If AI Makes Personalization Free, Personalization Becomes Worthless

There was a time when a highly personalized cold email told you something about the sender.

Someone had probably done some work. They knew what your company did. They might have read something you wrote. Maybe they noticed you were hiring. Maybe they understood your role. Maybe they spent ten minutes figuring out whether contacting you made any sense.

You could still ignore them.

But the message itself carried a little bit of information:

Somebody thought I was worth researching before they asked for my time.

Then generative AI showed up.

Now everybody can do that.

For almost nothing.

"I loved your recent post"

You have probably received some version of this:

I loved your recent post about scaling engineering teams. Your point about balancing velocity and reliability really resonated with me.

A few years ago, that sentence carried more information than it does now. Somebody probably read the post.

Today?

Maybe.

Or maybe software scraped your recent activity, sent it to a language model, generated a compliment, inserted it into a sequence, and moved on to the next 4,000 people.

The sentence can be completely accurate. It can be beautifully written. It can reference something you really said.

And somehow it can feel less personal than:

Hey, this is probably a long shot, but...

Why?

Because recipients know what AI can do now, too.

Stage	What changed
Before: research was expensive	Read the company. Find the person. Read their work. Write the message. Repeat by hand.
Then: AI crushed the cost	Research, summarize, draft, personalize and generate variants in seconds.
Result: the signal weakened	A personalized message no longer proves that meaningful human effort happened before it was sent.

AI didn't just make personalization cheaper

It made personalization less valuable as a signal of effort.

That's the part I find interesting.

When something is expensive to produce, receiving it tells you something about the sender.

A handwritten letter? Effort.

Thoughtful research about your company? Effort.

A carefully written explanation of why the sender thinks the two of you should talk? Effort.

Then technology drives the cost of producing the artifact toward zero, and the artifact no longer tells you much about the effort behind it.

This doesn't mean personalized outreach is useless. It means personalization by itself stopped being evidence that the sender has a reason to contact you.

Personalization and relevance are not the same thing.

I learned this the expensive way

At my previous company, I had a sales team doing outreach every day, and the response rate was essentially zero. We trained. We changed messaging. I wrote an AI training program for the team. I got involved in calls.

Nothing fixed the underlying problem.

And when we did get meetings, too many were with companies that never should have consumed the team's time. One had roughly $1,000 in the bank. One couldn't meaningfully buy from us through the person we'd reached. One involved someone presenting themselves as having authority they did not appear to have, for reasons I still do not understand.

We didn't have an email-writing problem.

We had a targeting problem.

(The longer version of that story, including the 2 AM email with corporate counsel copied, is here.)

The inbox has a supply problem

Imagine that every salesperson suddenly becomes capable of writing ten times as many decent messages.

Then imagine every sales organization gets the same capability. Then every agency. Then every recruiter. Then every founder.

What happens?

The recipient does not acquire ten times as much attention. There are still 24 hours in a day.

The inbox simply gets louder.

Time	Opening line
8:04 AM	"I noticed your impressive growth..."
8:17 AM	"Loved your recent post..."
8:31 AM	"Congrats on the funding announcement..."
8:46 AM	"Given your role at [COMPANY]..."
9:02 AM	"I saw you're hiring engineers..."
9:11 AM	"You're hiring six people to rebuild X while your team is also doing Y. We work on exactly that problem."
9:24 AM	"Just bumping this to the top of your inbox..."

That sixth message isn't better because the prose is prettier.

It's better because, if the evidence behind it is real, it contains a reason for the conversation.

So what still has value?

Relevance.

Not "I noticed you went to Ohio State."

Not "I saw your CEO was recently interviewed."

Not "Congratulations on your Series C."

Those things can make a message personal. They don't necessarily make it worth answering.

The useful questions are harder:

• Why are you contacting me?
• Why now?
• What do you understand about my company that makes this conversation worth having?
• What changed?
• What problem do you think we might actually have?
• And can you show me enough evidence that I don't have to take your word for it?

AI can manufacture language. It cannot manufacture reality.

This is the distinction I keep coming back to.

A model can generate a beautiful sentence about a prospect. It can generate 500 versions of the sentence. It can change the tone. It can mention the person's school. It can summarize their podcast appearance. It can make the email sound casual, or thoughtful, or like you spent an hour writing it.

What it cannot legitimately do is invent a reason the company should care and pass it off as reality.

A team either expanded or it didn't.

A new executive either arrived or they didn't.

A company either started rebuilding a system or it didn't.

A role either explicitly owns a problem or it doesn't.

A product either launched or it didn't.

A company either publicly described a priority or it didn't.

AI can help find those things. It can help connect them. It can help explain them.

But underneath the language, something real has to be there.

The scarce thing is no longer the ability to write a personalized message. The scarce thing is having something worth saying.

This changes what AI should do for sales

The obvious application of generative AI to sales was:

Help every salesperson send more messages.

And, to be fair, AI is extremely good at that.

But if everybody uses the same breakthrough to increase message volume, we shouldn't be surprised when recipients respond by ignoring more messages.

I think the more interesting application is almost the opposite:

Help every salesperson figure out which messages are actually worth sending.

Less outreach can be a feature

This sounds strange in a category that spent years optimizing for volume. More contacts. More sequences. More touches. More automated follow-ups. More personalization.

But suppose AI could instead help you eliminate most of the companies you were about to contact, because there is no strong reason to believe they need what you sell right now.

For a lot of sales teams, that might be more valuable than helping you write to all of them.

The salesperson gets time back.

The prospects who never should have been contacted get silence.

And the remaining companies get something much rarer: a message from someone who actually has a reason to be there.

1. Possible companies: thousands you could contact.
2. Seller fit: companies that plausibly buy what you sell.
3. Current evidence: something relevant is actually happening.
4. Defensible need: the evidence supports a problem you solve.
5. Human attention: spend time here.

There are no percentages on that funnel. That's intentional; we haven't measured them, and the point is the direction of optimization.

Don't use AI only to make the top of the funnel cheaper.

Use it to decide what deserves to survive the funnel at all.

This is not an argument against humans in sales

Quite the opposite.

The more synthetic communication there is, the more valuable genuine human judgment becomes.

A good salesperson can hear hesitation. They can ask the question the system didn't anticipate. They can understand politics. They can build trust. They can realize the original hypothesis was wrong halfway through a conversation and change direction.

Those are valuable things to spend human intelligence on.

Reading 200 company pages to decide which five might be worth calling?

Less so.

Writing 200 slightly different versions of the same email?

Definitely less so.

AI should remove the work that keeps the salesperson from doing the human part. It should not create an industrial-scale machine for pretending the human part happened.

The future probably isn't better spam

Generative AI gave us the ability to create vastly more messages that look like somebody spent time writing them.

The obvious response was to create vastly more messages.

I am not convinced history will regard this as our finest idea.

When every message looks researched, looking researched stops distinguishing you. When every message is personalized, personalization stops being special. When everyone can manufacture the appearance of effort, the appearance of effort becomes cheap.

So the advantage moves somewhere else.

Maybe the winning salesperson isn't the one who can send the most personalized messages. Maybe it's the one who can answer a much simpler question:

Which 20 people do I actually have a good reason to contact?

Then AI isn't being used to manufacture effort. It's being used to decide where human effort is worth spending.

I think that's a much more interesting future for sales.

And considerably less annoying for everyone with an inbox.

---

Start from the beginning: I Built Kairo Because Sales Stopped Working