You have used ChatGPT or Claude for a year. You have hit the quality ceiling. The work has not changed evenly across your team — four jobs flipped first, three are still mostly human, and the order has implications.
What changes first
Four jobs LLMs reshape before anything else:
- Brief-to-draft cycle time on long-form content. The brief that took two days now takes two hours, and the draft that took a week now takes two days. Your binding constraint moves from time-in-seat to the quality of the brief feeding the model.
- Variant production for paid and lifecycle. Five paid-search variants becomes fifty; one lifecycle email becomes a personalised flow. The constraint moves to the discrimination function — how you pick which variants to ship.
- Distribution copy compression. A long-form post into a thread, a thread into a LinkedIn post, a LinkedIn post into an internal Slack message. Compression with voice preservation is something LLMs do well; it lets you ship more from each long-form piece.
- Audience research synthesis. Customer interviews into themes, support tickets into pain-point clusters, competitor pages into positioning maps. The work that used to land on a strategist’s desk for a week now lands on it for an afternoon.
What changes last
Three jobs that are still mostly human, and that will be for a while:
- Setting the strategic frame. What the company is actually trying to do this year. A model with no theory of the firm cannot answer this; you can.
- Editorial taste. Not “is this grammatical” — “is this the voice we ship in.” Taste is bounded; LLMs do not generalise across the bound without your hand on the dial.
- The measurement framework. What gets counted, what counts as a win. The model can run the numbers; the framework is yours.
Why the order matters
If you change the four “first” jobs without changing the three “last” jobs, you ship faster but your strategy, voice, and measurement drift. If you try to change the “last” jobs first, you spend a year on AI strategy decks and ship nothing.
The Playbook orders the chapters to match — Chapters 1 and 2 set the worldview and the measurement framework, then Chapters 3 through 14 work through the four “first” jobs and their adjacent jobs in order of how much they change. Chapter 15 closes on governance because governance is what lets you keep the gains without losing the trust.
If you want the long form, the sample chapter is free in your browser, and the full Playbook is the rest of the work.