AI Research

Twenty-two people said the same thing in different words: nobody taught them how to work with AI. The study set out to map that gap — what separates collaboration from use, and what the missing instruction costs in trust, time, and confidence. It carried one honest complication: the research was itself conducted with a machine, and hiding that would have failed its own thesis.

The instrument studies itself

A hundred twenty-seven data points, coded by type rather than by person, converged into seven themes — each verified against raw transcripts before it earned the name. Where recruiting left gaps, synthetic personas filled them, labeled as what they were rather than laundered into the sample. Midway through synthesis, the researcher caught himself re-running a prompt for the fourth time, trusting it less with each pass — and coded the moment pink, for emotion.

What changed

The findings outlived the coursework. Trust, time, and skill turned out to be one system, not three problems — the conclusion that seeded Cue and became the working premise of this practice: augmentation, not automation.

Two of these threads didn’t stay loose. Iteration Death Spiral and Learning and Guidance Needs became the starting brief for Cue — the trainer this study made necessary.