Agent-Led Growth — Five Implementation Principles

Companion to overview.md. The "what" lives there; this is the "how."

Agent-Led Growth (ALG) is a product strategy that adds an agentic layer into the product to tie user intent to outcomes — accelerating acquisition and turning users into champions instantly.

PLG asks users to self-serve. ALG lets products self-operate.

These five principles are the operating manual. They are how a product team actually moves from "we have an AI feature" to "our product is agent-led."

1. Capture intent. Automate execution.

Decouple what from how. The user owns the goal; the agent owns the steps.

A traditional product surface forces the user to translate intent into a sequence of clicks, settings, and screens. ALG removes that translation step. The user expresses the outcome they want; the agent executes the steps required to reach it.

Concrete signal: if your onboarding includes the phrase "Click here, then here, then enter…" you are not yet agent-led. Replace the click sequence with a prompt that produces the same outcome.

2. Build generative experiences.

Stop asking users to start from a blank page. Generate the 80% baseline; let them edit the last 20%.

Empty-state screens are where activation goes to die. The agent should arrive with a draft — a populated form, a starter dashboard, a first-pass workflow — derived from what the user just said they wanted. Editing a draft is psychologically and operationally cheaper than producing one.

Concrete signal: count the empty states in your product. Each is a generative opportunity.

3. Embed knowledge. Engineer champions.

Train the agent on your product, your domain, and your best practices — so it executes the way your top users would.

Every product has a small set of expert users who get more value out of it than everyone else. They've internalized the conventions, the right defaults, the shortcuts. ALG turns those expert behaviors into agent behavior, then exposes them to every user. The agent makes every user a champion.

Concrete signal: if your power users have a Slack channel where they share workflows, that channel is the source material for agent training.

4. Mine conversational signals.

Clicks whisper; prompts yell.

Treat user prompts as the highest-fidelity demand signal you have. A click tells you a user pressed a button. A prompt tells you what they were actually trying to do. The roadmap is hiding in the unresolved prompts.

Concrete signal: review one week of agent prompts. The clusters that don't resolve are your next features. The clusters that resolve through awkward workarounds are your next fixes.

5. Design agentic UX.

Move beyond the chatbot. Voice, in-chat UI, screen assist, canvas — match the modality to the workflow.

A chatbot in the corner is not agentic UX. Real agentic UX includes:

• Chat for conversational intent
• Voice for hands-busy and accessibility contexts
• Tandem mode where the agent manipulates the live UI alongside the user
• Chatterblocks — interactive UI components (forms, cards, dashboards, wizards) rendered inline in the conversation
• Screen assist where the agent reads the user's current screen state and acts in context

The modality follows the workflow. A confirmation needs a card, not a paragraph. A complex multi-step task needs a wizard, not a long chat. A quick status check needs voice, not a screen.

Concrete signal: if "agent" in your product equals "chat box," you have not yet designed agentic UX.

Last updated: May 2026. Source: Foldspace Positioning Deck (May 2026), promoted into canonical 2026-05-19.