arrow_backBack to What I Build

Operational Map

The org chart on the wall is not the org. The Slack threads are.

This is what the work looked like in a real business: the founder could feel the company getting tangled, but nobody had one view of how 31 people and 7 products actually depended on each other. I turned the tools they already used into a live map: load-bearing people, product collisions, communication dead zones, and the questions that kept getting asked without answers.

What Usually Looks Normal Until I Point At It

Three people are the whole company and nobody has said it out loud

Two product teams are doing the same work in different Slack channels and don't know it

The real org chart lives in a Slack DM thread the owner started in 2021

How I Build It

01

Intake: Slack, Asana, CRM, and the org roster feed one live model of the company

02

The org runs through an instrument panel — the people everything depends on, the products quietly colliding, the communication dead zones

03

You keep a live interactive map, a written read on what it found, and a walkthrough — yours, not a slide deck I take back

Outcomes

Named the people everything depends on, with severity—not a vague warning about what breaks when they're out

Product-entanglement map showing which teams share load and where collisions are coming

Communication flow trace—where handoffs happen, where they fail, where knowledge disappears

Technology Stack

ReactPythonSemantic memory substrateLocal LLMsSlack/Asana APIsAdversarial research agents

Capabilities

Phase 0 — Operational Audit

The wedge. You give me read-only access to Slack, Asana, and your org roster. In two weeks you get observations, a walkthrough, and a live map of your organization with your load-bearing people, product collisions, and communication gaps rendered against your actual data. If you stop here, you keep the map.

How It Works

  • Days 1–3: Intake. Slack export, Asana export, CSV roster, product list ingested into engagement JSON
  • Days 4–10: Instrument panel runs—people graph, product ecosystem, communication flow, journey steps, load-bearing-people dossiers populate

Outcomes

  • Named load-bearing people with severity ranking and what breaks when each person is out
  • Product-entanglement graph showing which teams are secretly one team

Phase 1 — Live Deployment

The audit is a snapshot. Phase 1 stands up the live system — a local-first RAG stack with a multi-agent swarm (entity, relational, temporal, pattern detectors feeding a local-LLM aggregator) — against your live sources so the map updates as your org changes. Self-hosted on your hardware or ours; data never leaves your perimeter. Phase 1 output: ongoing question capture from Slack, surfaced as the primary signal for what your company is asking and forgetting.

How It Works

  • Stand up the multi-agent swarm on local or dedicated inference
  • Connect Slack and Asana as live sources, swap the one-time export for a streaming ingest

Outcomes

  • Live map that reflects current state, not a two-week-old snapshot
  • Question capture as the flagship signal—what is the company asking that nobody is answering

Related Case Studies

The research is public.

Benchmarks run on machines I own, measured against frontier systems — with the failed approaches reported next to the results. Read it before you take my word for anything.