The definitive guide · v2026.1

Agent-Led Growth.

The buyer's journey didn't change — the actor did. An AI agent now does the research, evaluation, and trial on the buyer's behalf. The human stays in the loop to approve. The complete playbook for getting picked by that agent.

100K→8M+
MCP server downloads in 6 months
97M / mo
Agent SDK downloads (Dec 2025)
57%
of B2B already using AI sales agents
63% vs 7%
Resend vs SendGrid in Claude Code

TL;DR

  • Agent-Led Growth (ALG) is the GTM motion where AI agents — Claude, ChatGPT, Perplexity, Cursor, Operator — do the research, evaluation, and trial on the buyer's behalf. The human stays in the loop to set the goal and approve; they no longer do the work.
  • This is not 'AI in sales.' That's supply-side — agents helping you sell. ALG is demand-side — agents doing the buying. Confusing the two is the most common strategy mistake.
  • The infrastructure is in production: MCP went 100K → 8M+ downloads in 6 months; Chrome 146 ships WebMCP; 57% of B2B uses AI sales agents. Most of your top-of-funnel traffic in 2026 will not be human.
  • ALG360's framework: Introduce → Evaluate → Execute — one pillar for each job the agent now does on the buyer's behalf. Get named on the shortlist. Win the agent's evaluation. Let the agent transact with you, not around you.
  • Early movers report 4–7× more conversions and up to 70% lower CAC. Recommendation moats compound with every model generation — by the time they're obvious, they're locked in.

Ch. 01Definition

What ALG is — and what it isn't.

Agent-Led Growth is the GTM operating model for a buying journey where the actor changed, not the journey itself. The human used to do the research, read the docs, build the feature matrix, and run the trial. Now an AI agent does all of that on the buyer's behalf. The human stays in the loop to set the goal and approve the outcome — but is no longer the one doing the work.

That changes who is on the other side of your site. When a buyer asks Claude "what's the best X for Y," the human never opens your homepage. The agent reads your docs, scans your pricing page, compares you against three competitors, and hands the human a recommendation. You won or lost before the buyer was ever on your site.

ALG is not "AI in sales." That's supply-side — agents helping you sell more efficiently (AI SDRs, content engines, pipeline automation). ALG is demand-side — agents doing the buying on the buyer's behalf. Supply-side ALG improves the economics of your current funnel. Demand-side ALG changes whose funnel it is. Confusing the two is the most common strategy mistake.

Approach
Who decides
Who executes
Human's actual role
AI-assisted GTM
Human
Human (with AI help)
Drafts emails with ChatGPT, asks AI to summarize a call.
AI-augmented GTM
Human
Shared
Sets up automations; reviews AI output before sending.
Agent-Led Growth
Human sets goal & approves
AI agent (on buyer's behalf)
Approves the agent's shortlist, evaluation, and trial. Does not do the work.

In the first two rows the human is the bottleneck. In ALG the human is the governor — they set the objective, define the guardrails, and approve the outcome. The agent does the research, evaluation, and trial. Crucially, an agent on the buyer's side is now the one comparing you to your competitors — and increasingly the one initiating the transaction.

Ch. 02Inflection

Why now: the infrastructure is production-ready.

Every dominant GTM motion has been unlocked by a wave of enabling infrastructure. CRMs made sales-led growth scalable. Product analytics made product-led growth measurable. Intent data made account-based marketing targetable. ALG is the motion unlocked by agent infrastructure: MCP, A2A, frontier-model retrieval, browser-side agent runtimes, and WebMCP.

MCP adoption curveMonthly downloads (log scale, millions)
0M1M10M100MNov '24Jan '25Feb '25Mar '25Apr '25Aug '25Dec '25
100K
MCP downloads, Nov 2024
8M+
MCP downloads, Apr 2025
97M / mo
Agent SDK downloads, Dec 2025
57%
of B2B orgs using AI sales agents
62% / 23%
experimenting / scaling agents (McKinsey)
Chrome 146
ships WebMCP (Feb 2026)

When infrastructure matures this fast, the motion it enables isn't theoretical. It's already redistributing pipeline. The gap between 62% experimenting and 23% scaling is exactly where the competitive advantage compounds — and it closes fast.

Ch. 03History

The evolution of GTM motions.

Motions don't replace each other — they layer. But each era has a dominant new motion that reshapes how the best companies grow. Each transition followed the same pattern: new infrastructure emerges, early adopters build novel workflows, the workflows become repeatable enough to name, the name becomes a category. ALG is at stage three.

2000s
Sales-Led
Enabled by
CRM (Salesforce)
North-star metric
Quota attainment
What changed
Scalable sales team management
2012+
Product-Led
Enabled by
Mixpanel, Amplitude
North-star metric
Activation rate
What changed
Product as acquisition channel
2018+
Account-Based
Enabled by
6sense, Bombora
North-star metric
Engagement score
What changed
Precision targeting of buying committees
2025+
Agent-Led
Enabled by
MCP, A2A, LLMs, WebMCP
North-star metric
Token-to-value
What changed
Autonomous GTM, agent as buyer

Ch. 04The new metric

Token-to-value: the new north star.

