A Part-Time Founder's Agents Found His First Pilot Customer on Their Own
A solo founder selling quoting software to industrial insulation contractors runs about 16 agents on evenings and weekends; 214 agent-written cold emails at a 3.3% reply rate produced his first pilot customer without him touching the funnel until the demo.
Who this is about: Solo founder of quoting software for industrial insulation contractors
What it takes
- Cost
- Free planClaude: Free plan. Pro is $20 a month billed monthly ($17 a month billed yearly), and Max starts at $100 a month. Team seats are $25 a month billed monthly.Price checked Oct 24, 2026 on claude.com. Prices change; check before you buy.
- Setup time
- A few weeksOur estimate from the source, for one business.
- Skill needed
- Developer skillsYou build or run your own code, even if AI writes it.
Explain it simply
While the founder works another full-time job, a group of AI helpers research companies, write cold emails, and keep his sales list up to date. He only steps in when someone wants a demo. A separate checker blocks emails that claim features the product does not have, so the outreach stays honest.
What was done
He works full time at an insulation and fire-protection company, so his agents run the business: lead research, personalized cold email, CRM upkeep, SEO content and blog publishing, LinkedIn comments on target CEOs' posts, Reddit scouting, call-prep sheets and call transcription. The research agent surfaced a company, enrichment qualified it, the email agent wrote the outreach, and he only joined for the demo.
How it went
214 cold emails at a 3.3% reply rate led to one qualified pilot customer, with the pilot set to start mid-August. He calls the funnel thin but says it delivered the one result he needed.
- Limits: the copywriting agent kept promising features the product does not have. "Never claim X" prompt rules reduced this but never stopped it, so he added a separate checking agent that rejects any draft with a claim not on an approved list.
- He also moved rules like do-not-contact out of the prompt and into code, so the send step refuses those addresses. One prospect clicked but never replied; a phone call turned into the best conversation of the campaign.
Prompt
1. Keep a hard list of allowed product claims. A separate QA agent checks every outbound draft against it and rejects anything not on the list.
2. Enforce do-not-contact in the sending code, not in the prompt.
3. If a recipient engages but does not reply after two emails, flag them for a phone call instead of sending a third email.His guardrails, in our words.
Business fit
Who this helps, and how:
Startup founders and e-commerce sellers: run research and outreach for a niche B2B product while still employed, with hard limits in code.
StartupsMechanics and tradespeople: a trade business selling to other contractors can use the same research-then-email chain with an approved-claims list.
Local & serviceAgencies and freelancers: add a claims-checking step before any agent-written email goes to a client's prospects.
AgenciesLocal service businesses: research commercial prospects and send a small number of tailored emails, then call the engaged ones.
Local & serviceIndependent contractors (barbers, bartenders, consultants): consultants can keep a pipeline moving with call-prep sheets ready before each meeting.
Agencies
Related
- Evidence grade BAn Austin Realtor Kept 6 of 28 Claude Workflows, Including a CRM Sweep That Turned Up a DealStrong evidence, using Claude, Claude Code and Follow Up Boss
- Evidence grade DA Claude Agent Takes Instagram DM Orders for a 7-Location Sushi ChainLight evidence, using Claude, Meta API (Instagram DMs) and The restaurant's CRM
- Evidence grade BA Consultant Turns a Company Name Into a Scored Lead in Under 20 MinutesStrong evidence, using Gemini, NotebookLM and Google Sheets
- AI for sales prospecting examplesDepartment hub: AI for sales, 34 examples
Similar tools
Other tools for this job, and how they differ. Chosen by our editors, never paid for.
- n8n
Open-source style automation for chaining research, enrichment, and send steps with your own rules in code.
- Make
Visual automation for the same research-to-email pipeline if you prefer a no-code canvas.
- Instantly.ai
Built for high-volume cold email sending and deliverability; pair it with a separate research step.
“Strong evidence” (grade B) means: The person who did the work describes the setup and results in detail. How we grade. Spotted a mistake? Suggest a correction.