3. LeanScale AI OS Overview
Note: The video covers material not in the guide below — please watch in full.
Training Guide
LeanScale AI OS Overview
A 15-page scoping document for sales territory design. A diagnostic on a startup's entire go-to-market. A 50-page playbook that the whole team runs on. None of it written by a human — all built by AI agents. Twelve weeks ago, we were using ChatGPT like everyone else. Now we have an entire agentic operating system.
It's Just Folders and Files
Here's what makes all of this less intimidating than it sounds: the entire system is built on folders and files.
When you hear "agent," just think of a folder with a set of instructions listed as a file. When you hear "repository" — or "repo" — that's just a folder. A transcript warehouse repo? That's a transcript warehouse folder. Same thing.
Four folders make up the backbone of the LeanScale AI OS:
- Transcript Warehouse — where every call transcript lands automatically
- Customer Warehouse — where all enriched customer intel lives
- GTM Library — all of our go-to-market playbooks
- LeanScale Context — everything about us as a company (brand guidelines, customer avatars, pain points)
(Instead of just describing each folder, let's walk through how the agents actually move through them)
The Agent Handoff Chain
Every call transcript — from Fireflies, Gong, Chorus, Fathom, whatever you use — automatically lands in the Transcript Warehouse. An agent watches that folder. Every time a new transcript drops in, it reads it, annotates it, and routes it to the right customer folder.
This isn't a simple automation. The agent is tagging context: Is this a discovery call? A renewal conversation? Did the customer mention a competitor? Push back on pricing? Show urgency? All of that gets annotated.
Then the Customer Warehouse agent picks up the handoff. It reads the notes the transcript agent left and enriches the customer's context files — tech stack, stakeholder map, priorities, pain points. These files aren't for humans to read. They're for the next agent in the chain.
See how this works? It's literally a handoff from one agent to the next — like team members passing work to each other.
Why Not Just Use the CRM?
You might be thinking: doesn't all this already live in the CRM? Here's the shift. The agent platform is tool agnostic. Your CRM isn't. Notion won't connect to your CRM. Salesforce's Agentforce isn't tool agnostic enough to pull from every source you need.
Through MCPs — model context protocol, basically plug-ins that let the AI connect to third-party tools — your agent platform can pull data from HubSpot, Salesforce, Google Drive, Snowflake, Intercom. One central workspace that connects everything.
That's what's been missing in AI. Not smarter models — one workspace that ties it all together.
The Context Memory Layer
Here's what this entire ecosystem actually is: a context memory layer for the agent platform. Every agent has memory of what's going on with each customer, with your company, with your playbooks, with your reps.
And because agents can write into files — not just read them — the system compounds. Every implementation teaches the agents something new. Every customer conversation makes the playbooks better. The agents are making the system better over time, so every interaction improves the next one.
(Let's see what this looks like in practice)
A Real Example
Say you need to do a sales territory design for a customer. You give the agent three things: the customer's enriched context file, the sales territory design playbook from your GTM Library, and the kickoff call transcript. Hit go.
The agent reads all of it — the playbook refined across dozens of implementations, the enriched customer context other agents prepared, and the specific needs from the kickoff call. Out comes a full scoping document with executive summary, current state analysis, data requirements, and business objectives.
The output quality is only as good as the input and context you give it. That's the whole point of the earlier agents doing enrichment — they're prepping data quality so the downstream agents can deliver.
And it doesn't stop at one playbook. Every function in go-to-market has one. Right-click, copy path, give it to the agent, pull in customer context, and the agent gets to work.
The Shift
This isn't about making AI smarter. It's about making sure the data layer and the context across your entire organization is already organized — so the agents can do real work. And the cool part? You use agents to organize the context for other agents.
Twelve weeks in, the compounding is real. It's still early. But this is the operating system we're building on.
Comment in Slack
Post your answer in your onboarding channel.
Before this training, what did you think an "AI operating system" meant — and how is it different from what you just saw?