Position Pilot
Most B2B founders spend weeks piecing together positioning, ICP, and messaging before they can build a campaign. Position Pilot compresses that into a single governed pipeline — with a human review gate between stages so strategy is validated before execution begins.
Overview
Strategy work isn't usually short of information. It's short of structured synthesis. Position Pilot takes structured business inputs and moves them through a sequence of specialised agents designed to turn raw context into a coherent GTM strategy.
Three layers, 15 agents
Positioning · ICP · Messaging · GTM · SEO/AEO — the core 5-agent engine, live and deployed
Pain Excavator · Trigger Mapper · Persona Architect · Messaging Strategist · Positioning · Narrative · Opportunity · Differentiation — 8 agents across companion research pipelines, built and published, not yet wired into the live product
Compliance Audit · Brand Rewrite — validates all generated copy against brand voice rules and auto-rewrites flagged sections
A real run — Fathom · 7 June 2026 · Strategy-layer output
An AI meeting-intelligence platform competing with Otter.ai and Fireflies.ai. Position Pilot took business context and produced positioning, ICP, messaging, GTM milestones, and search strategy — five real outputs from one run.
"For sales and customer-facing teams who lose critical information from meetings because manual note-taking distracts from the conversation, Fathom is the free AI meeting intelligence platform that automatically records, transcribes, and summarizes calls from Zoom, Google Meet, and Microsoft Teams — unlike Otter.ai or Fireflies.ai which charge for full features and require complex setup."
A prospect complains about being asked the same questions multiple times across conversations. Leadership reviews call recording costs and sees the team is paying for expensive conversation intelligence tools with low adoption.
Never Miss a Customer Promise Again — Free. Stop paying for meeting intelligence. Start using Fathom for free — get your first summary in your inbox after your next call.
Day 7: 500 free-tier sign-ups from targeted outbound. Day 14: 3 CRM integration partnerships. Day 30: 25% week-over-week user growth, 70%+ first-week activation.
"Otter.ai vs Fireflies.ai free features comparison" · "free alternative to Read AI" · "how to stop manual note-taking in meetings"
This example predates some of the architecture above — shown as what the system actually produced, not a simulation of what it would produce today.
Architecture
The governance layer above runs in n8n. It's also been rebuilt in LangGraph + FastAPI — same audit/rewrite logic, different architecture — to compare a code-first agent stack against a low-code one. langgraph-agents ↗
What I learned
The hard part of an agentic system isn't making more agents. It's deciding what each agent should own — what information should move between them, where humans should intervene, how failures become visible, and what evidence tells you the output is actually useful.