170 cold emails. 0 replies. Here's the ICP mistake I made — and what finally worked.
I've been building Velor — an AI customer support agent for small B2B SaaS teams. The product works. Getting first customers has been harder.
Here's what I got wrong.
The mistake: targeting up-market
I got excited about logos. Sent cold email to Rippling, Gusto, Deel, Mercury — 50+ emails to companies with 100+ employees. Zero replies. Then I tried the 50–200 employee band. Still nothing.
The problem wasn't the pitch or the copy. It was the target. Companies that size have:
— A dedicated support/CX team
— Intercom already contracted and budgeted
— Procurement that takes months
— A "Head of CX" who deletes cold email from unknown founders
170 emails, 0 replies confirmed it: those weren't my buyers.
What actually works: 2–10 person founder-run teams
When the FOUNDER personally answers support tickets, everything changes:
— They feel the pain every single day
— They make purchasing decisions in hours, not months
— A $500/mo tool that saves 4 hrs/week is an obvious yes
First trial customer: 2-person fintech. Founder was in their support inbox 4 hours a day. Velor deflected 70% of tickets in week one — no hallucinated answers.
The product
Velor grounds answers in your actual docs (not guesses), and takes real actions — looks up order status, escalates to a human, files disputes — instead of deflecting to "please contact support."
No signup: velorplatforms.com → "speak with an agent." Live demo, 30 seconds.
If you're a 2–10 person B2B team and you're personally answering support tickets, this is for you. Also happy to talk — sharing here partly because IH is full of exactly the founders I'm looking for.
170 emails with zero replies sounds like a copy problem until you realize great copy sent to the wrong buyer is just beautifully written spam.
Testing who actually has the pain before scaling outreach is probably the highest-ROI email optimization there is.
"Beautifully written spam" is the exact framing -- and it took me longer than it should have to see it that way. The instinct when you get 0 replies is to rewrite the email. The actual problem was I was writing to people who didn't personally feel the cost of the problem. When support is already someone else's job, even a great email is just noise.
The signal that worked for qualifying who had the pain before outreach: does the founder's name show up in their own support channel? Reply to a support email from them directly, see how fast they respond. If they're personally in the queue, they feel it. If support goes to a help@ alias with a 24-hour response time, you're reaching someone who has already delegated the problem -- and no amount of copy fixes that.
The founder-run distinction seems much more important than company size alone. Once the person experiencing the support burden is also the buyer, the pain, urgency, and purchasing friction all change at once. The 70% deflection result makes that ICP shift even more interesting.
That's the exact mechanism. When the buyer is the one answering tickets, they're not paying for software -- they're buying back time they can feel losing every hour. The pain is immediate and personal rather than reported by someone else on the team.
The 70% deflection held up mostly because the tickets that stayed required judgment the founder actually wanted to make -- relationship calls, edge cases, anything where context mattered more than the answer. The ones that deflected were real questions with known answers. That split was much harder to see when we were targeting companies with dedicated support staff.