The canon

The doctrine behind everything we sell.

The Lean Pyramid 6.0, Jidoka edition: our management model for companies where humans direct AI agents. The journey starts at the top, with purpose, and winds down through twenty-one practices until value flows at the base. Walk it.

Trust at the apexPeople get trusted. Agents get verified, and that verification protects the trust between people.

01 / 21 · Start with purpose

Strategy01 / 21

Human intent. Humans decide why the company exists and where it goes. No agent sets direction.

Start with purpose

Why before methods, methods before products.

People join causes, and a clear purpose is what holds a team through the hard times. In the agent era it works twice: a purpose written with clarity becomes the strongest instruction in an agent’s context, because agents amplify whatever we tell them. Change has layers, from tools down to mindset, and that is why a transformation starts here, never with tools.

Bring this into your company

Value at the base

Everything stands on value delivered to the customer. If value stops flowing, the pyramid has nothing to stand on.

Review and improve relentlessly closes the loop, and the cycle restarts.

Deliver value continuouslyCreate an organization safe to learnImprove relentlessly

The field manual

The whole canon, bound.

The book in progress turns these cards into a field manual, with the stories from the floor behind every practice. Leave your email and the first chapters find you before anyone else.

The lexicon

Sixteen words the model thinks in.

Tap a card to turn it.

The model · reading time 4 minutes

The Lean Pyramid, explained for executives

Samuel CrescencioFounder and Principal, Lean it 101

Every company now owns an uncomfortable question: agents can produce work around the clock, so who decides what they produce, and who answers for it? Most answers on the market amount to buying more tools. Ours is an operating model.

The Lean Pyramid organizes a technology company into three permeable layers. Strategy is human intent: people decide why the company exists and where it goes, and no agent sets direction. Management is human policy executed by the system: leaders define the policies of flow, capacity and quality, and the system runs them, measuring work items rather than people. Engineering is human craft with agent labor: engineers own the judgment and the standards, agents do the heavy lifting, and everything an agent produces gets verified before it counts.

Two words frame the whole structure. Value sits at the base, because everything above exists to deliver it continuously. Trust sits at the apex, because it is the scarcest output a system can produce and the first casualty of theater: inflated demos, dashboards that hide lateness, plans nobody believes. The sixth edition carries the name Jidoka, the Toyota principle of automation with a human touch, because deciding what to automate and what to keep in human hands is itself a human decision, taken deliberately.

In practice the model shows up as small, hard rules. Reversible decisions can be delegated, including to agents; irreversible ones a human takes, at the last responsible moment. Charts tell the truth, so a late release says it is late and shows its honest recovery range. Standards live in a vault that agents and humans both read, so the way of working survives any single person. None of this requires our platforms; all of it gets easier with them.

If your organization is deciding how much of itself to hand to agents, this model is the shape of our answer. The rest of this site shows it running.

See the model in operation

Get the next essay by email

New essays and recorded talks, sent when they are published. Written to be worth the inbox space.