever the sponsor already trusts. Nobody enjoyed writing that second column. That discomfort turned out to be the point.
Only once the room had argued its way through that gap — no shortcuts, no tool, just people disagreeing with each other about what's real in their own organizations — did the facilitator open an AI assistant and feed it everything the whiteboard had produced. What came back wasn't a new answer invented by the AI. It was the room's own messy, half-formed conclusions organized into a clean before-and-after comparison, and a first draft of a "next-generation" work description with fields the current template simply doesn't have: a line for communication style, a line for the kind of teammate this role needs to pair well with, a line for the career motivation a good fit for this role should already carry. Every one of those fields traced back to something someone in the room had said twenty minutes earlier. The AI hadn't diagnosed anything. It had taken a diagnosis we'd already reached the hard way and given it a shape that could survive being handed to someone else.
That ordering is the part I keep replaying. It would have been faster to skip the argument and ask the AI directly: what should a good work description include? It would have produced something plausible, even well-organized, and it would have been almost useless, because it wouldn't know that in this particular office, the quiet expert keeps losing assignments to the loud mediocre one, or that the actual friction point is a manager who reads "available" as "capable." Those specifics only exist because a room full of people were willing to sit in the discomfort of naming them to each other first. The AI's value showed up entirely in the second step — structuring, formatting, making an idea portable — and depended completely on the first step having already happened without it.
I do the reverse constantly in my own consulting work, and yesterday made me notice it. A client hands me a mess, and before I've let them sit in the discomfort of naming what's actually wrong, I'm already halfway to a framework, a template, a deck. The instinct isn't laziness so much as a kind of politeness — nobody enjoys watching a room struggle, so you reach for the tool that promises to shorten the struggle. But the struggle is where the real material comes from. Skip it, and the tool, however good, is just organizing air.
The broader lesson isn't about AI at all, not really. It's that the order of operations is the discipline, and the order that actually works is: people argue out the true, uncomfortable version of the problem among themselves first, and only then hand the result to a tool to give it durable shape. Reverse that order and you get something articulate and wrong. Keep it, and even a rough whiteboard turns into something you can actually use on Monday.
No comments:
Post a Comment