What Marg is: your AI workforce

Most AI tools hand you a chat box and leave the thinking to you. Marg hands you a staffed company.

That difference runs through everything in this manual, so it is worth making concrete. A chat box is one generalist who answers whatever you type. Marg is a master orchestrator on top and a specialist team under it for every function a startup runs, with a named expert inside each team for the actual job: a pricing modeler, a funnel architect, a deal strategist, a churn analyst, and a bench that keeps growing as the product does.

Those teams cover the work a founder otherwise carries alone:

• Advisory handles strategy, fundraising, unit economics, and the decisions that keep founders up at night.
• Marketing covers content, growth, SEO, and social.
• Sales runs pipeline, deals, outbound, and playbooks.
• Product owns discovery, prioritization, and release planning.
• Design delivers brand, UI, and UX research.
• Project management keeps delivery, scope, and workflow on track.
• Support looks after customer operations and compliance.
• Specialized picks up documents, recruitment, audits, and the odd jobs every company has.
• Paid media manages search, social, and programmatic advertising.
• Research digs into deep, multi-source questions and checks every claim along the way.

A bench that wide would be a burden if you had to manage it. The point of Marg is that you never do.

You never pick the agent

A chat box makes you the manager: you decide who does what and stitch the pieces together. Marg takes that job. You describe what you need in plain language, and the orchestrator classifies the request, picks the team, assigns the specialists, and returns one finished deliverable.

Watch it work on a real question. You ask, "What is our current MRR and churn rate, and is the churn a problem?" The orchestrator routes this to advisory, where a revenue analyst reads your billing data and a churn specialist reads your product analytics, if those tools are connected. The two do not hand you two spreadsheets. They hand the orchestrator their findings, which come back to you as one short answer: the numbers, whether the churn is noise or a trend, and what to do about it.

That settles who did the work. The harder question is whether you can trust what they concluded, and Marg answers it on every deliverable.

Every answer tells you how much to trust it

Each deliverable carries an evidence grade, so you are never guessing about its footing. HIGH means the team built it from your live, connected data. MEDIUM means it worked from what you described in conversation. LOW means it had to assume more than it could verify, and it says so rather than dressing the gap up as fact.

That grade is also why Marg is useful on day one, before you have connected anything. With no integrations, every team works from what you tell it and the grades read MEDIUM, which is honest rather than limiting. Connect your tools later (Stripe, HubSpot, PostHog, and others) and the same requests start coming back HIGH, built from live numbers. Chapter 24 covers that step, and nothing in this manual is closed to you until you take it.

What this means in practice

You stop spending your attention on who should do the work and spend it on which work matters, which is the only question a founder should be answering anyway. The next chapter follows one request all the way through the system so the mechanics are clear, and chapter 3 gets you installed and lands your first deliverable in chat.