Video Title: How to Build a Self-Improving Company with AI

Video Author: YC Root Access

Translation: Peggy

Editor's Note: In this latest YC batch talk, YC General Partner Tom Blomfield discusses not "how to use AI to improve employee efficiency," but a more fundamental question: When AI is no longer just a copilot, but can perceive, make decisions, call tools, receive feedback, and self-correct, how should the company itself be redesigned?

Tom's core judgment is that traditional companies still operate like "Roman legions": information moves up through layers of hierarchy, and commands are distributed down through management chains. But AI is breaking this organizational assumption. What truly matters is not making engineers write 20% more code, but extracting the business knowledge scattered across emails, Slack, meetings, documents, and human brains, and turning it into organizational context that AI can read, call upon, and iterate on.

In his view, future AI-native companies will be composed of a series of recursive, self-improving AI loops: the system perceives external changes from customer emails, support tickets, and product data, executes decisions through a rules layer, a tools layer, and quality gates, and finally automatically learns and corrects based on results. YC is already experimenting with similar mechanisms: agents don't just answer questions; they monitor which queries fail, determine whether new tools, databases, or indexes are needed, and automatically submit code, review, merge, and deploy. This means a company can continue to optimize itself even while the founder sleeps.

This also means AI's impact on companies will not stop at the tool level but will further change organizational structure. Tom proposes "burn tokens, not headcount" – the bottleneck for future startups may no longer be the number of employees, but token usage, the quality of business context, and the readability of organizational knowledge. The coordination functions performed by middle management will be largely replaced by AI, while ICs, directly responsible individuals, and human roles that handle high-risk judgments in the real world will become more important.

The most noteworthy point is not that AI makes companies more efficient, but that it is changing the very form of "company" itself. When software can be generated on the fly, processes can be automatically improved, and experience can be continuously precipitated into a company brain, what founders truly need to build may no longer be a team with clear hierarchy, but a smart system capable of continuous learning and self-optimization.

Below is the original text:

Rewriting How Things Work: Companies Should No Longer Operate Like Roman Legions

This part is somewhat based on a previous talk by Diana. That video from the weekend is already online and is fantastic. Also, Jack Dorsey posted some tweets about two or three weeks ago that I found very interesting, so I "borrowed" quite a few of his ideas and stuffed them into this talk.

This talk will be quite conceptual and high-level, mainly discussing how we should rethink building companies.

The design of a Roman legion was essentially about projecting power outward from the center of Rome, across two continents, all the way to Hadrian's Wall near Scotland. It relied on a nested hierarchy, with stable spans of control at each level. Each level had a clear person in charge, responsible for passing commands down and sending information back up.

If you look at most companies today, you'll find they still operate like a Roman legion: people are the channels through which information flows up and down. One thing that struck me from Jack Dorsey's tweets was our long-standing default assumption that hierarchical organization is the best way to organize economic value units. But I believe AI is essentially breaking this assumption.

A year ago, if you asked people what AI was good for, they usually talked about "productivity": like Copilot making engineers 20% more efficient, integrating Copilot into workflows to help teams deliver more software. But I think this is a problematic way of understanding it. It's like putting a more powerful engine into an old way of working. What's truly worth thinking about is not how to add an AI tool to an old organization, but to reimagine what a company itself is and how it should operate.

For example, what Garry was talking about just now – I truly believe he can now produce more code by himself than an entire engineering team. What really occupies my mind is how to extract the domain knowledge inside a company and define it as context, skill sets, or whatever you want to call it.

This domain knowledge, business knowledge, know-how – it's originally scattered in people's brains, Slack messages, emails, and Notion documents. This information collectively defines how your company operates. Once you can make this knowledge clear and readable, you can move from a hierarchical organization to an intelligent organization driven by AI-native software.

Making the Company Better While You Sleep: How AI Loops Automatically Discover, Fix, and Deploy

AI is not something attached to the side of a company. It's not just a tool for engineers to become more efficient. I believe we can reimagine the company as a set of recursive, self-improving AI loops. This point is crucial because once a company reaches this stage, it will continuously self-optimize even while you sleep.

For example.

Diana also mentioned this AI loop in her talk. First, there's a "sensor layer." This sounds fancy, but it can be quite simple: customer emails, support tickets, code changes, users canceling subscriptions, product telemetry data – these are all sensor data used to gather information from the outside world.

