“Human-in-the-loop” quietly assumes the machine is driving and a person occasionally supervises. For knowledge work, that is the wrong way round.
Reversing the loop
Machine-in-the-loop keeps the professional in charge of framing the problem, choosing the criteria and owning the decision. GenAI is invited in for what it is genuinely good at: expanding options, stress-testing assumptions and compressing research time.
Three moves that work
- Widen before you narrow. Ask for eight credible options, not one answer. Divergence is where models earn their keep.
- Argue against yourself. Prompt the model to build the strongest case for the option you rejected.
- Force the evidence. Require sources, assumptions and confidence levels in the output format itself.
What this protects
Decision quality degrades when teams outsource thinking to a fluent generator. Keeping the machine inside a human loop preserves accountability while still cutting the drudgery — the research sweep, the first draft, the comparison table.
The result is not faster decisions for their own sake. It is better decisions, made with more options on the table and fewer hours lost to assembly work.