Newsletter / Operations

How AI-native agencies use SOPs

60+ agencies, and almost all of them have a Google Drive folder called "SOPs" that nobody has opened in months. AI-native teams build something else entirely.

RI
Romans Ivanovs
Founder, Big Growth Group
Apr 2026 4 min read

I've sat inside 60+ agencies at this point. Almost all of them have a Google Drive folder called "SOPs" or "Processes" or "Playbooks" that nobody has opened in weeks.

They spend months documenting everything. Loom videos. Notion pages. Process maps that look beautiful in a screenshot and useless in practice. So why is the founder still the one who knows how things actually work, still the person the team defaults to when something breaks?

Here's what I've been thinking about a lot recently. There are two ways an agency can be structured. We show both to every client we work with.

Model A

The Dependency Model

The founder sits at the base, holding all the weight. Operations stacked on top. Delivery above that. People at the very top. It looks like a pyramid and the founder is the load-bearing wall. Pull them out and the whole stack collapses.

That's the business of almost every agency founder we talk to between $500K and $3M.

Model B

The Leverage Model

The founder moves to the top. Architect and owner, not load-bearer. Between them and the day-to-day sits a layer that doesn't exist in most agencies at this stage: data, AI, and systems. The founder sets direction. The machine handles execution.

That middle layer is where your documentation lives. But it's alive. It's queryable. It updates. It thinks with you.

A data layer is your delivery model, your client data, your process documentation, your decision frameworks, all connected, all searchable, all feeding into how work actually gets done. It updates when your processes change. It answers questions with context from your actual clients. And it doesn't rot in a Google Drive folder after three weeks.

How to solve this using AI

We've been installing this for clients using tools like OpenClaw and Obsidian, built on top of existing project management systems. Your team doesn't need to remember where the SOP lives. They ask a question and the system answers it, with context from your actual clients, your actual processes, your actual pricing.

When a process changes, and it always does, you update it in one place and it propagates everywhere. The bot updates the documentation. You just give the command. New hires don't need a two-week onboarding where someone walks them through 47 Loom videos. The knowledge system trains them as they work.

The founder and the seniors stop being the institutional memory of the business.

I had this conversation last week with a founder running $80K MRR, 65% margins, team of four. Beautiful numbers on paper. His words: "Delivery is not systemised enough. Not enough SOPs, not enough processes." So his instinct was to document everything.

He's identified the gap. What he's about to fill it with will be ignored in a month.

What he actually needs is proper infrastructure. A system where knowledge, processes, and client data all connect and talk to each other.

We, as founders, resist this because building infrastructure feels slow. Writing an SOP feels productive. You can knock one out in an afternoon and feel like you've made progress. Building a system where your knowledge, processes, and client data all talk to each other takes longer. But it's the thing that actually lets you step away.

That layer is data, AI, and systems.

The next step

Stuck in the Dependency Model?

The Founder Dependency Audit figures out what the Leverage Model looks like for your specific agency, and what to build first so the knowledge stops living in your head. We work through it together.

See where your agency depends on you

- Romans