Getting Started

What is AOS? The Agentic Operating System explained

AOS (the Agentic Operating System) is a private operating system you install inside your own company. It is a software product, not a SaaS chat tool. Instead of giving you one chatbot, it stands up a 61-agent company inside your business: eight departments of named specialists, each with a real job description and an acceptance bar for their work. What you get - 8 departments and 61 specialist roles. Office of the CAO, Go to Market, Brand & Marketing, Customer Success, BizOps, Product, Engineering, and Security & Infra. - A mission system. One sentence of direction becomes a scoped, reviewed workstream with an owner, an acceptance bar, non-goals, and the evidence required before anything ships. - Five approval points. The small set of decisions only a founder should make. Risky moves wait for your sign-off. - An evidence trail. Every artifact, draft, decision, and reviewer is captured, two clicks away if you ever need it. - Operating memory. Decisions and the reasoning behind them carry across missions, so the next mission that touches the same territory inherits all of it. - A self-review loop. Builder, QA, reviewer, and security checks run on every artifact before it reaches you. - A Friday founder summary. One page in your inbox at 4pm: what shipped, what is blocked, and what waits on you. Who it is for AOS is built for operating founders running companies between roughly one and a hundred million in revenue who have noticed they are the routing bottleneck. It runs alongside the human operating system you already use (EOS, Scaling Up, EMyth, and similar) rather than replacing it. How it is different from a chatbot A chatbot answers a question and forgets it. AOS scopes real work, assigns it to a named specialist, reviews it four ways, holds the risky parts for your approval, and remembers every decision. You should be able to look at the Mission Control dashboard and understand your entire company at a glance: who is doing what, how much it costs, and whether it is working. Ready to install? See How to install AOS in about 20 minutes.

How to install AOS in about 20 minutes

Installing AOS takes about 20 minutes from sign-up to a private instance with all 61 specialists ready to take orders. There is no PDF to sign, you agree at checkout. This guide walks through what happens and what you will set. Before you start Have these ready and the install goes faster: - The company name and the basics of what you do. - Your spend caps for the money approval point (defaults are $1,000 per move and $10,000 per week, both adjustable here). - One messy priority from the last 30 days that you would like the team to take on first. The install, step by step 1. Sign up and pay the one-time install. $1,000 for the first 100 founders (the Founding Founder License). This includes your first month of hosted maintenance. 2. Your private instance is provisioned. AOS stands up your own workspace with the eight departments and all 61 named specialist agents. 3. Set your approval thresholds. Confirm the money caps and which approval points apply by default. These are the only places customer data, secrets, and production access are allowed to cross. 4. Connect the tools you want the team to use. Integrations are walled off by default; you grant access deliberately. 5. Run your first mission. Type one sentence describing a real priority. The CEO agent drafts a founder brief and asks at most two clarifying questions. You answer or accept the defaults, and that is the only friction. What "done" looks like When install finishes you will have a working Mission Control dashboard, 61 specialists on the bench, your approval points configured, and your first mission already scoped. Most founders look at the dashboard once a day; many look once a week. The agents work either way. Not sure what to run first? The onboarding playbook walks you through it. Most founders pick their messiest priority from the last 30 days, and that is usually the right one. See Running your first mission for a full walkthrough.

Running your first mission

A mission is how you put the 61 specialists to work. You write one sentence of direction in plain English, your founder voice, and AOS turns it into a scoped, reviewed workstream. This article shows exactly what happens from your sentence to the finished work in your inbox. 1. You write one sentence Describe a real priority the way you would say it out loud. For example: "New customers take too long to get value. Fix onboarding." You do not need to specify a department, an owner, or a format. 2. The CEO agent drafts a brief The CEO agent reads your sentence, drafts a founder brief, and asks at most two clarifying questions. You answer or accept the defaults. The brief contains a working goal, three non-goals, and the rough shape of acceptance. Nothing executes until you confirm it. 3. Department leads scope the mission The brief moves to the right department leads, who write the operating plan: specific owners, acceptance criteria, and which of the five approval points apply. For an onboarding mission, Customer Success might own the weekly activation target, Support Ops owns the response, and BizOps owns the escalation path. 4. Specialists do the work in parallel You can watch the Mission Control dashboard if you want, but you do not have to. Before anything reaches your inbox, a four-step review runs on every artifact: the builder reviews their own work, QA tests against the acceptance bar, a department reviewer reads it with fresh eyes, and security scans for anything that crosses a boundary. 5. Approval points light up only when needed If the work hits one of the five approval points, say a refund policy change, the mission flags it, writes a recommendation with the rationale and the alternatives, and pauses that thread until you decide. You click yes or no and the mission continues with that branch confirmed. 6. Your founder summary lands The summary is written for a founder, not for an audit. Every decision, every approval, and every blocker is captured with its reason in the evidence trail, two clicks away if you need it. The next mission that touches the same territory inherits all of that context. How many missions can I run at once? There is no technical cap. The practical limit is your attention span for the five approval points; three to five concurrent missions is typical. You can pause a mission mid-flight at any time.

