Rust Just Set the Highest Bar for AI in Open Source. Every Maintainer's Inbox Is About to Change.

In Q1 2026, Gartner projected something that most product teams hadn't fully absorbed: 40% of enterprise applications would have embedded, task-specific AI agents by the end of the year, up from less than 5% in 2025. That 35-point jump in 12 months is not a product feature story. It is a distribution story — and it has consequences that reach every activation funnel, retention dashboard, and pricing model in B2B software.

The product-led growth playbook — the one that made Slack, Notion, and Figma into category leaders — was written for a specific kind of user: a human being who navigates an interface, discovers features, experiences an aha moment, builds a habit, and eventually pays or invites teammates. The entire PLG stack — the onboarding flow, the activation milestone, the DAU/MAU dashboard, the virality coefficient — is calibrated for human cognitive patterns and human decision-making cycles.

When your user is an AI agent, none of those calibrations apply.

Agent-Led Growth (ALG) is the fourth major go-to-market paradigm, following Sales-Led Growth (SLG), Product-Led Growth (PLG), and Community-Led Growth (CLG). According to the AgentLedGrowth Research team's 2026 State of AI Agents report, ALG describes the pattern in which AI agents — rather than human users — drive product adoption, trigger activation events, generate expansion revenue, and create the network effects that compound growth. As of mid-2026, it is not a future scenario. It is the current operating reality for a growing slice of the enterprise software market.

The product teams that understand this are rebuilding their activation infrastructure for headless growth. The ones that don't are watching DAU metrics that measure a customer base they're already losing.

The PLG Flywheel and Its Fundamental Assumption

Product-led growth works because software products can acquire their own users. When a Slack invite lands in an inbox, Slack acquires a user without a sales rep. When a Notion doc is shared, Notion extends its reach without a marketing budget. The PLG flywheel depends on one assumption so basic it usually goes unstated: the user who receives the invite, opens the doc, or tries the free tier is a person.

That assumption is breaking down. Not because AI has replaced human users — humans still buy software, use it, and make renewal decisions. But because an increasing share of the actual product usage — the API calls, the workflow executions, the data reads and writes — is happening autonomously, driven by AI agents acting on behalf of the humans who configured them.

According to Userpilot's 2026 analysis of user adoption metrics, the critical inflection point for any product team is when agent-generated usage exceeds 20% of total API volume. Below that threshold, agent and human usage patterns are similar enough that human-optimized activation funnels still work. Above it, the divergence becomes measurable: human DAU grows while agent API volume — the actual driver of product value and expansion revenue — is tracked with no dedicated metrics at all.

B2B SaaS companies are the most exposed. The State of AI Agents 2026 report places B2B SaaS companies at 45% AI agent adoption — higher than financial services (38%) or healthcare (29%) — which means the sector most dependent on PLG as a growth motion is also the sector where PLG's foundational assumption is under the most pressure.

The Three Eras of PLG

The evolution of product-led growth follows a clear arc. Understanding where you are on that arc determines which growth levers actually work.

SaaS Mag's 2026 analysis of PLG evolution frames the transition this way: "Software is heading toward a headless, agentic future where AI agents use SaaS tools instead of human end users." That characterization is accurate but incomplete. The transition is not binary — it is layered. Most enterprise software products in 2026 are simultaneously in PLG 1.0 for some buyers, PLG 2.0 for others, and beginning the PLG 3.0 transition for a third cohort that has embedded them into AI agent workflows.

The companies pulling away in 2026 are the ones that have identified which cohort is which and built differentiated activation paths for each.

How Agent-Led Growth Actually Happens

Human PLG follows a recognizable sequence: awareness → signup → activation → retention → expansion. Each stage has discrete interventions: the onboarding flow, the email nurture, the feature discovery modal, the in-app coach.

Agent-Led Growth follows a different sequence: API discovery → integration → first successful task completion → workflow embedding → organizational expansion.

The differences at each stage are operationally significant.

Discovery for human users happens through brand channels, word of mouth, and content marketing. Discovery for AI agents happens through model training data, MCP server registries, and the API documentation that crawlers have already indexed. Products that have not optimized their API documentation for AI legibility — clear, structured, complete, with explicit error-handling guidance — are invisible to the agents that would otherwise integrate them.

Integration for human users is measured in time-to-first-value, typically minutes to hours. Userpilot's 2026 data shows the median time-to-value for self-serve accounts under $5K ARR has compressed to 11 minutes. Integration for AI agents is measured in API call success rate in the first 24 hours. An agent that encounters ambiguous API responses, inconsistent error codes, or undocumented rate limits during initial integration will be pointed at a different vendor before the human buyer even knows there was a problem.

Activation for human users is an aha moment — the first time the user experiences the core value proposition in a way they understand and remember. Activation for AI agents is a completed workflow — the first time the agent executes a task from start to finish without encountering an error that halts execution. Products that measure activation only as human aha moments are missing the activation event that determines whether the agent integration persists.

