A couple weeks ago I sent you an email with the subject line "AI is destroying capital raising."

I meant it.

But {{first_name}}, a few of you wrote back with a fair question:

“Okay, so what do you actually use it for?”

Because I do use it. Every day. It isn't a side tool for me, it's most of how the work gets done now.

And I don’t pretend that I don’t use it (although that’s oddly popular right now).

So here's the honest version, starting with the time I talked a prospect out of the thing he was trying to hire me for.

He's a sponsor who wanted me to build out his investor relations system. Good call, real budget. About forty minutes in, he asked,

“How much of it runs itself, how fast could we switch it on?”

….and then of course,

“Could the whole thing basically run without me?”

Instead of selling him the dream, I told him this:

Over automating everything is where a lot of people get in big trouble.

Which is a strange thing to say to a man who is deciding whether to pay you.

But we want AI everything right now, and I've watched what that does to operators who get carried away with it...

...they end up with a very efficient machine sending very forgettable emails to people who were never going to answer.

Here are the four things I've actually learned. I paid for most of them.

Not the outreach, not the investor updates, not this letter you're reading right now.

I hand write it. I actually type it out, badly sometimes, and then I go back and fix it.

That’s better.

Jokes aside, I do use AI to build topics that are valuable to capital raisers and to make sure there aren't crazy mistakes in there, and that part matters more than people think. But the words are mine, for two reasons:

The first one is practical. Allocators read hundreds of these a month and they are getting much better at spotting what was AI written and what wasn't. The second they clock it, you stop being a person and become a campaign.

The second reason is the one I actually care about. If AI writes the email, you never think about the investor at all. You skip the ten minutes where you sit there and work out what this specific human being cares about, and why on earth they should care about you.

Those ten minutes are the job. The email is just the receipt.

This is where it has genuinely earned its keep for me, and it's the use nobody seems to talk about.

Every investor call I have gets transcribed. Then I run the transcript through Claude and pull out the part that never used to survive the week.

Not the check size or the deal terms, I'd remember those anyway. I mean the story around the person. How they found us, who introduced them, what they're actually nervous about, and how they feel about us, which is a completely different question from how we feel about them and a lot more useful to know.

Here's why that matters. You have a great call with somebody in March, and you call them again eighteen months later because you finally have something that fits them...

...and you remember nothing, because you've had four hundred conversations since.

With the transcript sitting there, you open the file before you dial and you already know to ask how their son is doing in school, and whether he's still playing soccer.

I'm not going to pretend that's a relationship. It isn't. It's the difference between calling somebody and calling somebody.

Upstream of a human being, before anyone has to talk to anyone, is where I think this stuff actually belongs.

I run models that sift through data on the individuals and groups who have invested in this kind of asset before, looking for the signals that say they'd do it again. That's real work, and it's work a person genuinely cannot do at volume. A machine doesn't get bored on record 4,000. I do.

Then it narrows. We're not reaching out to a family office in Rhode Island unless that Rhode Island family office actually loves what we're bringing them: right metro, right check size, real history in the asset class.

And then a human being writes to them. Every…single…time.

Not a bot.

Not AI.

A human.

The operators I watch struggling have this exactly backwards. They automate the conversation and do the targeting by hand, which is the most expensive possible way to get ignored.

I've bought data. There's a CSV sitting on my desktop right now with roughly 140,000 contacts in it, and it's a perfectly fine place to start.

But the first thing I push into a client's system on day one isn't purchased data. It's my own database, the ones I vetted, the ones I actually know.

And almost every time I sit down with a sponsor who is convinced he needs cold outreach, we find he's been sitting on a warm database he's ignored for two years. People who already invested in his deals. People who took a call back in 2023 and never heard from him again.

That's where the money is, and it's boring, so nobody wants it to be true.

AI is very good at enriching those records. It is not going to create them for you, and if there's nothing real underneath, all you've built is a faster way to get ignored.

Honestly? I don't think I've got this fully worked out. I change how I use these tools every couple of months, and some of what I just wrote will probably look dumb to me by Christmas.

But the shape of it hasn't moved: use AI on everything that happens before the conversation, and nothing that happens inside it.

That one has held up, and it's the only rule I'd hand somebody else.

Anyway…thanks for reading this one.

Have a great Friday and the rest of the weekend,

Andrew LeBaron

P.S. If you want a read on where your own raise is actually breaking down, grab time with me here: andrewlebaron.com/meetwithandrew

Target returns are projections only and not a guarantee of future results. Actual results may differ materially. This is not an offer to sell securities. For accredited investors only.