← WorkSystems · Applied AI

Growth Recon

Enter any company URL and Growth Recon returns a five-stage sales-intelligence dossier — company discovery, brand, competitive landscape, leadership, and an outreach playbook — generated live through a serverless proxy to the Claude API.

Both a portfolio piece and a working tool: it shows what AI orchestration looks like when it is wired into real revenue infrastructure, not just prompted in a chat window.

Live tool

Enter a company URL and run the five-stage dossier.

Runs on the model’s training knowledge (live web search is off for the public demo), so treat the output as a fast starting point for research — not verified fact. A demo rate limit applies.

What it is

Growth Recon compresses the first hour of pre-call research into about ten seconds. Point it at a company and it assembles the same brief a good SDR would build by hand — who the company is, how they position, who they compete with, who to call, and what to say — structured as a dossier a seller can act on.

The five stages
  1. 01

    Company Discovery

    Who they are, size, funding, leadership, and what they sell.

  2. 02

    Brand Analysis

    Positioning, tone of voice, and visual identity.

  3. 03

    Competitive Landscape

    Market size, rivals, advantages, and exposure.

  4. 04

    Leadership Intel

    The executives worth reaching — and how to approach each.

  5. 05

    Outreach Playbook

    Cold email, LinkedIn, content angles, and a 30-day plan.

How it is built

Serverless proxy

A single on-demand endpoint calls the model. The API key lives in a server-side environment variable and is never shipped to the browser.

Strict-JSON contracts

Each stage prompt specifies an exact field shape and demands JSON only. An assistant-message prefill forces a valid object; a tolerant parser recovers if the model ever wanders.

Sequential stages

The five stages run in order, each rendering the moment it lands — so the dossier fills in live instead of blocking on the slowest call.

Guardrails for a public demo

CORS is locked to the production domain, and per-visitor rate limiting caps abuse and cost. It runs on the current cost-efficient Haiku model — a few cents per full run.

Why it belongs here

The interesting part is not the prompts — it is the plumbing. A secret handled correctly, a strict data contract between model and interface, graceful failure, rate limiting, and cost control are the same disciplines that make revenue infrastructure trustworthy. Growth Recon is a small, honest proof that I build AI into systems, not slideware.