Your first week with AOS

Your first week with AOS is about getting the agent company pointed at real work and tuning how much it does on its own. You do not need to configure 61 specialists by hand. You set a few defaults, run a mission or two, and let the system show you how it works. Here is what to expect. Day one: provisioning and access Your private instance is stood up on dedicated infrastructure. Out of the box, integrations are walled off, so nothing touches your real tools until you connect them. The first thing to do is sign in, look at the Mission Control dashboard, and read the eight department names so you know who does what. You do not have to connect everything immediately. Connect only what your first mission needs. Set your guardrails Before you run anything important, set the dials that decide how much routes to you: - Money caps. Default is 1,000 dollars per move and 10,000 dollars per week. Lower them if you want tighter control early on. - Public claims. Decide whether anything published under your company name always waits for your sign-off. Most founders keep this on. - Token caps. Start conservative. You can raise them once you see your real burn. These map directly to the five approval points, which are the only places work crosses your boundary. Run your first mission Pick something small and real. One sentence of direction is enough. The department leads scope it into a workstream with an owner, an acceptance bar, and three non-goals, then the builder, QA, reviewer, and security checks run before it reaches you. Watch how the brief turns into scoped work. That single loop teaches you the whole model. See Running your first mission for a walkthrough. Read your first founder summary Every Friday at 4pm you get a one-page summary: what shipped, what is blocked, and what waits on you. Your first one is the moment AOS stops being abstract. Read it, approve or redirect what is waiting, and you have closed your first full cycle. Tune as trust builds The honest goal of week one is calibration. If too much is routing to you, raise the caps. If you want more eyes on things, lower them. If a mission drifted, tighten the non-goals next time. AOS is meant to earn autonomy gradually, so you stay comfortable the whole way. A simple first-week checklist - Sign in and read the eight department names. - Set money caps and the public-claims approval. - Set a conservative token cap. - Connect only the tools your first mission needs. - Run one small, real mission end to end. - Read your first Friday summary and act on what is waiting. By the end of the week you will know how a sentence becomes shipped work, where the approval points sit, and how to dial autonomy up or down. That is the whole operating loop.

How AOS remembers: operating memory

Operating memory is what makes AOS feel like a company that learns instead of a tool that forgets. Every decision, and the reasoning behind it, carries across missions. The next mission that touches the same territory inherits all of it. This is one of the biggest differences between AOS and a chatbot. What gets remembered - Decisions and their rationale. Not just what was decided, but why, so a later mission does not relitigate a settled question. - Acceptance bars that worked. Standards that produced good work become reusable patterns. - Facts about your business. Your positioning, your constraints, your non-goals, your tone, the tools you use. - What was rejected and why. Dead ends are recorded so the team does not wander back into them. Why this matters A chatbot answers a question and the context evaporates. The next session starts cold. AOS does the opposite. When you run a mission about pricing in March and another about pricing in June, the June mission already knows what you decided in March and why. You are not re-explaining your company every time. This compounds. The more missions you run, the sharper the institutional knowledge gets, and the less you have to repeat yourself. How it stays accurate Memory is only useful if it is trustworthy, so AOS treats it carefully: - It is tied to the evidence trail. A remembered decision points back to the mission and the reviewers that produced it, so you can always check the source. - It updates when you change your mind. Override a past decision and the new one becomes the standing answer. The old one stays in the record as history, not as current truth. - It respects your boundary. Operating memory holds decisions and reasoning. It is not a dumping ground for raw customer data, which stays governed by your access controls and the five approval points. How you use it Mostly, you do not have to do anything. Memory works in the background. But you can lean on it deliberately: - Ask a department to "use what we decided last quarter about onboarding" and it will pull the prior reasoning. - When you change direction, say so plainly. The new instruction becomes the standing decision. - If something in memory is stale, correct it once. The correction carries forward. The payoff The result is an operating system that gets to know your company. Early missions teach it your standards. Later missions move faster because the context is already there. Instead of starting from zero every time, you are always building on what you already settled. That is what turns 61 specialists into a company rather than a pile of disconnected answers.