I spent ~2 years building an AI SDR (starting in 2020). This was an extremely hard business to run, and I learned so much.

Long story:

5 years ago, Apollo was a fraction of the size it is today. Exporting data from a data provider to an email sequencing tool felt archaic.

Someone was going to come along and build an all-in-one tool, and I wanted to take a crack at it.

I spent a few months building it but wasn't feeling much pull from the market. "Why is this better than ZoomInfo + Outreach?" was what I'd hear.

And then...OpenAI launches GPT3. Holy crap I thought...this changes everything.

Most cold emails use the same body, but personalize the opening sentence...which is done manually. If AI could just do that for you, you could automate the whole thing!

So I add this feature in, launch a sequence with it, and the results are..nuts. I'd send maybe 50 emails a day and book 3-5 meetings. I never saw anything like this.

I'd end each email with "ps. this email was written with AI" and prospects loved it.

I thought I had PMF. Within 3 months I hit 100K ARR! I'm starting to think what color bugatti Im gonna buy to match the lambo I'll buy. I'm sorting homes price hi-low on Zillow.

I 50x'd my pricing and no one flinched. I went from being a PLG saas app at $20/mo to $1,000/mo. "The app pays for itself"

But then, I started to notice some problems. After the first couple days, retention TANKED. No AE or SDR was using it.

When I'd ask why, I'd get the response every founder hates: "Sorry I've just been busy". This, I learned, is code for "your app sucks".

At first, the pushback was a lack of features. "We need a Salesforce integration. We need account mapping. We need blah blah blah".

Easy! I'll add that in. Then surely usage will increase. But still, crickets.

So, I started getting on calls with people to setup their first campaign. And this is when I realized this business is dead.

"We're going to automate away prospecting" is a killer pitch when selling to sales orgs. You don't even need to personalize it. People want it.

BUT, when you have OTHER orgs using your tool - sooo many variables get introduced. How good is their lead list? How refined is their ICP? Do they have healthy email domains? How strong is their pitch?

With all those dialed in, you see maybe a 3% reply rate.
With one of those missing, you're screwed. And people blamed my AI SDR for it.

From this I learned, don't overpromise. I see sooo many of these YC AI slop companies that raise millions in funding for this same premise, but suffer from insane churn.

You’re better off blowing minds with a simple win than underwhelming with an overhyped promise.


This post was originally shared by on Linkedin.