Compass Advisory Group
I help companies decide where AI belongs in their business, then build it — one thread at a time, something real every one to three weeks. No hours reports. No strategy decks. You own everything.
It started in a cigar bar, over a whiskey. My friend David — twenty years running a good software company, no panic in him at all — set his glass down and asked: “How should I use AI in my company?” Refusing to answer fast was the answer.
The real cost: Your people stop believing in AI — right before you need them to.
One question exposes it: “Which number does this move?”
The real cost: Standing still in a moving market — and standing still is moving backward.
One question exposes it: “What's different since the plan arrived?”
95%
of enterprise GenAI pilots delivered no measurable return.
MIT, The GenAI Divide (2025) — $30–40B of enterprise investment examined.
And it wasn’t the models. It was where they got bolted on.
The method
Five systems you run × five things they run on. You can draw it in an hour. Everything anyone has ever sold you lives in one column — Vendors. Your answers live in the other four.
Every cell has a depth: six tiers, from “your people already use their own AI” to fully AI-native. You tier cells, not companies. Tier 4 in sales and tier 0 in finance isn’t failure — it’s an honest map.
Deeper gold, deeper tier. Outlined cells we built ourselves. Dark cells are empty on purpose. This is the method from my book, Where to Put AI in Your Business.
| Process | People | Tech | Data | Vendors | |
|---|---|---|---|---|---|
| Generate Revenue | |||||
| Deliver Value | |||||
| Support Customers | |||||
| Enable Operations | |||||
| Govern & Protect |
Hover or tap any cell to read its tier — this is my own company’s map, cell by cell, honestly. Not a case study.
The pattern
Every tool gets an address — a cell on the map, not a vibe. Buying beats building two to one.
Everything AI produces flows to one place. Efficiency pays for it; the database is the prize.
One cell at tier five — the one where your advantage lives. Buy everywhere else so you can afford it.
No AI move without an address on your map.
“If I can’t fill in the cell with confidence, AI doesn’t get near it.”
A blurry process doesn’t get automated. It gets defined first. That’s what the 95% died of: fog, automated. Some addresses stay empty — on purpose.
The offer
The unit of work is an AI Development Thread: a lane, not a timesheet. One build target actively in motion at all times, and exactly one. When it lands, the next starts — same day. You are buying concurrency and continuity, not hours.
| Service | Investment | What it is |
|---|---|---|
| Onboarding | $10,000 one-time | What you have, what's worth keeping, what to build first, and how it all connects. ~30 days. |
| Development Retainer — 1 Thread | $5,000 / month | One builder, one thing at a time, always moving. Something real every one to three weeks. |
| Working Session | $2,000 / month | Ninety minutes, twice a month. The cell map — what to build next, what to stop — plus what shipped in AI and what it means for you. |
| Production Support | $1,500 / app / month | For the MVPs that earned it: monitoring, fixes, updates, and someone to call. |
| Additional Thread | $4,500 / month | Another lane. Two problems in motion at once instead of one. |
Not priced in hours, deliberately. Generation is no longer the bottleneck — direction, judgment, and verification are. The measure is what got built and how fast.
The work is done by me, directly. Where a build needs specialist depth I lack, I bring in help at my cost, under my accountability.
You will never receive an hours report — and you should never want one.
Most ideas die at stage one — the cheapest place to die. Stages one through five cost nothing beyond the retainer. Killing a build costs nothing, carries no penalty, and is a good outcome. Only what earns production gets production — with SLAs, on your signature.
Who you're hiring
Adam Anderson has spent 26 years in cybersecurity and technology: founded or invested in ~30 companies, advised 100+ CEOs and heads of state across nine countries, sold his own firm in a multimillion-dollar exit. Today he runs an AI-native software company (ThreatCaptain) — the operator's vantage point behind the advisory work. Author of Where to Put AI in Your Business and the forthcoming Shovels and Maps. Two-time TEDx speaker.
This is how I run my company — and I built the software that runs it.
Your whole map — 25 cells, tiers, buy / build / don’t — in about an hour. The cells you can’t fill in with confidence are your real to-do list.
Forty-five minutes together. Bring your org chart and your P&L frustrations; I’ll bring the map. If there’s nothing worth building, I’ll tell you that, too.