The thesis

Agent-Led
Growth

The growth architecture after product-led. Four motions, five components, one compounding flywheel that changes the economics of how software grows.

by Jean-Philippe LeBlanc

The problem

The closed growth model
has hit a ceiling.

Product-Led Growth solved the top of the funnel: self-serve sign-up, frictionless onboarding, usage as the qualifying signal. But it created its own constraints. The motions still depend on humans clicking through flows that were designed for humans. The growth surface is bounded by the product's own UI.

Agent-Led Growth opens the growth surface to external autonomous agents operating against a trusted growth layer. Agents are measured on verified outcomes, not impressions or intent signals. They are paid on results, not retainers. The operator doesn't manage the agent. They define the contract and observe the outcomes.

Growth Motions

Four surfaces.
Each independently addressable.

Diagram of the four Agent-Led Growth motions: acquisition, activation, conversion, and expansion
01

Acquisition

Agents find and register qualified leads against the growth layer. The surface replaces the landing page; the agent replaces the ad.

02

Activation

Agents guide registered leads through first-value milestones, triggering activation events that the growth layer can verify.

03

Conversion

Agents move activated leads toward revenue events: upgrades, purchases, commitments. Verified against outcome contracts.

04

Expansion

Agents identify and execute expansion opportunities within existing accounts, measured on incremental verified revenue.

Architecture

Five components.
One system.

Diagram of the five system components: scoped growth API, lead lifecycle state machine, immutable event log, operator trust levels, and programmable payouts
I

Scoped Growth API

The bounded interface where agents connect. Explicit allowed actions, rate limits tied to trust levels, response schemas that return only what the agent needs. Not a public API. A surgical surface.

II

Lead Lifecycle State Machine

Every lead occupies a defined state. Transitions are explicit, logged, and governed by rules. The shared language between you and every external operator.

III

Immutable Event Log

Every action produces an event. Every event is recorded in a log that cannot be altered. The source of truth. If there is a dispute, the log settles it.

IV

Operator Trust Levels

Graduated permissions based on demonstrated performance. New agents start narrow. Verified results earn wider scope, more volume, more action types.

V

Programmable Payouts

Outcome contracts that specify what counts, how it is verified, and what payment triggers on verification. No invoices. No retainers. Results move money.

The endgame

The compounding
flywheel.

Each cycle tightens the feedback, reduces the cost per verified outcome, and widens the aperture for what agents can do. The loop does not stop.

01

Agents produce outcomes

02

Outcomes generate data

03

Data improves the surface

04

Better surfaces attract better agents

The loop compounds

The window is open.

Start accumulating the trust graph before the category names itself. The sandbox is live.

Experimental. Early. Open.