A Novel About Making AI Work
Every business is being told to bet on AI. This is the story of one that did, and of the five questions it had to answer to survive.
Every business is being told to bet on AI. Felix Canis made that bet years ago, when he signed off on an optimization that cut forty truck drivers. He never met one of them. It worked. That was the problem.
Now he runs Confluence Logistics out of the old Clark Bar factory on Pittsburgh's North Side, a network of two hundred freight companies built to carry the loads that can't be late. At 4:47 one morning its AI routes a NICU shipment through hubs that don't exist. Someone has been poisoning the data it learns from, and the only system left that works is a dispatcher's spiral notebook.
Felix bets the company on rebuilding it the other way, so the drivers and dispatchers who live with the AI's decisions can see what it decides and stop it when it's wrong. The attacks keep coming. Forged emails in his allies' names turn them against each other. The money runs short. And the deeper Felix digs, the less the attack looks like anything he knows how to fight.
Someone in the story writes each one down, usually after it has already cost them. None of them is about the model. Ask them about any AI system you buy, build, or work beside.
Who can see it?
Before an AI decision reaches a customer or a coworker, can the person closest to the work see what it decided, and why?
Who can stop it?
When it's wrong, who has the authority to stop it, and can they do it in time?
Who pays?
When the system makes a mistake, whose paycheck or budget absorbs the cost? Settle it before something breaks.
Does the record carry it?
When someone catches a near miss, is it written down where the next shift and the next model update will find it?
Whose dashboard saw zero?
When your numbers say a problem cost nothing, find out who quietly paid for it.
Anyone trying to make AI work in a business, in any industry. Freight is the setting, not the audience.
The owner deciding whether to trust a vendor's black box with the business.
The manager rolling out AI tools her team doesn't trust yet.
The worker whose judgment the system can't see.
The executive who has to sign for all of it.
No technical background needed: the book comes with a glossary for business readers and a Practitioner's Toolkit you can run with your own team.