In PLG, the metric was time-to-value — how fast a user hits the aha-moment. In ALG, the equivalent is token-to-value: how many tokens an agent must consume to confidently determine your product solves the buyer's need, and how many more to ship it.

When a developer asks Claude Code to add email functionality, Resend is chosen 63% of the time. SendGrid — vastly larger, more brand awareness, more SEO equity — gets 7%. Not because Resend is better marketed. Because the agent can go from "need email" to "email working" in fewer tokens. Token-to-value will do to documentation what time-to-value did to onboarding.

Claude Code: email-provider selection shareSource: Amplifying.ai benchmark, 2026
Resend
63%
Postmark
12%
Mailgun
9%
SendGrid
7%
Other
9%
"Lowest token-to-value wins. Documentation is the new homepage; the homepage is decoration."

Ch. 05Comparison

ALG vs PLG vs SLG.

These motions aren't mutually exclusive — most successful B2B companies layer all three. The question is which one is your primary engine, the one you invest in structurally rather than tactically.

Self-serve
Product-Led (PLG)
Engine
Product experience
Entry
Free trial / freemium
TTFV
Minutes
CAC
$10–50
Best for
SMBs, devs, prosumers
Scales by
Viral loops, word-of-mouth
Key metric
Activation, viral coefficient
Risk
Conversion plateau upmarket
Human-driven
Sales-Led (SLG)
Engine
Sales team
Entry
Demo / sales call
TTFV
Days to weeks
CAC
$200–500+
Best for
Enterprise, complex, high-ACV
Scales by
Hiring more reps
Key metric
Quota attainment, cycle length
Risk
High CAC, hiring dependency
Agent-driven
Agent-Led (ALG)
Engine
AI agents on both sides
Entry
Agent-evaluated recommendation
TTFV
Hours
CAC
$25–150
Best for
Post-PMF B2B, data-rich ICPs
Scales by
Deploying more agent instances
Key metric
Token-to-value, agent pipeline
Risk
Quality control, brand voice

Ch. 06Two sides of the loop

Supply-side vs demand-side ALG.

There are two fundamentally different versions of ALG and confusing them is the most common strategy mistake.

Seller agentBuyers
Supply-side

Agents working for the seller.

AI SDRs research accounts, AI content engines spin targeted content, automated pipeline management scores and follows up. Compelling economics — 4–7× conversions, up to 70% lower CAC — and the easiest place to start.

Who deploys
The seller
Changes
Funnel efficiency
How to win
Better data, sequences, agents
Key metric
Meetings booked, pipeline generated
Maturity
Production-ready
VendorsBuyer agent
Demand-side

Agents working for the buyer.

A procurement team asks an agent to evaluate CRMs. A developer asks Claude Code to add payments. A marketer asks an AI to build a competitive analysis. The agent makes — or decisively shapes — the buying decision.

Who deploys
The buyer
Changes
Market structure
How to win
Better docs, simpler integration, transparent pricing
Key metric
Token-to-value, agent selection rate
Maturity
Early but accelerating
"Supply-side ALG improves the economics of your current funnel. Demand-side ALG changes whose funnel it is."

Ch. 07The framework

The ALG360 operating model: Introduce → Evaluate → Execute.

ALG isn't a product — it's an operating model. ALG360 sequences it into three pillars that map to the buyer-agent's actual workflow: get introduced to the agent, win its evaluation, then let it execute against you instead of around you.

Ch. 08In production

What agents actually do, all day.

Theory is useful. Here's a real day in an ALG360-instrumented motion, mapped to the three pillars and the ALG360 capabilities that run them.

Introduce
ALG360 · Agent Readiness Scan
  • ·Weekly retrieval sweep across Claude, ChatGPT, Gemini and Perplexity
  • ·Citation, schema and MCP-discovery tracking on hundreds of buyer prompts
  • ·Machine-readable surfaces (schema, .well-known, structured pricing) shipped on a queue tied to retrieval gaps
  • ·Human SEO/AEO referred to Teknicks — out of scope for ALG360
Evaluate
ALG360 · Evidence Library
  • ·Agent-readable docs, pricing and comparison surfaces
  • ·Trust signals and proof points in one source of truth
  • ·Continuous token-to-value scoring as competitors ship
Execute
ALG360 · MCP / API Handoff
  • ·Agent-native trial endpoints with zero-friction provisioning
  • ·WebMCP instrumentation on the top conversion paths
  • ·Agent-confirmable checkout wired to your billing stack
A concrete workflow
Signal → booked meeting
  1. 07:02Signal

    Agent Readiness Scan flags an account researching your category in Perplexity and ChatGPT this week.

  2. 07:03Evaluation

    Buyer's agent pulls your evidence library — docs, pricing, comparison page — and stacks you against two competitors.

