What is AI-led growth (ALG)? When the buyer evaluating your product is an agent
Written by Arnon Shimoni
✓ Expert
Last updated on:
AI-led growth (ALG) describes a distribution motion where an AI agent, rather than a person browsing a website, discovers your product, evaluates whether it fits a task, and adopts it. The human is still the buyer. They are no longer the one doing the shortlisting.
It is the third motion after product-led growth and sales-led growth, and it inverts an assumption both of those rest on: that the evaluator is a human who can be persuaded. An agent cannot be persuaded. It can only be given something it can read, test and call.
Field | Detail |
|---|---|
What it is | AI agents discovering, evaluating and adopting software on a user's behalf |
Evaluator | A model or agent, not a human browsing |
Wins on | Machine-readable docs, a working free tier, an API that behaves predictably |
Loses on | Sales-gated pricing, docs behind a login, mandatory demo calls |
Pricing implication | Usage must be programmatically discoverable and entitlements enforceable in real time |
Related motions | PLG (self-serve human) and SLG (rep-led). See quote to cash |
How ALG differs from PLG and SLG
SLG | PLG | ALG | |
|---|---|---|---|
Who evaluates | A buying committee | An individual user | An agent acting for a user |
Discovery via | Outbound, referrals, analysts | Search, community, word of mouth | Model training data, docs indexes, tool registries, MCP servers |
What persuades | A rep and a business case | A fast time to first value | Structured documentation and a callable endpoint |
Blocked by | No budget | A signup wall | A login wall, ambiguous docs, unpriced endpoints |
Speed | Weeks to months | Minutes to days | Seconds |
Volume shape | Few large contracts | Many small accounts | Bursty, concentrated, machine-paced |
The last row is the one with billing consequences. An agent that decides to use your API does not ramp gently the way a human team does. It can go from zero to substantial volume within a single task, then stop entirely when the task completes.
What agents actually need from you
Documentation a model can parse. Structured, versioned, complete, and reachable without authentication. If your docs require a login, you are invisible to the evaluation.
Pricing that is published and machine-readable. An agent cannot request a quote. A pricing page saying contact sales removes you from consideration entirely.
A usable free or trial tier. Evaluation means calling the endpoint. If the first call requires a card and a sales conversation, the agent picks a competitor that answered.
Predictable error semantics. Agents retry. Ambiguous errors produce retry storms, which produce surprise bills and a support ticket.
Programmatic access to spend. An agent operating under a budget needs to check consumption and remaining allowance through an API, not a dashboard.
Real-time entitlements. The answer to "may this caller do this action right now" has to be available in milliseconds.
What ALG demands from pricing and billing
Three requirements follow directly from the motion, and none of them are optional.
Per-action metering. Value delivered to an agent is measured in actions completed rather than seats occupied. Seat-based pricing has no meaning when the user is a process. See usage metering.
Spend controls that hold. The failure mode of agentic consumption is a loop. Caps, alerts and hard stops need to be enforced by the billing system in real time, not reconciled at month end. This is a contractual expectation now, not a nice feature.
Idempotency. Agents retry aggressively and often at-least-once. Billing the same action twice because a retry succeeded after a timeout is the fastest way to lose an agent-driven account. See agentic billing.
Why this is happening now
Two things changed at once. Models became competent enough to select tools rather than just call ones they were told about, and the ecosystem standardised on protocols for exposing tools to them.
The result is that a category of purchasing decision that used to run through a human evaluation funnel now runs through a retrieval and selection step measured in seconds. The company whose documentation is complete, whose pricing is public, and whose free tier actually works gets selected. The one whose pricing page says contact sales is not in the consideration set at all.
None of this removes the human. Someone still owns the budget, signs the contract, and complains when the bill is wrong. ALG changes who does the shortlisting, and shortlisting is where most of the competitive outcome was always decided.
What to do about it
Unauthenticate your docs. If a model cannot read them, you are not in the running.
Publish real prices. Per unit, with the unit defined precisely enough that an agent can estimate cost before calling.
Make the free tier callable. No card, no demo, no waiting list, for at least enough volume to evaluate.
Expose consumption over the API. Current spend, remaining allowance, and the rate applied.
Enforce caps in real time. Both to protect your customer from their own loop and to protect your margin.
Make retries safe. Idempotency keys on anything that bills, with a defined deduplication window.
Where Solvimon fits
Solvimon meters agent actions in real time, enforces entitlements and spend caps at call time rather than at period close, and deduplicates retried events so a retry storm does not become a disputed invoice. Consumption and remaining allowance are readable over the API, which is what an agent operating under a budget needs.
Credit wallets, per-action rates and hybrid structures are native contract shapes, so the pricing an agent reads on your docs is the pricing the system enforces.
Frequently Asked Questions
What is AI-led growth?
AI-led growth is a distribution motion in which AI agents discover, evaluate and adopt software on behalf of a human user. The human still owns the budget and signs the contract, but the shortlisting is done by a model.
How is ALG different from PLG?
In product-led growth an individual human evaluates the product through a self-serve experience. In AI-led growth an agent does the evaluation, typically in seconds, based on documentation, published pricing and whether the API can actually be called.
Does AI-led growth replace sales?
No. It changes who builds the shortlist. Contracts, budgets and enterprise procurement still involve people, but the consideration set is increasingly assembled by a model before a human is involved.
Why does contact-sales pricing hurt in ALG?
An agent cannot request a quote or attend a demo. If the price is not published in a form the model can read, your product cannot be compared and is dropped from consideration.
What does ALG require from billing?
Per-action metering rather than seats, real-time entitlement and spend-cap enforcement, idempotent handling of retried calls, and programmatic access to current consumption and remaining allowance.
Why do agents cause billing disputes?
Agents retry aggressively, so the same logical action can arrive multiple times. Without idempotency and a deduplication window, retries become duplicate charges, and a runaway loop becomes a surprise invoice.
Is ALG the same as agentic billing?
No. ALG is the go-to-market motion. Agentic billing is the infrastructure that meters, prices and invoices what agents consume.
Related
Agentic billing. The infrastructure that meters and invoices agent activity.
AI agent pricing. Pricing models for agent-delivered value.
PLG billing. The self-serve motion ALG sits alongside.
Entitlements. Deciding in real time what a caller may do.
Quote to cash. How the motions converge after the evaluation.
Seats are a relic. Why per-user pricing stops working for agent workloads.
Ready for billing v2?
Solvimon is monetization infrastructure for companies that have outgrown billing v1. One system, entire lifecycle, built by the team that did this at Adyen.
