At 22, I was making around 200 cold calls a day.

Not emails.

Calls.

I was working in financial research in Argentina and my boss had moved me into sales. We were selling independent financial research to banks, trading desks, CFOs and large companies, which sounds difficult because it was difficult 😂

Most of these institutions were already used to getting research for free.

The junior salespeople before me hadn’t really managed to make the model work.

And now there was me, 22 years old, calling people who had probably worked in finance longer than I had been alive and trying to convince them they should pay attention to what I was selling.

It was exhausting. A lot of people said no. Some people weren’t particularly nice about it either.

And if you gave me the choice today between doing all of that manually again or having access to the sales technology we have now?

Obviously I’d use the technology.

I’m not crazy.

I would automate half of what I was doing before lunch.

But there’s one part of the AI conversation that I keep coming back to, especially as we automate more and more of the junior work inside companies.

That’s the bit I’m not sure we should automate without thinking about what disappears with it.

I think I learned sales somewhere around call 37.

Then a little more around call 86.

Then probably something else around call 143.

You hear the same objections enough times and eventually you stop hearing them as isolated objections. Patterns start appearing.

You notice which opening makes somebody stay on the phone for another 20 seconds.

You learn the difference between someone politely trying to get rid of you and someone who genuinely means, “Call me in three months.”

You hear what CFOs care about versus what traders care about.

You learn what people say when they don’t understand your product.

More painfully, you learn what they say when you don’t understand your product.

And after hundreds and then thousands of these conversations, something starts happening that is quite difficult to put into Salesforce, HubSpot or an Excel cell.

You build instinct.

I could look at my spreadsheet and see the data, yes.

But after enough calls I was carrying another database around in my head.

That database turned out to be quite valuable.

According to my old numbers, I eventually closed more than $150K in annual recurring revenue in about a year and a half.

At 22, that felt enormous.

Looking back now though, the revenue isn’t really the part I find most interesting.

It’s the reps.

This is where it gets complicated.

Look at what modern AI sales tools are promising.

Artisan, for example, describes its AI BDR as something that can find leads, research them, send personalized outreach and book meetings alongside the sales team.

IBM’s explanation of AI SDRs describes basically the same evolution: AI handling parts of the early funnel, identifying prospects, engaging leads, qualifying opportunities and then handing the right ones over to humans.

This is useful.

Seriously.

I would have killed for some of this at 22.

Finding the correct contact manually wasn’t character building.

Copying information into Excel wasn’t character building.

Forgetting to follow up with somebody because row 413 disappeared into the spreadsheet definitely wasn’t character building.

We should automate that stuff.

A huge amount of what people call “paying your dues” was actually just bad software.

But then you get to the harder questions.

What about the conversations?

What about rejection?

What about trying to explain something badly 50 times and slowly hearing yourself become better at explaining it?

What about developing enough pattern recognition that you can hear one sentence from a prospect and already have a pretty good idea where the conversation is going?

I’m much less sure we want AI to remove all of that.

6sense surveyed 872 BDRs for its 2026 State of the BDR report.

The headline number is crazy:

So the debate about whether salespeople are going to use AI is basically finished.

They already are.

Outreach volume has climbed at the same time.

According to 6sense, the average BDR went from roughly 17 attempts per contact in 2024 to around 34 in 2026.

AI makes that possible.

Research is faster. Messages are faster. Follow-ups are easier. You can touch many more prospects without adding another five people to the sales team.

But this is the part of the report that really got my attention:

More outreach by itself wasn’t a reliable predictor of better performance.

Even the most common AI use case - generating messages and content - wasn’t reliably associated with higher quota attainment.

One use of AI that was associated with stronger performance?

Reviewing and analysing conversations.

I think there’s a pretty important distinction hiding in there.

AI seems extremely good at helping us learn from experience.

I’m less convinced we should use it to eliminate the experience completely.

99%

of BDRs surveyed by 6sense now use AI.

But using AI more does not automatically mean selling better.

The interesting question isn’t really:

“Are salespeople using AI?”

It’s:

What are the best salespeople using AI for?

Sales is just where I experienced this personally.

You can see the same thing happening across almost every knowledge job.

A junior developer doesn’t only learn by producing clean code.

They learn because the code breaks.

Then they spend three hours trying to understand why.

Then it breaks again for a completely different reason.

