<100 subscribers
Share Dialog
Share Dialog



I haven’t written a line of code in a year. I’ve shipped 25 projects anyway.
When I heard Jason Lemkin on Lenny’s Podcast, I realized we’re running the same playbook in different domains.
All started when Lemkin’s SaaStr lost three sales reps during their biggest conference of the year. His response?
“I’m done hiring humans in sales.”
Eight months later, his go-to-market team runs on 1.2 humans and 20 AI agents. Same revenue. One-tenth the headcount.
Lemkin crossed from experiment to reality. And it contradicts most of what we hear about AI in enterprise sales.
Listen to the episode below. It’s a template for implementation.

We replaced our sales team with 20 AI agents—here’s what happened | Jason Lemkin (SaaStr)
18 days ago · 103 likes · 1 comment · Lenny Rachitsky
Here’s what actually happens when you replace humans with AI agents.
Walk into Lemkin’s office and you’ll find ten empty desks. Each one has a nameplate for an AI agent instead of a human rep.
His agents process real sponsorships and event tickets for SaaStr. Actual money, real transactions, running right now.
When Lemkin says “same results,” he’s not promising transformation or magical 10x gains. He’s describing the new baseline.
The common assumption is AI must deliver 10x improvements. Lemkin says that’s wrong.
Match human performance at one-tenth the cost, and the decision makes itself.
Lemkin’s prediction:
“The classic SDR junior kid that is hired out of college to send emails... we don’t need them... They should be extinct next year.”
Junior reps with three months’ experience barely know the product anyway. AI handles inbound lead qualification 24/7 with instant responses.
The repetitive email outreach that defined SDR work for a decade? Agents do it better at scale.
But here’s where it gets interesting. Lemkin sees SDRs disappearing, then redefined.
Future SDRs manage “10 agents not 10 people” and earn $250,000 a year for orchestration expertise.
The SDR role defined B2B sales for a decade. Lemkin’s already building the next version—one where humans direct agents instead of doing the work agents handle better.
Lemkin’s take:
“AI isn’t coming for elite performers. It’s coming for the mid-pack and the mediocre.”
An agent trained on company documentation knows the product better than a rep three months in who still can’t explain it clearly.
As he put it:
“My god I never read everyone’s emails before. These are the worst emails that I’ve ever read.”
The A-player sales rep becomes far more valuable when armed with AI agents. Companies that understand this will compete fiercely for top talent. Those that don’t will automate their way to average.
The middle disappears. The top becomes indispensable.
I see this in my own work. Remember those 25 projects I mentioned? 10 were production systems in the last 3 months alone.
As a senior dev, I manage Claude Code agents the way Lemkin manages his sales agents.
Meanwhile, seasoned developers around me still say: “I sometimes chat with Copilot, but I prefer to write my own code.”
That’s the midpack talking. The same pattern shows up everywhere.
Lemkin again:
“Professionals are panicked about AI because they’ve never actually used it. They strategize, but they don’t build.”
Most AI conversations happen in strategy documents. Lemkin says that’s noise.
The actual skill employers will pay for? Implementation.
His advice for becoming “hyper employable” is direct: pick a tool and train it yourself.
Here’s the process:
Feed your agents the company docs
Answer their training questions
Spend an hour each morning reviewing their mistakes
Because they will make mistakes. Every single day.
That daily review cycle builds rare expertise. It teaches you what these tools can and can’t do. Their actual limits, not the marketing promises.
The person who implements gains a skill that’s suddenly in demand. Vendor evaluation becomes obvious. Deployment becomes practical instead of theoretical.
Implementation is where value lives.
Lemkin chose his vendor for a simple reason:
“They offered to help us the most.”
Their tech wasn’t necessarily better. They just committed to doing the work alongside his team.
He emphasizes that agents will fail without proper training. What matters is partnership—what he calls a “forward deployed engineer” who ensures your agent actually works.
His buying advice is direct:
Call vendors and ask specifically who will help with deployment
Get concrete commitments, not vague promises
Judge vendors by the engineers they put in your office, not the bullet points on their website
These aren’t plug-and-play SaaS tools. Their value comes from implementation, not feature lists.
Lemkin found the threshold: AI matching human performance at dramatically lower cost.
The replacement narrative misses the point. The real shift is who orchestrates the agents—and who gets orchestrated out.
The GTM playbook that defined the 2020s is obsolete. The question is how quickly you can rebuild around agents.
This week’s challenge:
Pick one repetitive sales task. One tedious workflow you hate. Build an agent for it.
Skip the months of vendor research. Forget the strategy documents.
Pick a tool
Feed it your docs
Spend three days training it
The agents are working 24/7 right now. Your competitors are deploying them. Lemkin already crossed from speculation to implementation.
Someone’s going to build the 1.2-human GTM team in your market.
Let it be you.
How do you actually coordinate multiple agents without them stepping on each other? That’s a deeper topic—one I’ll cover next time.
Thanks for reading Orchestrated Code! Subscribe for free to receive new posts and support my work.
Subscribe

Available in the Substack app and on web
Thanks for reading Orchestrated Code! Subscribe for free to receive new posts and support my work.
Subscribe

I haven’t written a line of code in a year. I’ve shipped 25 projects anyway.
When I heard Jason Lemkin on Lenny’s Podcast, I realized we’re running the same playbook in different domains.
All started when Lemkin’s SaaStr lost three sales reps during their biggest conference of the year. His response?
“I’m done hiring humans in sales.”
Eight months later, his go-to-market team runs on 1.2 humans and 20 AI agents. Same revenue. One-tenth the headcount.
Lemkin crossed from experiment to reality. And it contradicts most of what we hear about AI in enterprise sales.
Listen to the episode below. It’s a template for implementation.