Then there's the policy or decision layer, which are the rules: what the AI can do, which things must request human permission, which operations must be logged. Next is the tools layer, similar to the skills and code Garry mentioned – essentially deterministic APIs, like querying a database, checking a calendar – the set of tools AI can call upon.

Then comes the quality gate, like deterministic checks Eva mentioned, security filters, and human review for high-risk items. Finally, there's the learning mechanism: the system interacts with the real world, finds where it doesn't work, and feeds that feedback back to the start of the loop.

If each step can run with no or minimal human intervention, the system gets better and better while you sleep.

I can give you some real examples we're running now. Initially, we made an agent you could ask questions to. It had some deterministic tools to query our database. For example, a simple question: When was the last time I did office hours with this company?

Later, it got smarter. For instance, if I'm doing office hours with a company that needs to know someone in the petrochemical industry, the system could query the database in different ways, combine it with RAG methods, find five relevant founders, and recommend them to you.

But this is still just a sidekick, an assistant agent. It's still last year's AI usage: AI making me, as a group partner, more efficient, boosting my productivity by 20% or 30%.

My real "aha moment" came when we added a monitoring agent on top of this system. It looks at every query made by every YC employee, determines which queries succeeded and which failed. Then it asks: Why did it fail? What would make this query succeed? Do we need a new deterministic tool? Do we need to update the skills file? Do we need a new database? Do we need a new index?

These things now literally happen automatically at night. It writes code, submits a merge request to YC's codebase, has another agent review it, then merges and deploys it. So the next day, when a human asks the same question, the query succeeds.

For me, that was the pivotal moment. It's not just about making a human 20% or 30% more valuable. It's the AI completing the loop itself and finding ways to self-improve.

I believe if you can identify which parts of your company can operate like this, and minimize the human role in execution and supervision, you can put tokens into that problem, and the company itself will continuously improve.

There are many other examples. For instance, with product analytics data, you can have an agent analyze it to find the biggest friction points in the sales funnel. It can research best practices, set up an A/B test, run it for a week, pick the best-performing version, and deploy it live.

This happens over and over. Your product would have a self-optimizing product loop.

Customer support is the same. Customer suggestions keep coming in. You can use an agent for triage. This agent acts somewhat like your Chief Product Officer and Chief Technology Officer, making judgments: we don't want this suggestion, discard it; but this one fits our roadmap and can be done tonight. So it writes the code, deploys it live, directly delivers it to the customer, with no human intervention needed throughout the process.

So if you can view every part of your company as a self-improving, recursive AI loop, it becomes something completely different from a "Roman legion" style hierarchical company.

Burn Tokens, Not Headcount: AI-Native Companies Will Reshape Organizational Structure

So, if you want to do this, what does it imply?

The first point is: consume tokens, not stack headcount. We're already seeing that by Demo Day, many companies' revenue per person is about 5 times higher than 18 months ago. I think this trend will continue into Series A and B rounds. Soon, what truly constrains you won't be the number of employees, but token usage.

The crudest way now is to measure everyone's token usage. Of course, this metric is stupid when taken to extremes and easily gamified. But directionally, I think it's right. We're in an exploratory phase of "what's possible," so everyone should experiment maximally, pushing this crazy new intelligence to its limits.

Once you turn it into a leaderboard and link promotion or firing to this metric, it will be gamed and distorted. But directionally, figuring out who in the organization is maximizing token usage and who isn't, is indeed a way to judge where you should spend your time with employees.

I believe middle management is over. At least for these coordination problems, I don't think we need middle management anymore; AI should handle this.

For me, there are two important roles in the future. Jack Dorsey mentioned three, but I didn't like the third, so I removed it. I think the two truly important roles are: everyone must be an IC, an Individual Contributor, a builder, a doer. And the key is having a directly responsible individual. For anything to move forward, you need a clearly named person responsible, not a committee or a group.

I think companies can be built entirely on ICs. Middle management is truly over. And building a self-improving company is this vision.

By the way, I think everyone is still at the frontier of this. I'm very curious to see how far you've all gotten. It feels like everyone is still exploring the boundaries. I'm not sure if anyone has yet built a truly self-improving company across every function. Maybe I'm wrong, and you can prove me wrong.

If I were you, what would I do first?

The first very important thing is to make the entire organization readable and understandable for AI. What does that mean? It means you must record everything.

