I Built Kairo Because Sales Stopped Working

I did not set out to build another sales platform.

I was trying to figure out why nobody would answer my sales team.

At my previous company, we had people doing outreach every day. They weren't bad at their jobs. They weren't lazy. They weren't sitting around waiting for leads to magically appear.

They were prospecting. They were sending messages. They were following up.

And the response rate was essentially zero.

For months.

They kept coming back to me baffled. Honestly, so was I.

The meetings we did get were somehow worse

Occasionally, somebody would respond. For a brief moment, this felt like progress.

Then we'd get on the call.

These aren't edge cases I dug up years later to make the story better. These were actual meetings my team had worked to create.

Every one of them represented hours of somebody's life. Research. Prospecting. Follow-up. Preparation. The meeting itself. Then the postmortem where everybody tried to figure out what the hell had just happened.

The problem was not that we needed more leads. We had plenty of people to contact. We were contacting the wrong people.

So I tried fixing the sales team

That seemed like the obvious place to start.

Maybe the messaging was bad. Maybe the reps needed better training. Maybe the research wasn't good enough. Maybe they weren't explaining the value clearly.

I wrote an AI training program for the team.

No meaningful improvement.

I got on calls with them.

No meaningful improvement.

We worked on outreach.

No meaningful improvement.

Eventually I had to consider the possibility that the problem wasn't the people doing the selling.

The environment had changed.

AI made outreach cheap

This is the part I underestimated at first.

Generative AI made it incredibly cheap to produce outreach that looked researched. Suddenly everybody could send:

I saw your recent post about scaling engineering teams and really liked your point about balancing speed with reliability...

Thousands of times. With different names, different companies, different "personal" observations.

And recipients knew it.

The volume went up. The meaning of "personalized" went down.

Being able to write a better cold email was no longer enough. The problem had moved one step earlier. Before asking "What should we say?", we needed a much better answer to "Who the hell should we be talking to?"

I needed to know something the prospect cared about

Not their favorite football team. Not where they went to college. Not that their CEO had appeared on a podcast.

Something about the company that mattered. Something current. Something connected to an actual problem, change, risk, priority, or investment.

And, critically, something I could point to right away, so the person on the other end knew this wasn't another mass email wearing a personalized hat.

That was the beginning of the analysis system that eventually became Kairo.

I started pulling together public information from everywhere I could get it responsibly: job postings, hiring patterns, company pages, technology usage, news, people, public statements. Anything that could tell me what was actually happening inside a company without pretending I could read anyone's mind.

Then I tried to connect the pieces.

At first, I wasn't sure it worked

This part matters.

I did not build the first version, look at a few outputs, and immediately decide I'd discovered fire.

It looked promising. It found things that looked interesting.

But I was close to the system. I knew how it worked. I knew what I wanted it to find.

That makes it very easy to fool yourself.

Then somebody tried to sell me data analysis.

The company that wanted to sell me research

A company approached me offering corporate data analysis services.

I told them I wasn't interested. I already did my own.

You can probably imagine how convincing that sounded. "AI-powered research" was already everywhere. So was proprietary data. So were systems that supposedly found things nobody else could.

The reaction was basically:

Yeah, right, buddy.

Fair enough.

I told them mine had found things about companies that other people missed. Eventually their COO got involved, and the request was simple.

Prove it.

So I ran their own company through the system and sent them the analysis.

Analysis: [REDACTED]

Reconstructed example. The company name, exact wording, role details and identifying facts have been intentionally changed. It is more heavily sanitized than the original: it preserves the kind of analysis that was delivered without preserving enough exact text to identify the company through search.

• Engineering delivery is under pressure. Public hiring material points to a push for faster delivery, clearer operating discipline, and fewer blockers between engineering teams and roadmap commitments.
• Product and engineering alignment appears strained. Multiple roles emphasize cross-functional trust, shared accountability, and tighter coordination across product, engineering, design, and customer-facing teams.
• Product prioritization is being rebuilt. Public roles call for stronger prioritization frameworks, more disciplined product operations, and better use of tooling to guide roadmap decisions.
• Customer retention is receiving direct attention. Commercial roles include explicit responsibility for identifying churn risk early and intervening when engagement or outcomes fall below expectations.
• Outsourced engineering created control concerns. Leadership responsibilities include tighter vendor discipline, clearer quality gates, knowledge transfer, and keeping core architecture and intellectual property in-house.
• Data quality requires manual correction. Public hiring language calls out classifier quality, error patterns, freshness, null values, duplicates, and ongoing monitoring.
• Enterprise requirements are outgrowing parts of the platform. Leadership roles call for deeper enterprise capabilities around identity, access control, auditing, governance, and isolation.
• Product teams are spending meaningful time on incident triage. Several roles include root-cause analysis, reactive troubleshooting, and reducing recurring product issues.

Then they disappeared.

For hours.

No response. No "interesting." No "thanks." Nothing.

I remember wondering whether I'd overplayed it.

Then, around 2 AM, I got an email.

The CEO emailed me at 2 AM

Their corporate attorneys were copied.

The basic message was: how did you get this?

They believed I had obtained internal company information improperly. At one point, the word "espionage" entered the conversation.

From: CEO · CC: Corporate counsel · Sent: around 2:00 AM

Subject: Your analysis of our company

We are extremely concerned about the information contained in the analysis you provided and how it was obtained. Some of these conclusions involve matters that are not broadly known inside the company. Please explain the source of this information and whether you or anyone acting on your behalf accessed non-public systems or information.

Reconstructed for illustration, not a quotation from the original email.

That was unexpected.

There was just one problem with the accusation. Everything I had used was public.

No hacked systems. No leaked documents. No employee feeding me information. No private database I shouldn't have had.

Job postings. Public company information. Public statements. Things the company itself had put on the internet.

The only difference was that I had put the pieces together.

That was the moment

The strangest part of the exchange was that, according to them, some of what I'd found wasn't obvious even to the people running the company.

That stuck with me.

A company can publish an enormous amount of information about itself without ever seeing what all of it says once the pieces sit side by side.

A job posting lives with recruiting. A technology change lives with engineering. A customer role lives with sales. A product announcement lives with marketing. A new executive lives somewhere else entirely.

Inside the company, those facts are spread across teams, systems and people. From the outside, you can sometimes connect them. And sometimes they tell a story that no single source does.

I remember thinking something very close to:

Oh. This might actually work.

That became the real idea behind Kairo

Not "use AI to send more sales emails." Definitely not "generate better personalization." The world has enough of both.

The idea was much simpler:

1. Find the companies where something relevant is actually happening.
2. Figure out whether it matters to what a specific seller does.
3. Show the evidence.
4. Explain the connection.
5. Give the salesperson enough real understanding to start a conversation for a reason.

Because the lesson from my old sales team wasn't that salespeople needed to work harder. They were already working.

It wasn't that they needed more prospects. They already had them.

It wasn't even that they needed more personalized messages. AI had made those cheap.

They needed a better reason to contact someone in the first place.

The most expensive sales mistake is not a bad email. It is spending time on a company that never had a reason to buy from you.

That's the problem I started trying to solve.

I just didn't know yet that it was going to become a company.

---

Next: Public Data Is Abundant. Evidence Isn't.