Eventually they develop that strange engineering instinct where they look at something and say, “I don’t know exactly why yet, but this is going to cause a problem.”

A recruiter doesn’t build judgment because an ATS gives them a ranked candidate list.

They build it after interviewing hundreds of people and discovering that the candidate who looks incredible on paper can be terrible in the actual job, while the person with the slightly weird CV sometimes ends up being the best hire on the team.

A marketer doesn’t develop taste because ChatGPT generates 20 headlines.

You develop some of that taste by writing a lot of terrible headlines yourself and eventually understanding why they’re terrible.

An analyst doesn’t only learn from the finished financial model.

Sometimes the learning happens at 11pm when two numbers refuse to reconcile and you have absolutely no idea why.

None of these experiences feel efficient while you’re doing them.

That might be precisely why they’re useful.

And this is something I think about a lot while watching companies hire AI automation specialists and build more automated workflows.

There is enormous value in removing repetitive execution.

The question is what happens when the repetitive execution was also secretly the training program.

This is the bit I keep thinking about.

Look at which tasks AI touches first inside most companies.

And yes, a lot of it is annoying. Nobody dreams of becoming an analyst because they want to spend an afternoon cleaning a spreadsheet.

But those tasks historically did another job that was almost invisible.

They gave people exposure.

You saw hundreds of examples.

You made mistakes that didn’t destroy the company.

Someone senior corrected you.

You tried again.

Eventually you stopped being junior.

That was basically the apprenticeship model of knowledge work, even if nobody called it that.

So if AI does all the first-pass work, we need to think seriously about what replaces the learning that used to come with doing it.

Where does the future senior salesperson learn how buyers actually think if they have never really prospected?

Where does the future senior recruiter develop judgment if an algorithm ranked every shortlist from day one?

Where does the future senior analyst get that weird instinct that tells them something doesn’t look right here before they can even explain why?

Where does a future engineering lead get intuition about bad code if AI fixed most of the bad code before they ever had to wrestle with it?

I don’t think the answer is:

“Well, I suffered, so young people should suffer too.”

That’s stupid.

I don’t want someone making 200 cold calls because I did.

We have better tools.

Use them.

The more useful distinction is probably this:

There is work that is pointless.

And there is work that is painful but useful.

They are not always the same thing.

That’s what companies are going to have to get much better at identifying.

I’d automate the research.

Completely.

I’d automate CRM updates.

I’d have AI prepare a one-page briefing before every important call so I knew exactly who I was speaking with, what the company did, what had happened recently and what might actually matter to them.

I’d have it summarize every conversation.

I’d have it track the objections that kept coming up.

I’d have it tell me where people consistently lost interest in my pitch.

I’d use it for role play before an important meeting.

I’d probably ask it to review ten calls every Friday and tell me what changed during the week.

I’d absolutely automate follow-ups so nobody disappeared because I forgot to write something down.

In other words, I would probably use AI everywhere.

But I would still make a lot of the calls myself.

Maybe not 200 😂

But enough.

Because I didn’t only need the meetings.

Those are two very different things.

There’s a tendency with AI to measure progress almost entirely by how much human effort disappears.

This task took three hours.

Now it takes 20 minutes.

Amazing.

And sometimes that is absolutely the right metric.

But maybe there’s a second question worth asking before we automate everything that looks inefficient:

I don’t think we’ve figured that out yet.

And when basically every BDR already uses AI, junior developers are increasingly coding with AI, recruiters are getting AI-generated shortlists and analysts are getting AI-generated first drafts of their work, we probably need to figure it out fairly quickly.

Technology should remove friction.

It should remove pointless administration.

It should remove work that humans are doing purely because the software used to be terrible.

But I’m not convinced it should remove every struggle.

Sometimes the struggle is where the judgment comes from.

And eventually, judgment is the thing we pay senior people for.

What job taught you something through repetition that you probably wouldn’t learn the same way today?

I’d genuinely like to hear those stories.

Reply to this email and tell me.

— Julian

CEO to CEO is where I write about hiring, AI and what building companies actually looks like once you remove the hype.

👉 Subscribe / read more at juliansigal.com

6sense — 2026 State of the BDR Report

IBM — How AI SDRs are redefining sales

Artisan — AI BDR

HiresLink — AI Automation Specialists