Naming the category, then building everything that depends on it.

A product that had outgrown its own story

Paragon was an embedded iPaaS for SaaS companies building AI features — copilots, agents, and RAG-based products that needed to read and act inside their customers' third-party tools. I took this on as an independent GTM exercise while the company searched for its first product marketing hire, to pressure-test whether its positioning matched the product reality.

The primary buyers were platform and product engineering leaders: VPs of Engineering, Heads of Platform, and CTOs. They owned the build-versus-buy decision; product leaders shaped the AI roadmap, while security and compliance leads could later block a deal over SOC 2, GDPR, deployment, or credential handling.

The opening came from a visible mismatch between the product and the market story. I compared the homepage, sales deck, and job postings against the CEO's own public explanation that the product had been rebuilt because RAG required massive data ingestion and agents needed universal tool calling. The product already supported four integration modes — data sync, agent actions, event triggers, and orchestration — on multi-tenant, per-end-user authentication infrastructure. The outward-facing materials still sold an embedded iPaaS competing on connector count. That gap was the evidence and the opportunity.

Competing on a number it could not win

Paragon was underselling its hardest-to-copy advantage and fighting on raw tool count. Meanwhile, AI-native challengers were not built for the same enterprise-scale, multi-tenant requirements — but that distinction existed nowhere anyone outside engineering could repeat it. There was no clear positioning, no category name, and no competitive story.

Position first. Then everything else.

The pivotal decision was to name a new category: Integration Infrastructure Platform. I rejected staying inside "embedded iPaaS" and repositioning incrementally around faster connectors or better auth — that keeps you competing with established players on connector count, on a product that has outgrown the category. I also rejected framing it purely as an AI-native tool-calling layer, which would have put Paragon head-to-head with a competitor on the one metric where it looked smaller, and hidden the broader advantage: one system supporting four integration patterns.

I built the positioning framework first, using the Dunford model, anchoring the argument to the CEO's public explanation of why the product had been rebuilt for AI rather than to the outdated website copy. That fixed the competitive alternatives, the unique attributes, and the category claim. Those decisions then translated into a messaging framework and buyer persona work for the platform buyer, the product leader, and the security blocker.

With positioning settled, the rest of the system followed: competitive landscape and battlecard for the competitive story; analyst brief and category-creation POV to pressure-test the claim and name the category publicly; sales narrative, launch announcement, and pricing and packaging narrative so the position held across every downstream motion an AI-native scaleup actually runs. Ten documents, one coherent story.

Twenty-seven years of positioning and competitive work did the diagnosis and the judgement: it identified the mismatch, chose the category, and tested whether the language could withstand technical scrutiny. Because this was external work without a formal internal review process, I made public source traceability and cross-document consistency the review function — every claim traceable, every document checked against the others for contradictions, terminology drift, and assertions that needed a confidence label.

A system that agrees with itself

The complete ten-document system — positioning, messaging, persona, competitive landscape, battlecard, analyst brief, category POV, sales narrative, launch announcement, and pricing narrative — was produced in six hours, from first reading the public materials to finished, cross-checked documents. A seasoned director would typically need about a week to diagnose the category, draft the interlocking documents, and manually validate the copy.

The work was not launched, and no one at Paragon reviewed or adopted it. So there is no pipeline, win-rate, adoption, or market-response figure to report, and I won't invent one. Its result is the artefact itself: a self-consistent proof of what a first product marketing hire could build from public evidence alone, at a pace that would otherwise be difficult to achieve.