The ICP: AI-pilled software teams at any scale

One ICP sits above the four personas: AI-pilled software teams at any scale. What they ship does not matter. The framing follows PostHog's ICP definition, which we share.

"AI pilled doesn't mean building an AI product, it means the team is happy using AI first tools. Doesn't matter what they're building."

What AI-pilled means

• The team ships with coding agents daily and does not consider that remarkable. Claude Code, Cursor, Codex, or similar is in the inner loop, not on trial.
• They reach for AI-first tools by default and expect their vendors to be agent-native: MCP servers, agent-readable docs, a setup you can hand to the agent.
• They have already crossed the trust threshold Polylane depends on: they are comfortable with an agent reading their code, querying their infrastructure, and opening pull requests, as long as they review and merge.
• AI-pilled is not brainless tokenmaxxing. These teams hold judgment, review everything, and are ruthless about noise. The existing anti-persona still stands: anyone who wants a fully hands-off replacement for engineering judgment is out.

Why this is the ICP

• These teams will outperform everyone else. They move faster, and betting on the winners compounds.
• They have strong opinions about what they need, so they pull the product in the right direction instead of dragging it toward feature checklists.
• They are the teams with the problem Polylane solves. Coding agents multiply code, deploys, and production surface per engineer without multiplying ops attention. The gap between what an AI-pilled team ships and what it can watch is the wedge.
• The product only works for them. The whole loop, agents that read your code, watch your infrastructure, investigate, and fix, presumes a team willing to let an agent do real work. A team that is not AI-pilled stalls at the first rung of the connect ladder.

Who qualifies

Scale is not the qualifier

The old instinct was to qualify on stage: high-growth startups. The qualifier is temperament, not headcount. A four-person seed team and a platform team inside an enterprise both qualify if they ship with agents and want agents watching production. What still disqualifies is unchanged in the anti-personas: procurement-led buying, dashboard-staring, legacy datacenter estates, non-GitHub teams.

How the ICP maps to the personas

The ICP describes which teams; the personas describe who inside them. All four personas already share the trait (what all personas share): they use an AI coding agent daily. AI-pilledness is the cut across every persona, at every scale:

• Pete is an AI-pilled team of one with intent to commercialize. The hobbyist version of Pete, building for fun with no commercial intent, is the free-tier "no" row above.
• Max is the AI-pilled team's heaviest user and its informal tooling gatekeeper.
• Laura is the buyer who made her team AI-pilled on purpose and buys leverage instead of headcount.
• Marcus is the proof that "any scale" reaches into the enterprise: an AI-pilled platform team inside a large org qualifies even though the procurement-led version of the same org does not.

What this changes

Nothing about the motion: bottom-up, self-serve, win developers one at a time. What it sharpens is qualification and messaging. Lead with the agent-native surface (MCP, the one-prompt setup, autofix through the coding agent the team already uses), spend in the channels where AI-pilled teams already are (coding-agent communities, MCP directories, build-in-public circles), and stop qualifying on growth stage or company size.