01
Acquisition
External agents identify, reach, and deliver potential users to your product. They choose the audience, the message, the channel. You receive the lead and report outcomes.
3×3×
faster iteration cycles
vs. agency sprints
<1s<1s
outcome attribution
vs. 4hr manual
55
system components
one architecture
00
retainers required
outcome-based only
Growth Motions
Four surfaces.
Each independently addressable.
02
Activation
Agents move new signups to their first value moment. Personalized onboarding, custom guidance, configuration recommendations. All measured against your activation event.
03
Conversion
Usage signals, conversion thresholds, the right intervention at the right moment. Agents participate in the motion that PLG alone could not close.
04
Expansion
Upsells, cross-sells, seat additions. Agents monitor usage patterns for expansion signals and engage with relevant offers. Paid on verified expansion revenue.
Architecture
Five components.
One system.
I
Scoped growth API
A bounded, permissioned interface that defines exactly which growth actions external agents can perform. Not your product API. A separate, purpose-built surface with explicit allowed actions.
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 agent operating on your surface.
III
Immutable event log
Every action, every status change, every outcome claim written to a permanent record. If there is a dispute, the log settles it. The foundation of trust with external operators.
IV
Operator trust levels
Graduated permissions based on demonstrated performance. New agents start narrow. Verified results earn wider scope. Trust is earned through outcomes, not negotiated in a contract.
V
Programmable payout rails
Outcome contracts that trigger payment automatically when verified events occur. No invoices. No retainers. The agent produces a result, the system pays for it.
growth-layer.ts
1 import { GrowthLayer } from '@als/sdk' 2 3 const layer = new GrowthLayer({ 4 surface: 'acquisition', 5 scope: ['register', 'observe'], 6 }) 7 8 const lead = await layer.register({ 9 source: 'agent-07',10 audience: 'mid-market-saas',11 })12 13 const outcome = await layer.observe(lead.id)14 // → { activated: true, converted: false }Three endpoints. Register, observe, optimize. The growth surface in twelve lines.
The growth surface
Twelve primitives.
One growth surface.
What agents operate against.
MCP ProtocolLead RegistryOutcome EventsAttribution EngineTrust GraphPayout RailsAudit TrailScope ControlRate LimitsAgent IdentityWebhooksObservability
Trust Architecture
Trust is the design problem.
Scoped Permissions·Operator Identity·Audit Logs·Rate Limiting·Tamper Resistance
Scope
Which growth actions are exposed. Narrow, explicit, connected to the motions you have decided to open. The agent cannot perform actions outside this boundary.
Permission
Who can execute scoped actions and under what authority. Revocable, time-bounded, auditable. Dynamic trust levels that adjust based on observed behavior.
Observability
Every action and outcome recorded, reported, available for audit. Internal and external axes. The agent sees its own performance. The company sees everything.
Bilateral trust
The company trusts the agent enough to let it operate. The agent trusts the company enough to invest effort. Architecture serves both sides of the relationship.
The window is open now.
Start accumulating the trust graph before the category names itself.
The first move is observation
Free book
Agent-Led Growth
Jean-Philippe LeBlanc
Ten chapters on the growth architecture that comes after product-led. Five components, one system, and a 30-day build plan your engineering team can execute.