  3. 07:04Shortlist

    Your name comes out of the model with a recommendation. The human never opens your homepage.

  4. 07:05Provision

    Agent hits the trial endpoint, gets a working sandbox in seconds, and brings the buyer back a live demo.

  5. 07:06Confirm

    Two-step checkout: agent proposes the cart, human one-taps to confirm. Pipeline attributed to the agent-driven motion.

  6. Day 2Expand

    Usage signal triggers an expansion prompt. The agent surfaces the next plan; your team owns the relationship.

Human time per account: ~5 minutes of review. Agent time: ~2 minutes of execution. A traditional SDR spends 30–45 minutes on the same loop.

Ch. 09Measurement

Metrics that matter (and which module owns them).

Measure agent performance separately from human performance so you can optimize each independently. Early on, focus on cost-per-meeting and quality. As you scale, shift focus to pipeline generated and meeting conversion.

Token-to-value
Evaluate
Lower is better

How many tokens an agent needs to recommend you

Agent retrieval share
Introduce
% of prompts retrieving you

Across Claude/ChatGPT/Gemini/Perplexity

Agent pipeline generated
Execute
Track monthly trend

Pipeline attributable to agent-initiated motion

Cost per meeting
Execute
$15–50

vs. $200–400 for a human SDR

Response rate
Execute
5–15%

Cold agent-driven outbound

Meeting conversion
Execute
20–40%

Of conversations → booked meetings

Agent-completion rate
Execute
>80%

Of MCP-instrumented paths an agent can finish

Quality score
Evaluate
>80% acceptable

Human review of agent-generated artifacts

Ch. 10The math

The economics of an agent-led motion.

A fully-loaded human SDR runs $75–110K/yr. An AI SDR stack — tooling, orchestration, and oversight — runs $24–60K/yr. Early adopters report 4–7× conversion lift and up to 70% lower CAC. The motion doesn't replace humans; it changes what humans get paid to do.

Human SDR vs ALG360-run AI SDRIndexed: human SDR = baseline
Fully-loaded SDR
Human 92 · AI 42
Human
AI SDR
Cost per booked meeting
Human 300 · AI 32
Human
AI SDR
Time per account (min)
Human 38 · AI 7
Human
AI SDR

On the Growth tier, a single replaced SDR headcount roughly funds the entire ALG360 subscription — before counting introduce-side wins, conversion lift, or expansion. The math is conservative; the compounding is not.

Ch. 11Anti-patterns

Six ways to do ALG wrong.

1

Bolting AI onto a broken GTM system

If your ICP is undefined and your pipeline stages don't reflect reality, agents will automate the chaos.

What ALG360 does: The Agent Readiness Scan maps the system before any agent touches it — free on the Starter tier.

2

Treating agents as standalone tools

An AI SDR tool in isolation isn't ALG — it's a silo with a chatbot.

What ALG360 does: ALG360 stitches Introduce, Evaluate and Execute into one signal layer.

3

No human-agent interface design

Without clear escalation, you get either too much oversight (defeating the point) or too little (damaging the brand).

What ALG360 does: Execute ships the governance and consent layer, not just the bots.

4

Optimizing for volume over quality

A 2% error rate at 500 messages/day is 10 brand-damaging interactions every day.

What ALG360 does: Quality score is a tier-1 metric in the ALG360 product.

5

Ignoring the recommendation moat

If you only ship outbound agents, you compete on outbound efficiency forever.

What ALG360 does: Introduce + Evaluate build the recommendation moat first — Execute compounds it.

6

Measuring activity instead of outcomes

'The agent sent 2,000 emails this week' isn't a metric. '14 qualified meetings at $35 each' is.

What ALG360 does: ALG360 reports outcomes by default; activity is diagnostic only.

Ch. 12The path

How to get started.

You don't rebuild the motion overnight. ALG360 is software you adopt in three steps — scan, ship, wire — designed to compound. Every team starts with the free Agent Readiness Scan.

Step 01

Run the Agent Readiness Scan

Free · Starter

Sign up and run the scan on your domain. You'll see exactly where you're being missed in Claude, ChatGPT, Gemini and Perplexity — and the highest-leverage fixes to ship first.

Step 02

Ship Introduce + Evaluate

Growth · monthly SaaS

Track agent retrieval weekly, build your agent-readable evidence library, and watch token-to-value drop across your competitive set. Continuous scoring as your competitors ship.

Step 03

Wire Execute

Enterprise · monthly SaaS

Turn on MCP / API handoff, agent-native trial endpoints, and agent-confirmable checkout. The buyer's agent transacts with you instead of around you — and the recommendation moat compounds.

Ch. FAQFrequently asked

The questions every team asks.

Trusted by post-PMF B2B teams

Thermo Fisher ScientificWebMD Health ServicesIndex Engines

Get picked by the buyer's agent.

Start with the free Agent Readiness Scan. You'll see exactly where you're being missed in the model, where token-to-value is bleeding, and the highest-leverage motion to ship first.