At a YC dinner last week, a founder casually mentioned they’re not hiring sales anymore.
“So.. how do you do outbound?”
“14 AI agents.”
I thought they were messing with me. Then they showed me the dashboard. Agents handling outbound sourcing, onboarding sequences, support triage, feedback loops. Real pipeline. Real revenue.
This wasn’t a side experiment. It was the whole operation. And they’re not the only ones doing it.
The pattern I’m seeing
The old YC advice was “don’t hire until it hurts.” The new version I keep hearing: don’t hire, automate.
Founders are peeling off the low-value work — the stuff that used to be someone’s entire job — and handing it to agents. Not replacing their core team. Not firing anyone. Just never making that next hire in the first place.
The metric that keeps coming up is revenue per employee. Levelsio is the extreme version — one person, no employees, real revenue. But the principle scales: the best founders I know are asking “can I automate this?” before “who should I hire for this?”
I’m seeing this across YC batches, across stages. The founders moving fastest right now aren’t the ones with the biggest teams. They’re the ones who figured out which work doesn’t need a human.
Manual first, automate second
Here’s the part most people skip.
Automating before you understand the job is a mistake. If you can’t crack it yourself, you can’t hire someone to crack it either. Same thing applies to agents.
I’ve said this before about outbound: don’t automate your first outbound. Do it manually. Feel the responses. See what clicks. Open rates. Reply rates. What messaging actually lands. You need that intuition before you hand it off — to a person or a machine.
Same principle as “do things that don’t scale.” The automation comes after the reps. After you know what good looks like. After you can tell whether the automated version is actually working.
I posted about the 14-agent story and it went everywhere. Three days later I had to post the correction: automation is great, but not on day one. The hype was already outrunning the nuance.
The sequence matters. Manual first. Nail it. Then automate. Then maybe hire someone to scale what’s already working.
What gets automated first
Not everything. Not the hard stuff. The low-leverage stuff.
The pattern: founders automating work that’s necessary but doesn’t require taste or judgment.
- Outbound sourcing — finding leads, enriching contact info, sending initial sequences
- Onboarding flows — welcome emails, setup guides, drip campaigns
- Support triage — routing tickets, answering common questions, surfacing edge cases for humans
- Feedback collection — aggregating signals from support, reviews, and usage data
These are functions where agents can run a playbook. You set the metrics — open rates, reply rates, resolution times, demo conversion — and you measure them the same way you’d measure a junior hire. Numbers don’t hold up? Fix the playbook or pull them off the job.
You’re not replacing people. You’re automating the work that would’ve required a hire you haven’t made yet. When the hire eventually comes, that person is doing higher-leverage work from day one.
What stays with the founder
Sales is a founder function. Not just closing — learning. Every sales conversation teaches you what the market actually wants. What language resonates. Where the objections live.
You can’t delegate that to agents. Not because the technology isn’t there, but because the learning is the point. Knowing how people react to your product in real time is one of the few things that compounds for founders. The moment you lose that signal, you’re guessing.
I’ve seen founders outsource sales too early and lose the thread. They stopped hearing what the market was telling them. Roadmap drifted. Messaging went stale. Not because they hired badly, but because they cut themselves off from the source.
The founder role is shifting. Less “doing the work” and more “knowing what good looks like.” You set the bar. You watch the metrics. You step in when things feel off. But you can’t do any of that if you never did the work yourself.
I’m building around this exact idea — automating work I’ve already done manually, piece by piece. Less “org redesign,” more tinkering. Figuring out what agents can actually own vs. what still needs me. Still early.