Writing
Managing AI that does the work
A series on moving people from doing the work to governing it. Organizations are adopting AI faster than they can govern it.
The chapters
One question per chapter.
-
Chapter 1: Every organization has an AI policy. Almost none has a way to make people follow it.
For CEO or managing partner of a mid-sized firm
Most firms now have a written AI policy, but the work still happens in dozens of private chat windows that nobody reviews or records. A policy becomes real only when there is one place where AI work is defined, approved and recorded. Three steps a leader can take before buying anything.
-
Chapter 2: Authority must remain explicit. How AI can act in your operations without acting on its own authority.
For Compliance officers, risk officers and heads of IT in regulated mid-sized firms
A model that understands a request has not been given permission to act on it, and most AI deployments in operations quietly treat the two as the same thing. AI work is defensible only when authority is granted, bounded and recorded, when a stop counts as a valid outcome, and when every action traces to a version, a run and a named approver. Three steps to take before buying anything.
-
Chapter 3: Pay for AI when the work is designed, not every time it runs
For CFO and CTO of a mid-sized firm, reading together
Most automation built on language models calls a model on every record, so the bill grows with volume and the output cannot be reproduced. Calling the model when the work is designed, and again only when a source changes shape, leaves a deterministic process in between that costs less to run and can be proven afterwards. Three steps to take before buying anything.
-
Chapter 4: The cost of repeated translation
For COO or CFO of a mid-sized firm in a regulated or paper-heavy sector
Every hand-off between people, applications and companies is a translation that costs hours and loses meaning, and most firms pay for it without ever seeing it on a ledger. The fix is to keep business meaning in definitions that people and software both read, so software carries the coordination while people keep the authority. Three steps to measure the cost before buying anything.
-
Chapter 5: Typed, checked AI calls for .NET
For .NET architects and senior developers who build and maintain business applications
Calling a model from C# takes an afternoon; keeping that call trustworthy inside a business pipeline is the hard part. Give every call a contract: the goal, the shape and the rules in versioned files, every answer checked against them, retries that carry the findings, every attempt kept, and wording that changes with a save instead of a release. Three steps a development team can take before buying anything.
-
Chapter 6: A Tuesday, before and after
For Operations leaders
One recurring task, a government portal filing, followed through an ordinary Tuesday twice: once as it runs today, by hand, and once with the work carried by software that stops for a person before anything that matters. What changes for the person doing the work, the supervisor who owns it and the auditor who asks about it, and what stays exactly as it was.
-
Chapter 7: Websites other software can use
For Founders and product leaders
More and more of the visitors to a company's website are software acting for someone else. Today that software reads pages written for people and guesses. The direction we expect is a website with two faces, pages for people and a governed service interface for software, both describing the same capabilities. This is a direction, not a product; the first step any company can take is to publish a machine-readable description of what it offers and how a request is made.