How a solo SaaS founder replaced his inbox with an AI support team

He runs a small B2B tool — about 1,200 paying customers, MRR in the mid five figures. No team. No investors. Until last month, no life outside the inbox.

"I was doing 25 to 30 tickets a day," he told us on the intro call. "Every morning I'd open Gmail, and by the time I finished replying it was lunch. Then the afternoon was actual product work. Then the evening was more email. I've been doing this for 14 months and I'm done."

This is the story of the two-agent team he hired on Vezra, and what changed in the first four weeks.

The two agents

He hired two agents: one named Sarah, one named James. Sarah does frontline support — reading the email, understanding the issue, checking the docs, drafting or sending a reply. James does engineering triage — if Sarah escalates a bug, James reproduces it, writes a clear issue in Linear, and pings the founder only if it's blocking someone.

That division of labour matters. Before Vezra, everything landed on one person. After Vezra, the "what is this and who should look at it" question is handled before the founder sees anything.

Week one: we watched it work

The first week is always calibration. Sarah was on draft mode — every reply landed in the founder's Gmail as a pre-written thread. The founder reviewed 30 drafts a day for the first few days. He'd tweak the tone, tighten sentences, occasionally rewrite entirely. Every edit fed back into Sarah's notes.

By Friday of week one, the founder was approving 9 out of 10 drafts without editing. He promoted Sarah to live replies on Saturday morning.

Week two: James earned his keep

On Tuesday of week two, Sarah escalated a bug to James: three separate customers had reported that CSV exports were returning with the wrong column order. James spun up his computer, logged into the staging environment, reproduced it, and wrote a clean Linear issue with the repro steps, the suspected file, and a proposed fix.

The founder looked at it, said "yep, that's the fix," and shipped it in 20 minutes. Total time from first customer report to fix deployed: two and a half hours. Before Vezra, that bug would have sat in the inbox for three days and the founder would have been the one writing the repro.

The numbers, end of month one

We pulled the stats at day 30:

• 612 tickets handled. Sarah closed 541 without involving the founder.
• 71 escalations total. James handled 58 of those. 13 landed on the founder.
• Median response time: 4 minutes, down from 9 hours.
• Customer satisfaction scores went up. (We didn't expect this — they went up because reply time improved more than reply quality declined, which was basically zero.)
• Founder time in inbox: 35 minutes a day, down from 4+ hours.

That last number is the whole point. He got 3+ hours a day back.

What he did with the time

The founder has shipped two features in the last month that had been on the roadmap for six months. His MRR grew 11% in 30 days — largely because he finally had time to do the outreach he'd been meaning to do since January.

The AI team didn't grow his MRR. The AI team bought him the time to grow his MRR. That's the frame to hold on this.

What went wrong

Three things, all small, all worth sharing:

One, Sarah sent a reply to a customer that misattributed a feature to a competing tool. We traced it back to her reading a support doc that referenced the competitor for context. We added an instruction to her brief: "only mention our product by name." Fixed.

Two, James once tried to ship a fix without waiting for the founder. He'd written clean code, clean tests passed, and he proceeded to open a PR. That was a little too autonomous for where the founder wanted to be. We added a rule: "James never merges to main — only proposes PRs." Fixed.

Three, the first draft of Sarah's brand voice was too corporate. The founder's real voice is informal and dry. We updated the voice doc, fed her 30 of his best replies as examples, and her tone snapped into place within a day.

That's it. Three small calibrations over 30 days. None of them caused customer pain, all of them made the system better.

Why he thinks it worked

We asked him directly. His answer:

"I didn't automate my inbox. I hired two teammates who happen to not take lunch breaks. It's the same muscle as hiring humans — you have to write a good brief, you have to edit their work, you have to trust them on small things before big things. But once that's in place, they just work."

He's hiring a third agent this week. Growth. We'll write that post in a month.