Simply put, all our partners' emails now – if you email a YC partner, that email goes into the YC database. Every Slack message, every DM, every office hours session – we've been recording all of them for the past three or four months. Everything that happens, if it's recorded, it happened for AI; if it's not recorded, it didn't happen for your intelligent system.

Just a moment ago, I was chatting with some founders here, discussing a lot of good content about their companies. Every time I talked, I thought to myself, I really should be recording this conversation. Because someone needed an introduction, and now I can't even remember who the introduction was for. I agreed to it, then told them to email me later because I knew I'd forget – I still had 20 more people to talk to.

So, this might require using phones, recording devices, smart glasses, or putting microphones in every room. In short, everything needs to be recorded so AI can read it.

Then, as Garry said, you also need speaker diarization and summarization. You can't just stuff 100,000 hours of audio into a context window. You have to process it, aggregate, compress, and distill it into important parts, then leave some clues for the AI.

For example: How many of you have read the YC user manual? I hope at least everyone in this room has opened it once. It's okay. Most of that manual was written five to ten years ago and is a bit outdated.

Last weekend, Harsh suddenly had an idea: since we've accumulated about 2000 hours of office hours recordings over the past three months, why not regenerate a version of the user manual?

So you give the system a set of instructions: process, compress, and synthesize the recordings, classify them by topics like fundraising, hiring, co-founder disputes, and then have it write a new user manual. By the end of the weekend, he had generated a 150-page user manual that was clearly better than the current version.

More importantly, now we can update it every month. Our user manual has become a self-improving system. Every new piece of advice is compared against the existing manual, either absorbed or discarded. Thus, the user manual becomes a continuously updated living brain, holding the weekly advice we give to founders.

Of course, it doesn't stop at the user manual level. You can feed this as context into an AI agent. Suddenly, you can ask a super-intelligent AI a question and receive the combined wisdom of 16 YC partners. But this requires that knowledge to be readable by AI. So you must record everything.

The second point is similar: if something can create a self-improving artifact that can be read by AI, keep it; if it can't, discard it.

The third point is that every function should be able to generate its own software. In the past, we might have said "dashboard," but now it's more than just dashboards – it's on-demand generated software. Codex 5.5 is now good enough that for most simple internal software and dashboards, you can generate them to a very high quality in one go. I tried it over the weekend with some of our internal stuff, and the results were truly unbelievable.

So, all internal operations teams should sit on top of this layer: possessing an intelligent understanding of the business, and then generating their own dashboards and workflows.

And I would treat this software as completely disposable. What should be very carefully preserved is the data. As Garry said, he saves all his emails as Markdown, never discarding anything. But the software itself is ephemeral, temporary. You can generate it, and you can regenerate it.

What is truly valuable is the understanding of the business in the human mind: how this function operates, how we run a YC event, etc. As for the software actually used to execute the event, you can generate it for this event, use it, and discard it. A month or two later, the model gets smarter, you throw away the old software, give it the original instructions again, and generate a new version of the software.

So I believe the valuable things are the business context and skills. The software built on top of them is ephemeral.

So, in this world, what is the role of humans?

I think what we're discussing is essentially a "company brain." I know many people in this room are working on something similar. The middle part – all your data, all emails, DMs, skills, know-how – that's the company brain.

Humans are on the edges of this brain, responsible for interacting with the real world. That means humans are the point where this intelligent system touches reality. Humans can enter scenarios the model can't yet access, like meetings, or novel, complex situations. I was going to use phone calls as an example, but it's actually quite easy for AI to enter phone calls now.

More typical are unfamiliar situations, ethical judgments, and high-stakes moments. For example, a founder comes to us saying they are considering parting ways with their co-founder. In these truly high-risk, emotionally intense moments, you still want a human present.

This is the human's place. For many of your companies, sales conversations are similar. For the next 20 years, I believe sales will still require a human in the room.

So I believe humans will live on the periphery of the company brain, responsible for bringing intelligence into the real world.

I'm already over time, and the host is probably about to pull me off stage. I'll leave you with one final question: If you were to re-found your own company today, would you design it in this form from the start?

Most of you have companies that are small enough that you could absolutely do this. So I think you have no excuses. And I know there are several people in the room right now who are dismantling and rebuilding their own companies.

I'll stop here and pass it over to Pete. Thank you.

Video Link