Hi all,
You’re receiving this message because you’re part of a group of friends and colleagues who supported me in starting Revi Systems earlier this year. I had to take a hiatus for a bit due to some family matters, but I’m back now and full steam ahead. What started as a group bcc email has turned into this newsletter, Exponential Operator. I’ll use this as a space to write about the ideas I’m exploring every day.
If you no longer wish to receive these emails, no hard feelings! Just click unsubscribe at the bottom. If you find the post below interesting, please forward this email to someone else who would enjoy reading it.
-Sawyer
I start every sales call the same way by asking a question - “how are you guys using AI at work today?”
More often than not I receive a laundry list of tools, automations, and agents they’ve been using to varying degrees of success. My job is to help turn this jumble of subscriptions into an integrated operating system.
For a while, I asked a follow up a few minutes later - “how do you hope the fully AI-enabled version of your organization will operate?”
The folks I asked would usually say something outcome oriented like scaling up without adding headcount, or automating away busywork. These are good goals but don’t say much about how the company actually wants to operate with AI. How should humans and AI agents collaborate? Who decides what AI works on? How do we know if things are going well?
Everyone has a clear idea of what results they want, but the operating model is usually vague.
On one hand, it’s my job to design the AI operations system, not theirs. But I actually think it’s a problem that so few strategic operators can envision what a great AI-forward company looks like. It’s not just your current org, but with fewer people and more efficiency - whatever that means. The entire structure and rhythm of business are different. Without a clear north star, it’s nearly impossible to build and execute a strategy to get to the next level.
So when I came across this article published by Floodgate Fund’s Ann Miura-Ko, I knew that I had to share it. It’s a simple piece - I recommend you read it. She lays out a 5-level framework to assess how AI-forward a company is - essentially, to what extent does a company embed AI into its operations and make use of the technology’s most advanced capabilities?
She describes 5 levels of AI adoption from L0, “AI as theater”, to an aspirational-but-so-far-unachieved L5 “Virtually self-driving organization”. The framework breaks down how AI-forward a company is along 4 axes:
What can AI see?
What can AI do?
Who can extend the system?
How has the organization changed?
At L2, where most companies I’ve worked with are: AI has access to shared context within team boundaries, agents automate specific functional workflows, non-engineers benefit from but don’t build their own tools, and the organization is a more efficient (but structurally unchanged) version of itself.
That’s not a bad place to be, but I find many of the most interesting and transformative interventions within organizations I’d characterize as further along at stages 3 or 4.
Inside the companies operating with AI at the highest level today,
Organizational data and context is stored in structured, AI-accessible systems of record
Agents are able to work proactively and cross-functionally in a wide range of domains
Anyone can ship internal software tools
Role hierarchies collapse and rhythms of business accelerate
This framework isn’t prescriptive. There’s no roadmap that tells you how to get from L1 to L2, L2 to L3, etc. What it does provide, however, is that north star to indicate a general direction organizations should try to make progress towards.
It’s helpful to also realize that being AI-forward isn’t an all-or-nothing proposition. You don’t have to do everything at once, and it’s very likely some teams will start off well ahead of others.
Back when we started Aomni in 2023, L2 was considered the cutting edge. By the time we built Duet in late 2025, we had broken into L4 “Compounding operating system”. Partially this is due to the tech becoming more capable.
Agentic harnesses in particular have been a catalyst for crossing the gap from L2 to L3. But we’ve also made a lot of progress in understanding where AI provides the most leverage within an organization (and what it’s really bad at).
I fully expect that by the end of this year, we’ll see our first successful L5 companies. Not just tiny startups - those exist already - but actually meaningfully large organizations. Larger organizations are naturally going to move slower due to real cultural, capability, and cost constraints. Nevertheless, there’s only one direction worth moving in - and that’s up the AI adoption ladder.
That’s all for today. Thanks for reading!
江湖再见🫡
-Sawyer