0:09
Hi, I'm Diana and I'm a partner at YC.
Search this video
0:12
Over the past few months, it's become clear to me that AI is not just going to change how quickly software gets built or what workflows get automated.
0:19
It's going to fundamentally change the way startups should be run from what roles it will exist to what products are possible to build.
0:27
In this episode, I'm going to discuss how founders should think about building an AI native company.
0:33
What roles their team should have and what concrete internal practices they can adopt right now to move much faster.
0:39
Currently, most people talk about AI in terms of productivity.
0:44
They'll talk at length about how it can make engineers more productive or say we need to add copilot to existing workflows and ship more features.
0:50
This framing misses the shift we're currently seeing which is less about productivity boost than entirely new capabilities.
1:01
The right person with AI tools can now build features that used to require an entire team or were just impossible.
1:08
Thinking about AI in terms of new capabilities has several implications for how founders should run their companies.
1:14
At a high level, the way to think about AI is that it should not be a tool your company just uses.
1:19
It should be the operating system your company runs on.
1:25
Every workflow, every decision, and every process should flow through an intelligent layer that is constantly learning and improving.
1:32
What this means concretely is every important process in your company should be captured by an intelligent closed loop.
1:39
A closed loop captures information, feeds it back into an intelligent systems, and improves the process over time.
1:45
If you've ever studied control systems, you'll be familiar with the difference between an open loop and a closed loop system.
1:51
Open loops are control systems without feedback loops.
1:55
In the old world, companies basically ran as open loops.
2:00
You made a decision, executed it, and didn't always systematically measure the outcome, and adjust the process.
2:07
Open loops are inherently lossy.
2:07
A closed loop, on the other hand, is self-regulating.
2:13
It continuously monitors its output and adjust its process to better meet the stated goal.
2:18
Closed loops are extremely powerful for correctness and stability.
2:22
With self-improving agents, your company should run as a closed loop.
2:24
To build these closed loops, you will need to make your entire company queryable.
2:28
In other words, the whole organization should be legible to AI.
2:33
Every important action should produce an artifact that the intelligence at the center of the company can learn from and use to self-improve.
2:44
This means recording your meetings with an AI note-taker, minimizing DMs and emails, and embedding agents throughout communication of all channels.
2:50
It also means building custom dashboards with everything in the company: revenue, sales, engineering, hiring, ops, everything.
2:59
Here's a concrete example of how it could work.
3:01
Take engineering management and sprint planning.
3:03
If you have an agent that has access to your linear tickets, all your Slack engineering channels, all customer feedback from emails or tools like Pylon and GitHub, high-level plans in a Notion or Google Doc, sales calls and recordings from daily stand-ups, then the agent can analyze what was actually shipped in your previous sprint and how well they met customers' needs for real.
3:29
From there, you can go a step further.
3:31
With full visibility into what shipped, what worked, and what didn't, agents can start looking ahead.
3:36
They can propose sprint plans for engineers that are way more predictable and accurate and on track.
3:42
The days of eng manager status roll-ups that are super lossy are gone.
3:47
Having managed engineering teams myself, and now seeing this across multiple YC companies. This is a game changer.
3:52
What used to require constant coordination becomes legible and queryable by default.
4:00
I've seen teams that do this cut their engineering sprint time in half and get close to 10x more done in that time.
4:08
The overarching principle here is that to get their full capabilities, you need to provide models with as much context as you would provide an employee.
4:16
When you do this, your company stops operating as an open loop where information is fragmented and manually interpreted.
4:22
It becomes instead a closed-loop system.
4:25
Status, decisions, and outcomes are continuously captured and fed back into this intelligence layer.
4:32
The result is a system that always has an up-to-date view of what's actually happening.
4:37
There's also a new paradigm emerging for how the highest velocity companies build product. AI software factories.
4:45
If you're familiar with the test-driven development or TDD, this is the next evolution of that.
4:51
With software factories, humans write a spec and a set of tests that define success.
5:00
And then AI agents generate the implementation and code and iterate until the test pass.
5:09
The human defines what to build and judges the output.
