01 / Central thesis
The workflow is the structure now.
You do not start with roles and fill them. You start with workflows and decide which can run on AI, which need a human-in-the-loop, and which still belong to a human end-to-end.
Field notes
The frameworks in one place.
The book moves quickly. These are the operating models to keep open while you audit the business.
02 / Workflow modes
Three answers for every workflow.
- Full AI: the agent runs the workflow end-to-end.
- Human sherpa: the agent does the work; a human checks and ships.
- Human executor: judgement, taste, relationships and accountability remain human.
03 / Failure modes
Two ways AI adoption fails.
Too little leverage: a tool is bolted onto the org chart. Nobody uses it. Nothing changes.
Leverage you do not own: one employee becomes a 10x version of themselves with private workflows. The day they leave, the leverage leaves with them.
04 / Agent Anxiety
The bottleneck moves to executive judgement.
Once execution is no longer scarce, judgement becomes the constraint. Taste, prioritisation and decision volume become the work of the leader.
05 / SaaS compression
The middle gets squeezed.
Small and medium businesses can now build operating infrastructure that used to belong only to the largest companies. Hyper-niche tools and enterprise trust survive; generic middle software gets exposed.
Closing line
AI does not fix a bloated org chart. It removes the need for one.
Audit prompt
Run this against your next hire.
RoleWhat hire are you about to approve?
WorkflowWhat recurring work is the role meant to cover?
ModeFull AI, human sherpa, or human executor?
First systemWhat could ship in 30 days?
Need the outside lens?