We replaced our sales team with 20 AI agents—here’s what happened | Jason Lemkin (SaaStr)
18 days ago · 103 likes · 1 comment · Lenny Rachitsky
Here’s what actually happens when you replace humans with AI agents.
Walk into Lemkin’s office and you’ll find ten empty desks. Each one has a nameplate for an AI agent instead of a human rep.
His agents process real sponsorships and event tickets for SaaStr. Actual money, real transactions, running right now.
When Lemkin says “same results,” he’s not promising transformation or magical 10x gains. He’s describing the new baseline.
The common assumption is AI must deliver 10x improvements. Lemkin says that’s wrong.
Match human performance at one-tenth the cost, and the decision makes itself.
Lemkin’s prediction:
“The classic SDR junior kid that is hired out of college to send emails... we don’t need them... They should be extinct next year.”
Junior reps with three months’ experience barely know the product anyway. AI handles inbound lead qualification 24/7 with instant responses.
The repetitive email outreach that defined SDR work for a decade? Agents do it better at scale.
But here’s where it gets interesting. Lemkin sees SDRs disappearing, then redefined.
Future SDRs manage “10 agents not 10 people” and earn $250,000 a year for orchestration expertise.
The SDR role defined B2B sales for a decade. Lemkin’s already building the next version—one where humans direct agents instead of doing the work agents handle better.
Lemkin’s take:
“AI isn’t coming for elite performers. It’s coming for the mid-pack and the mediocre.”
An agent trained on company documentation knows the product better than a rep three months in who still can’t explain it clearly.
As he put it:
“My god I never read everyone’s emails before. These are the worst emails that I’ve ever read.”
The A-player sales rep becomes far more valuable when armed with AI agents. Companies that understand this will compete fiercely for top talent. Those that don’t will automate their way to average.
The middle disappears. The top becomes indispensable.
I see this in my own work. Remember those 25 projects I mentioned? 10 were production systems in the last 3 months alone.
As a senior dev, I manage Claude Code agents the way Lemkin manages his sales agents.
Meanwhile, seasoned developers around me still say: “I sometimes chat with Copilot, but I prefer to write my own code.”
That’s the midpack talking. The same pattern shows up everywhere.
Lemkin again:
“Professionals are panicked about AI because they’ve never actually used it. They strategize, but they don’t build.”
Most AI conversations happen in strategy documents. Lemkin says that’s noise.
The actual skill employers will pay for? Implementation.
His advice for becoming “hyper employable” is direct: pick a tool and train it yourself.
Here’s the process:
Feed your agents the company docs
Answer their training questions
Spend an hour each morning reviewing their mistakes
Because they will make mistakes. Every single day.
That daily review cycle builds rare expertise. It teaches you what these tools can and can’t do. Their actual limits, not the marketing promises.
The person who implements gains a skill that’s suddenly in demand. Vendor evaluation becomes obvious. Deployment becomes practical instead of theoretical.
Implementation is where value lives.
Lemkin chose his vendor for a simple reason:
“They offered to help us the most.”
Their tech wasn’t necessarily better. They just committed to doing the work alongside his team.
He emphasizes that agents will fail without proper training. What matters is partnership—what he calls a “forward deployed engineer” who ensures your agent actually works.
His buying advice is direct:
Call vendors and ask specifically who will help with deployment
Get concrete commitments, not vague promises
Judge vendors by the engineers they put in your office, not the bullet points on their website
These aren’t plug-and-play SaaS tools. Their value comes from implementation, not feature lists.
Lemkin found the threshold: AI matching human performance at dramatically lower cost.
The replacement narrative misses the point. The real shift is who orchestrates the agents—and who gets orchestrated out.
The GTM playbook that defined the 2020s is obsolete. The question is how quickly you can rebuild around agents.
This week’s challenge:
Pick one repetitive sales task. One tedious workflow you hate. Build an agent for it.
Skip the months of vendor research. Forget the strategy documents.
Pick a tool
Feed it your docs
Spend three days training it
The agents are working 24/7 right now. Your competitors are deploying them. Lemkin already crossed from speculation to implementation.
Someone’s going to build the 1.2-human GTM team in your market.
Let it be you.
How do you actually coordinate multiple agents without them stepping on each other? That’s a deeper topic—one I’ll cover next time.
Thanks for reading Orchestrated Code! Subscribe for free to receive new posts and support my work.
Subscribe

Available in the Substack app and on web
Thanks for reading Orchestrated Code! Subscribe for free to receive new posts and support my work.
Subscribe
2 comments
I haven't written code in a year. Still shipped 25 projects. Jason Lemkin did the same thing with sales—1.2 humans running what took 10. The pattern works in any domain https://paragraph.com/@xr0am/12-humans-20-ai-agents-same-revenue
AI is coming for the mids