The car stopped so fast the coffee left the cup in Calloway's hand. Half a second. No horn, no shout, nothing visible through the windscreen - and then the wheel eased itself back into traffic as if nothing had happened. Somewhere, a system had made a decision and unmade it inside a fraction of a second, without ever having to explain itself to anyone.
Calloway doesn't run the fleet. He runs Slipstick, a firm on the twenty-second floor of a Charlotte tower whose own AI has quietly become load-bearing - a recruitment platform nobody remembers approving, screening candidates on logic nobody can fully account for. He's spent his career trusting numbers offered with enough confidence, from sources with no reason to lie. The car in the opening pages is the first crack in that trust. What he finds when he turns that same scrutiny on his own building is worse.
No regulator has opened an inquiry. No client has walked. No journalist has called. Nothing has gone wrong yet - which is exactly the problem, because "wrong" has already happened, quietly, for a long time, in the form of a system nobody named, controlled, or owned. Calloway calls it a wild system rather than no system at all, and knows that naming a thing correctly and doing something about it are two different disciplines. He's only ever practiced the first.
What follows is the slow, contested work of building an AI management system inside a firm that didn't think it needed one: Reyes making the case for definitions nobody wants to sit still for; Whitfield and Priya keeping score on a league table neither of them will admit to reading; Voss watching a review go through the motions of scrutiny without any of the substance; Duffy signing off on things faster than anyone should be comfortable with. Every meeting, every dodged question, every "that's not really my department" is a clause of ISO 42001 playing out as office politics, before anyone in the room knows the standard's name.
Threshold is the fifth novel in The Collision Series, which teaches international management standards through story instead of textbook. This one takes on ISO 42001, the international standard for AI management systems, and asks the question every organisation deploying AI is one incident away from having to answer out loud: who is accountable for a decision nobody made, exactly, but everybody's system just made anyway?
You won't find a chapter that explains the standard. You'll find a firm finding out, the hard way, why it exists - and a way in for anyone who has ever sat in a room where a decision got made and unmade in half a second, by something that never had to explain itself, and has started to wonder whose job it is to ask it to.
Part of The Collision Series. Also available: Watershed, Exposure, Knife Edge, Collateral, and Turbulence.
Developed in collaboration with Claude (Anthropic). John Huckstepp is a Fellow of the International Institute of Risk and Safety Management and a Chartered Fellow of the Chartered Institute of Personnel and Development, and co-author of the internationally recognised risk management and leadership competence framework. That expertise, combined with AI-assisted narrative development, is what makes ISO 42001 legible through story without ever compromising the standard itself.