Retention for human users is habit formation — the product becoming part of the user's daily routine. Retention for AI agents is workflow embedding — the agent's reliance on the product becoming structural, a dependency that makes substitution expensive. The most durable AI agent integrations are those where the product's API is in the critical path for a workflow the organization cannot easily replace.

The Activation Metrics That Break When Your User Isn't Human

The standard activation metric stack was designed for human users. Applied to AI agents, it produces data that is actively misleading.

DAU/MAU ratio. A high DAU/MAU ratio signals habitual human engagement. When agents drive API calls, DAU/MAU can look excellent while human engagement is declining. The metric conflates two very different engagement patterns. As Signal previously analyzed, separating human events from agent events is now a prerequisite for accurate retention measurement — not a nice-to-have.

Feature adoption rate. Human activation funnels track whether users discover and engage with key features. AI agents don't "discover" features — they call endpoints. A feature that is never displayed in the UI but is accessible via API may have zero human adoption and 100% agent adoption. Feature adoption rate measured at the UI layer misses agent usage entirely.

Session depth and session length. These metrics measure human attention. An AI agent executing 10,000 API calls in three minutes with no UI sessions looks like a churned user by session metrics and like an extremely high-value customer by revenue metrics. Products that rely on session metrics to identify expansion opportunities are missing their fastest-growing customer segment.

NPS and satisfaction surveys. You cannot send an NPS survey to an AI agent. The human who configured the agent may not have direct visibility into the agent's experience of the product. Satisfaction measurement for agent-integrated products requires instrumenting the API — measuring error rates, latency, and task completion rates — rather than surveying the human buyer.

Agent Churn: The Metric Your Weekly Report Misses

Human churn is slow. It builds through declining engagement, unresolved support issues, competitive evaluation, and eventually a cancellation decision. The timeline typically spans weeks or months, and leading indicators — drop in DAU, declining feature breadth, reduced seat counts — give product and retention teams a detection window.

Userpilot's retention analysis identifies the structurally different pattern of agent churn: "An integration breaks, a workflow gets replaced, or the agent gets pointed at a competing tool, causing prompt volume to drop immediately, and by the time the metric registers on a weekly retention report, the decision may already be made."

Agent churn is instantaneous. When a workflow breaks or a better API becomes available, the agent is redirected in a configuration change that takes seconds. The prompt volume drops to zero immediately. There is no declining engagement curve, no retention email that catches the problem in time, no win-back opportunity that arrives before the replacement is already embedded.

This has three practical consequences for retention teams.

First, monitoring cadence must change. Weekly retention reports are designed to catch the slow decline of human churn. Agent churn requires real-time monitoring — API health dashboards, integration status alerts, prompt volume anomaly detection — that operates on a minutes-to-hours basis, not days-to-weeks.

Second, churn signals look different. Human churn signals are behavioral: declining logins, fewer features used, reduced session depth. Agent churn signals are technical: increased error rates, endpoint latency spikes, authentication failures. Teams looking for behavioral signals will miss the technical signals that precede agent churn.

Third, the recovery path is different. Human win-back campaigns use relationship, pricing, and value reinforcement. Agent re-integration requires debugging the technical failure that caused the disconnection. Products that have not built agent-specific support workflows — API debugging tools, integration health diagnostics, agent-optimized error documentation — will not win back the agent users they lose.

Repricing for Agent Users: From Access to Outcomes

Per-seat pricing was designed for human users because seat count is a reasonable proxy for human value received. One person = one license = one set of value delivered. For AI agents, the proxy breaks completely: a single enterprise can deploy 100 agents, each executing thousands of API calls daily, under a single seat.

The market is already correcting. According to the State of AI Agents 2026 research, 43% of SaaS companies now use hybrid pricing models — combining seats, usage, and outcome-based components — with adoption projected to reach 61% by end of 2026. The direction is clear: pricing is moving from "pay for access" to "pay for work done."

Products that have not added agent-tier pricing are leaving expansion revenue on the table. The hybrid pricing models that work for agent users share three characteristics: they measure actual task completion, not login sessions; they price at the workflow level rather than the feature level; and they include infrastructure pricing that scales with agent volume, not human headcount.

The PLG-to-enterprise transition that drove the hybrid GTM playbook of 2025 required product teams to develop enterprise sales empathy. The ALG transition requires something different: AI system fluency — the ability to understand how AI agents consume software, what they need from APIs, and how to instrument workflows that neither begin nor end in a human interface.

The ALG Playbook: Seven Steps to Redesign for Headless Growth

Transitioning from PLG 2.0 to include ALG does not require abandoning the human-user activation funnel. It requires building a parallel track.

1. Audit your metrics for human bias. Identify every metric currently in your activation and retention dashboards. Flag metrics that require a human session — DAU, NPS, feature adoption at the UI layer. Build a second tier of metrics that capture agent behavior: API call volume, workflow completion rate, error rate, task success percentage, and integration depth.

