Liam Ottley's 5-Layer Playbook For Building An AI Operating System
Liam Ottley generated $1M NZD in 7 days using the bandwidth his Claude Code AIOS freed up. The methodology — context, data, intelligence, automate, build — is the cleanest framework yet for wrapping AI around an existing business.
Liam Ottley's been the loudest voice claiming AGI is already here, and the loudest about what specifically he means by it: an AI Operating System — Claude Code wrapped around an existing business, built up in layers, eventually thick enough to run from a Telegram chat on your phone. He just dropped the methodology in full.
The specific result that makes this worth taking seriously: $1 million NZD in revenue over 7 days during a webinar launch he ran using the bandwidth his AIOS had freed up. He planned the entire webinar, built the offer, made the slides, set up the funnel, filmed YouTube traffic videos, and shipped — in a week. That's the move you can't make until the operator trap is solved.
"This isn't a business model. It's a methodology — a layer that wraps around your business model."
Here's the full framework, demystified.
The operator trap (the problem AIOS solves)
Every founder is stuck in the same shape: 80% of bandwidth spent in the business (operations, admin, fires) and 20% on the business (new initiatives, growth bets, new products). Liam's data-point — and the one most founders will recognize — is that the businesses growing fastest are the ones that flipped the ratio. 80% on the business, 20% in.
AIOS is the explicit framework to flip it. Five layers, deliberately stacked, each built on the one beneath.
Layer 1 — Context
The foundation: Claude Code needs to know who you are, what you sell, who your team is, what your values are. This is the layer that eliminates the "let me re-explain my business for the 40th time" problem most people have with ChatGPT.
What goes here: company overview, products, services, team roster, voice/brand guide, customer personas, current quarterly strategy.
Layer 2 — Data
On top of context, connect every data source so Claude can query quantitatively. Liam's stack:
- Stripe (revenue)
- Bitly (link clicks)
- Google Sheets (KPI rollups)
- YouTube (channel performance)
The test of whether this layer works: you can ask "What's the connection between Bitly clicks, YouTube videos posted, and revenue this month?" and get a real answer. Without this, Claude is making business-shaped suggestions without business-actual data.
Layer 3 — Intelligence
The meeting and communication layer:
- Fireflies (meeting recordings + transcripts, logged to a local database)
- Slack (every message pulled into the same queryable store)
This is the layer that powers the Daily Brief — Liam's signature output. Every morning he gets:
- A 5-10 page PDF report with full analysis across departments and businesses
- A Telegram summary with key signals, revenue moments, strategic insights, content ideas
- An auto-generated funnel breakdown image showing where every channel sits
The daily brief works only because layers 1-3 are stacked. You need context to know what matters, data to know the numbers, and intelligence to know what was discussed. It's the most valuable first-checkpoint output once you have all three.
Layer 4 — Automate
This layer starts with a task audit. Liam keeps his on a whiteboard: every task he does across his content business, agency, education business, and SaaS portfolio. The goal is to cross off 60-70% — the target percentage of tasks fully automated or heavily augmented.
The key Claude Code primitive here is the /explore command (which Liam emphasizes especially for non-technical founders). You say "I do this task every week / explore. Help me figure out how to automate it." Claude searches the codebase, the web, and your business context to propose an implementation.
Two automation patterns to layer in:
- Cron-scheduled tasks (every 30 min, every hour, every day) running automatically without prompting
- Plug-and-play modules — once you've automated a workflow, package it as an installable Claude Code skill that other founders can adopt directly
Layer 5 — Build
The payoff layer. Once you've freed 50-70% of your bandwidth, you have a choice: rebuild your lifestyle around the freed time, or pour it into new growth initiatives. Liam picks the second — that's what produced the $1M week.
Building inside the AIOS workspace is dramatically faster than building from scratch because Claude already has full business context, API access (Vercel deploy, ConvertKit, Calendly, etc.), and skills installed. You're not starting from zero on every new project. You're starting from the top of your context stack.
The three KPIs Liam tracks
Without measurement, AIOS is vibes. He tracks three numbers:
| KPI | What it measures |
|---|---|
| Away-from-desk autonomy % | How much of your job can you do from your phone via Telegram? Test by trying to work from a coffee shop / beach / weekend trip. |
| Task automation % | Of the items on your task audit, how many are now fully automated or heavily augmented? Target: 60-70%. |
| Revenue per employee | The new flex of the AI era. Either revenue going up (employees flat), or revenue flat (employees down), or both. |
The revenue-per-employee point is the spiciest one. Liam: "There's a bloodbath coming to the small business world in terms of headcount." He explicitly says people running his methodology are reconsidering hires. The arbitrage window is whoever ships the AIOS first while service costs are still high.
Who this is built for (and who it's not)
AIOS is for small businesses — the nimble ones who can drop a tech stack, swap APIs freely, and don't have ERP lock-in. It's specifically not for big companies with legacy systems. Liam: "This is the benefit of the small business."
The founder profile this works best for: agencies, consulting shops, content businesses, SaaS founders. Anyone with a service or product business under ~50 people who's still personally operating.
What to actually do this week
Liam's framework can be started in one weekend:
- Spin up a Claude Code workspace dedicated to your business — a single folder with
CLAUDE.md, business context files, and adata/directory. - Stack layer 1 (context): write a single context document covering your business model, team, products, strategy. Treat it as living — it'll grow.
- Stack layer 2 (data): connect at least one data source (Stripe is the most common entry point). Validate that you can ask quantitative questions and get real answers.
- Set up the daily brief automation: cron a Claude Code task to run at 7am that summarizes the past 24 hours and sends it to your Telegram. This is the highest-leverage first artifact.
- Run the task audit before going further. List every task you personally own. Pick 5 to automate this month. Then layer in the automation step-by-step.
Do not skip ahead to layer 4 before layers 1-3 are solid. The automations only work because the context, data, and intelligence layers make them possible.
The Bottom Line
AIOS is the cleanest framework currently published for using Claude Code as a business operating layer rather than just a coding assistant. The methodology — context, data, intelligence, automate, build — is repeatable. The KPIs are measurable. And the upside (a $1M week from freed bandwidth) is the proof of concept Liam set out to demonstrate. The founders running this first, while service costs are still high, are the ones the next 18 months belong to. The framework is public now. The arbitrage window is shorter every week.