What the machines took, and what they handed back

Every leadership team I have worked with this year has the same new problem, and none of them describe it as a problem. They describe it as progress. The strategy deck that used to take three weeks now takes an afternoon. The market scan that used to need an analyst now needs a prompt. The options paper that used to arrive with two choices now arrives with nine, each one plausible, each one supported, each one formatted beautifully.

Then the meeting happens, and nothing gets decided.

The bottleneck moved and nobody told the calendar

For about forty years, the expensive part of a decision was assembling the material. Finding the data, modelling the scenarios, writing the paper. Organisations built themselves around that cost. Steering committees, pre-reads, analyst teams, a month of lead time before anything reached a board.

That cost has collapsed. What has not collapsed is the cost of choosing, because choosing was never an information problem. Choosing is what happens when the information runs out and somebody has to accept a loss.

Nine good options is not nine times better than two. It is one decision, with more places to hide.

The organisations struggling most with AI right now are not the ones that adopted it too slowly. They are the ones that adopted it enthusiastically into a decision process built for scarcity, and are now drowning in supply.

What actually got harder

Three things, and none of them are technical.

Judging the material is harder. When a paper took three weeks, its weaknesses showed: you could see where the author had stretched, because stretching is laborious. Generated material is uniformly confident. It has no tells. The reader now has to supply the scepticism the production process used to supply for free.

Owning the recommendation is harder. Somebody used to write the paper, and that person carried the argument into the room and defended it. When the paper is assembled rather than argued, the authorship gets diffuse, and diffuse authorship is how decisions quietly become nobody’s.

Stopping is harder. If another option costs a morning, there is always an argument for one more option. Teams that used to be forced into a decision by the cost of analysis now have to choose to stop, which is a much rarer skill.

What became more valuable

The things that were always the differentiator and were previously buried under production work.

Naming the real question, when nine papers all answer slightly different ones. Setting the criteria before the options are ranked, because criteria written after the fact just ratify the option somebody already liked. Pricing the trade-off honestly, in the currency the organisation actually spends: attention, credibility, people who will leave if this goes the wrong way. And committing, out loud, with a date, so the decision stops being reopened every fortnight.

None of that is new. It is the part of the job that was always there and was always the least visible, because it took twenty minutes and the paper took three weeks.

What to do on Monday

Change one thing in your next options paper. Put the decision criteria on the first page, agreed and dated, before the options are listed. Then rank the options against them in the room.

You will find out very quickly whether your team has been deciding, or whether it has been selecting from whatever arrived. In my experience it is roughly half and half, and the half that discovers the truth about itself has a much better second half of the year.

The machines took the production. They handed back the part that was always the job.