5:10
The actual code is the agent's job.
5:13
Some companies have already pushed this to the point where the repos contain no hand-written code, just specs and test harnesses.
5:24
Strong DM's EI team is an example of how to do this.
5:27
Their end goal was a system that essentially eliminated the need for a human to write or review code.
5:31
And so, they built their own software factory where specs and a scenario-based validations drive agents to write, test, and iterate on code until it meets a probabilistic satisfaction threshold. And it works.
5:46
This is how you achieve the 1000 X engineer that Steve Yegge talked about by surrounding a single engineer with a system of agents that enable them to build things they would have never been able to build before.
6:02
The era of the 1000 or even 10,000 X engineer is here.
6:06
One implication of building your company this way with AI loops everywhere, a queryable organization and software factories, is that the classic management hierarchy no longer makes sense.
6:16
In the old world, you needed middle managers and coordinators to route information inefficiently up and down an organization.
6:27
In the new world, the intelligence layer serves that purpose.
6:29
If your company is queryable, artifact-rich, and legible to an AI, you should have almost no human middleware.
6:39
This matters because your company's velocity is only as fast as its information flow.
6:46
Every layer of human routing you can remove is a direct speed gain.
6:52
A great example is what Jack Dorsey is doing over at Block.
6:54
After going deep on the tools, he's come to the same conclusion many have.
7:01
This is about more than just incremental productivity gains.
7:04
His view is that if you keep the same org chart and management structure, you'd miss the shift entirely.
7:10
The company itself has to be rebuilt as an intelligence layer with humans at the edge guiding it rather than routing information through it.
7:18
Going forward, Jack suggests every company will have three employee archetypes.
7:23
The first is the individual contributor or IC.
7:25
Basically, the builder operator.
7:27
This is someone who directly makes and runs things.
7:29
In an AI-native company, this is not limited to engineers. Everyone builds.
7:37
Eng, ops, support, sales.
7:40
Everyone comes to meetings with working prototypes, not pitch decks.
7:42
Second, is the DRI, the directly responsible individual.
7:47
Focus on strategy and customer outcomes.
7:50
This is not a classic manager, is the person with a clear responsibility for the result.
7:54
One person, one outcome, no hiding.
7:57
The third is the AI founder type.
7:59
This person still builds, still coaches and leads by example.
8:05
If you're the founder, this needs to be you at the forefront.
8:10
Showing your team what massive capability gains look like, not delegating your AI strategy to someone else.
8:16
With this structure, companies will be able to get outsized results with much smaller teams.
8:21
Maximizing token usage, not headcount, will be the critical shift.
8:26
The best companies will be the ones that are token maxing.
8:28
Think of the trade-off this way.
8:30
One person with AI tools can be the equivalent of what used to take a large engineering team at a pre-AI company.
8:37
That means dramatically leaner engineering, design, HR, and admin teams.
8:43
And so, you should be willing to run an uncomfortably high API bill.
8:48
Because it's replacing what would have taken a far more expensive and inflated headcount.
8:53
But, don't just take my word for any of this.
8:59
You cannot outsource your conviction on the power of these tools.
9:01
You need to develop it yourself by actually sitting with coding agents and using them until you start to break your own priors about what is now possible to build.
9:10
If you are an early-stage founder, you have a huge advantage in getting ahead on this.
9:17
You don't have legacy systems, entrenched org charts, or thousands of people to retrain.
9:20
You are small enough to build your company right from day one.
9:25
The opposite is the case for existing companies.
9:27
They have to maintain and grow a live product while unwinding years of standard operating procedures and core assumptions about how software gets built.
9:37
Some companies can achieve this by spinning up small internal skunkworks teams that can build AI native systems from scratch, separate from the core business.
9:45
Mutiny is a great example of this.
9:47
But, for most, every change to their core processes risk breaking something that already works.
9:54
So, by their nature, these large companies will have a much harder time going AI native.
10:00
Startups don't have that constraint, and that's a major edge to take advantage of.
10:07
You can design your systems, workflows, and culture around AI from the start, and as a result, operate 1,000 times faster than the incumbents.