Why Building an AI-Native Company is So Hard? (And how to do it) - 6/18/2026

📆 Upcoming Hustle Fund Events

• 🇺🇸 June 24 – San Francisco, California – Finance Fundamentals featuring Shiyan Koh, GP of Hustle Fund

• 🇨🇦 July 8 – Montreal, Canada – Founder Friends Montreal

• 🇨🇦 July 8 – Toronto, Canada – Inside an AI-Native Company featuring Amir from Amir from Humblytics

Wanna volunteer? Join the events team in your city.

📰 Today's Edition: Why Building an AI-Native Company is So Hard? (And how to do it)

At a recent event we co-hosted with Legible.co (one of our portfolio companies), CEO Lawrence Coburn dropped some bombs about what it actually means to be an "AI native" company. Instead of the usual tech buzzwords, he shared 4 frameworks from brilliant thought leaders: 

1. The Death of Middle Management (Jack Dorsey & Roelof Botha)

It’s like Death of a Salesman (by Arthur Miller), but actually for middle management. 

Larger companies (including larger startups) are struggling with their org structure right now. Dorsey and Botha both believe the traditional org chart is collapsing. In the old days…hierarchies existed because information used to be expensive to route. 

The Roman army (that’s like super old) needed Centurions because a General couldn't talk directly to every soldier. But now when AI can instantly process, route, and act on information across an entire organization, those information bottlenecks disappear. Imagine if the Romans had had OpenClaw. Et tu Automation? 

Coinbase didn't just flatten their org, they nuked it. They eliminated all manager roles and collapsed from a traditional hierarchy to just five levels from the CEO. And we see this transition happening with many major tech companies.

In this new world, there are only three types of knowledge workers:

• Builders (the doers)

• Owners (the deciders) 

• Player-Coaches (the guides)

In this world, the human job shifts completely. Instead of managing information flow, humans become scouts at the edge—reading customer cues, detecting market shifts, bringing real-world intelligence back to the AI-powered core.

But in order for this to happen, your company has to become "legible" to machines. Every meeting recorded, every process documented, every communication channel structured. If AI can't read it, AI can't help with it. Water cooler conversations no longer determine company direction (until Granola buys Alhambra Water). 

2. Closed Loops and Token Maximization for the win (Diana Hu) 

Diana Hu believes that strategic direction remains human-driven and execution resides with AI. But this can only happen if most processes are closed-loop. What does this mean? 

Closed loop: AI initiates, acts, receives feedback, and iterates to completion autonomously (like code generation that tests and refines itself)

Open loop: AI generates output without systematic feedback integration; these are traditional one-way processes where AI generates output but doesn't get feedback about results. So, the AI doesn't know if its output was successful or useful. 

For example, you ask AI to write something. Then you copy-paste the results into a document, but the AI never learns whether that content was effective. This doesn’t help the AI get better.

Diana's insight is that AI-native companies should build as many closed loops as possible. Let AI run processes all the way through to completion rather than just generating outputs that humans handle manually. This allows the AI to learn, improve, and operate more autonomously.

The closed loop concept ties into her "token maxing" philosophy - maximize AI compute usage rather than human involvement, since closed loops let you get more value from your AI investment through learning. (This is why you had to cancel your team offsite to pay for more tokens.)

This should also lead to one of her other tenets of everything being queryable, where every process and artifact (AI outputs like files, images, etc) must be accessible to AI systems for maximum learning effectiveness.

Lastly, she believes that companies should have a no backfills policy (at least in the short term): When employees leave, you should absorb their workload with AI rather than hiring replacements. This "token maxing" approach prioritizes compute spend over people complexity.

3. How well is your company adopting AI? We should have levels. (Ann Miura-Ko)

Ann Miura-Ko sees AI adoption as a progressive scale with evaluation criteria almost like self-driving cars. Where is your company with regard to these things? 

• What can AI see in your company? (legibility and auditability)

• What can AI do autonomously?

• Who can extend the system? (Can non-engineers deploy code?)

• How is your organization evolving?

• Is your AI allowed to eat the free lunch and use the pool? (maybe that’s only at Google)

The goal in building an AI native company is to have self-driving operations. E.g. Level 5 means that there would be maximum autopilot functionality while keeping the C-suite in the loop for key decisions. 

Is your AI allowed to use the corporate swimming pool?

However, even if you get to Level 5, you have to be cognizant of the “feature factory”. Easy feature creation can lead to bloat without strong North Star guidance from management. In an age of cheap software, focus and restraint become premium assets in driving towards Level 5. 

4. Changing the Human Operating Model (Kieran Flanagan)

Lastly, Kieran emphasizes that AI adoption requires a change in human-operations, not just new tools. He believes that everyone needs to adopt a “Remote-first” workplace. This means: 

• Async-First: Remote-friendly, documentation-heavy workflows naturally lead to AI-readable operations. In this world, live meetings only happen for complex discussions and brainstorming (such as what the company lunch should be), not status updates. Everything else is written / documented that both AI and people can read. 

• Mission-Led Pods: Structure teams around workflows rather than departments.

• Naming Discipline: Consistent document-naming allows AI to find and interpret relevant data across the organization.

• Gating: But, not all company knowledge should be AI-accessible! Preserve some organizational "soul" and competitive advantages (and the company secret handshake). 

Transforming the workplace to be AI-native isn't just about adopting new technology. It involves redesigning the org structure as well as how work is done and communicated. 

Hearing these 4 frameworks, while straightforward, has made me realize we have a lot to clean up at Hustle Fund. 

Removing the hippocorn farting rainbows to make this blog post legible, 

Elizabeth, Chief Hippo

P.S. Lawrence and team have recently launched a consulting business called Legible.co to help companies progress towards becoming AI Native - feel free to reach out to them and tell them we sent you.

🎥 Watch This