2. Build an agent-specific activation path. Map the sequence from first API call to first completed workflow. Reduce friction at every step: clear API documentation, structured error codes with explicit retry guidance, sandbox environments for integration testing, and webhooks for workflow completion confirmation.

3. Define the agent aha moment. For human PLG, the aha moment is when the user experiences the core value proposition. For agent PLG, it is the first time an agent completes a target workflow without human intervention. That moment must be tracked, celebrated with the human buyer, and used as the activation milestone that replaces or supplements the human aha moment.

4. Shift retention monitoring from weekly to real-time. Implement API health monitoring with anomaly detection on prompt volume, error rates, and task completion rates. Configure alerts at the integration level, not the account level. Human accounts and agent integrations warrant different monitoring cadences.

5. Create agent-tier pricing. Introduce pricing that scales with agent usage rather than human seats. Per-call, per-task, and per-outcome pricing models all work; the right choice depends on where your product sits in the agent workflow. Ensure the contract structure allows for agent volume without requiring manual seat count negotiations.

6. Instrument for agent-specific support. Build debugging tools that give human buyers visibility into what their agents are doing and where they're failing. Provide agent-optimized error documentation — detailed, structured, with specific remediation steps — rather than human-optimized support articles written for users who can describe what they clicked.

7. Design for agent-to-agent virality. The viral loop in PLG 1.0 was file sharing. In ALG, it is output interoperability: agents that produce structured outputs consumable by other agents in adjacent workflows create referral dynamics that don't exist in human-user PLG. Products whose API outputs are well-structured, richly typed, and consistently formatted become natural upstream dependencies in multi-agent workflows.

Who Is Already Winning in the Agent-Led Era

The ALG leaders of 2026 share a pattern: they built exceptional human-user products, then systematically extended those products' reach into AI agent workflows without disrupting the human-user experience.

Cursor's trajectory from zero to $500M ARR in under 24 months — achieving $200M before hiring a single enterprise sales rep — demonstrates what ALG compounding looks like at speed. Cursor is not just a coding tool for human developers. It is the IDE layer through which AI agents like Claude Code, GitHub Copilot, and dozens of enterprise AI development tools operate. Each agent that depends on Cursor's architecture creates a structural integration that is far more durable than individual developer preference.

Salesforce Agentforce, which crossed $1.2B ARR at 205% growth in Q1 FY2027, represents the enterprise pattern: a legacy SaaS platform that used its deep CRM data integration to become the enterprise AI agent orchestration layer. Agentforce's retention moat is not the AI itself — it is that agents built on Agentforce have access to the full history of customer interactions stored in Salesforce CRM. Switching to a competing agent platform requires migrating that data context, creating a switching cost that is exponentially higher than canceling a software subscription.

Block's Buzz, launched in July 2026, represents the most aggressive ALG architecture: a workspace where AI agents hold cryptographic identities and collaborate as organizational peers, designed from the ground up for organizations where humans and AI agents operate on equal footing. Block is not building for human users who want AI assistance. It is building for organizations that need a coordination layer for a workforce where the distinction between human and AI employee is already becoming operationally irrelevant.

What the Activation Funnel Looks Like When You Rebuild It

The product teams that successfully navigate the PLG-to-ALG transition share one operational discipline: they instrument separately for human users and agent users, and they evaluate each activation funnel on its own terms.

For human users, the questions are: What is the aha moment? What friction prevents users from reaching it? What in-product experience drives habit formation?

For agent users, the questions are: What is the first successful task completion? What API behavior prevents integration completion? What workflow embeds the product into an organizational dependency?

These are different questions. They require different instruments, different monitoring, different support, and different pricing structures. Products that answer both — rather than optimizing the human funnel and hoping agents figure it out — are building the activation infrastructure for a decade of compounding growth.

The AI churn paradox — AI-powered apps generate 41% more ARPU but churn 30% faster among users who don't deeply engage — maps directly onto the ALG activation challenge. The "deep engagement" that prevents churn in AI-powered products is not session depth or feature breadth. It is workflow embedding: the moment when the agent's use of the product becomes structural, when switching requires rebuilding workflows rather than canceling a subscription.

Agent-Led Growth is not a replacement for Product-Led Growth. It is the next layer — the growth motion that compounds on top of PLG rather than replacing it, and that is already generating expansion revenue for the companies that have built for it. The Fortune Business Insights forecast of the agentic AI market reaching $139 billion by 2034 from $9.1 billion today implies a 40.5% CAGR — the overwhelming majority of which will flow to products that built their activation and monetization infrastructure for agent users before the growth curve became obvious.

Takeaway: Agent-Led Growth is the fourth GTM paradigm, and the transition from PLG 2.0 to ALG is already underway. Gartner's projection of 40% enterprise AI agent embedding by end of 2026 means that a product team not tracking agent activation, agent retention, and agent churn alongside human metrics is operating blind on a growing share of its customer base. The companies winning in the agent-led era have rebuilt their activation infrastructure for headless users — and they're pulling away from those still optimizing onboarding for a human who may no longer be the primary product consumer.