The Orchestrator

The chapters that follow are about operating with evidence, and none of them works unless you already believe that failure is normal enough to design for.

There is a reckoning approaching for the companies that adopted ahead of their comprehension, and its opening movements are already public. S&P Global’s 2025 enterprise survey found that 42 percent of companies had abandoned most of their AI initiatives that year, up from 17 percent the year before — the first broad, measured retreat of the adoption wave.¹² The layoffs announced with AI pride are being quietly reversed: HR researchers now find that most organizations that conducted AI-led layoffs have rehired at least a quarter of the roles they cut, and more than half of those rehires happened within six months, while Forrester forecasts that half of AI-attributed layoffs will end up undone.¹³ Even the chief executives who forecast the replacement economy are revising it — Sam Altman now says he was “pretty wrong” about the pace, and Dario Amodei’s forecast that AI could eliminate half of white-collar work has softened into a claim about work expanding.¹⁴ Every company that adopted on a promise is starting to run the actual numbers, and the numbers do not say the technology was a fraud. They say the announcement described a system nobody had built — the judgment layer, the escalation paths, and the review capacity never made it past the slide.

The reckoning is not an argument for cynicism, and this is the part I want to get exactly right. Cynicism is not analysis; it is hype with the sign flipped, and it will be exactly as wrong, in the other direction, on the same schedule. And the lesson is not that everyone should have waited. It pays to adopt ahead of time — the companies that did, inside boundaries they could name, with evaluation they could show, are not backtracking; they are compounding. What does not pay is premature overadoption, and those are the organizations now doing the walk-backs. The technology was never the hard part. The gates were.

HQ 6 — Assembled. The human and AI each wrote portions of this chapter. I assembled, reviewed, and take responsibility for the whole; the voice and arguments are mine, and I know which parts are which.