A Founder Automated Outbound and Meta Ads With Claude Code and Hermes: 100 Leads Contacted a Day
A founder lets helpers find leads and draft ads, and he reports how it went.
What happened A startup founder has Claude Code find, enrich and score leads into his CRM while Hermes sends personalised outreach on a cron and drafts Meta ad campaigns. He reports 400 daily clicks from Meta Ads and 100 leads contacted a day at about a 10% response rate.
Who this is about: Early-stage startup founder, Barcelona (Alejandro Gómez; bio: founder of Invisor)
What it takes
- Cost
- From $20/monthClaude Code: Included with Claude Pro ($20 a month billed monthly), Max, Team and Enterprise plans, but not the Free plan. You can also pay for API usage instead.Hermes is open source; you pay for models, hosting and ad spend.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
A founder with no sales team set up helpers for marketing. One finds people who might buy, looks them up, and scores them, and another messages them on a schedule. A helper studies the startup and drafts Facebook and Instagram ads. A separate set of steps checks his site's search rank against rivals.
What was done
Outbound: Claude Code searches for leads where his ICP is (LinkedIn, web search, niche data sources), enriches them through LinkedIn and pushes them into his CRM with a fit score.
Hermes runs cron jobs that contact every lead with a message personalised from the enrichment data.
Inbound ads: Hermes studies the startup and audience, makes creatives with GPT Image and creates draft campaigns through the Meta Ads CLI.
Inbound search: a Claude Code workflow audits his landing page, analyses competitors and finds new keywords.
A different founder, Arthur, had Claude Code read 83,982 tax-preparer websites from the IRS list and score 4,507 LinkedIn leads. Week one of that pipeline brought 3 signups and 0 paid conversions. He says none of the email text is written by AI.
How it went
He reports 400 daily clicks from Meta Ads and 100 leads contacted every day at roughly a 10% response rate, and says SEO research that took hours now takes minutes. Revenue isn't stated; he says the next step is to generate it.
- Volume numbers only: no customers, signups or revenue are given.
- Ad campaigns are drafted by the agent; the post doesn't say who approves spend.
- He doesn't name or pitch his startup in the post, so it isn't marked promotional.
- That same post says a regex title filter had thrown out 2,460 decision makers from 14,778 leads. He suspected onboarding, changed it, and was watching the next signups. This is a separate grade D result, not part of the 100-leads-a-day figures above.
Prompt
Find [50] companies that match this ideal customer profile: [ICP].
For each: company, website, decision-maker name and role, and the public LinkedIn or web evidence that they fit. Score fit 1-10 with one line of reasoning.
Write the results to [CRM or CSV] as new leads, skipping anyone already there. Draft a personalised first message for every lead scoring 7 or more, under [80] words, based only on what you found.
Don't send anything; I'll review the drafts first.Setup notes
- Review the first batches of messages and ad drafts before letting anything send or spend on a schedule.
- Check LinkedIn's terms and local email rules before automating enrichment and outreach.
Business fit
Who this helps, and how:
Founders without a sales team: keep a daily flow of scored leads and first messages going.
Startups- Solo foundersdraft Meta ad campaigns and SEO checks without an agency.
Your turn
Ideas to try on your own work. These are starters, not results anyone has measured.
- Which leads would you still want a person to score before a message goes out?
- What should an ad draft never say without your yes?
- Which part of finding customers would you still do yourself?
Related
- Evidence grade BA Consultant Turns a Company Name Into a Scored Lead in Under 20 MinutesStrong evidence, using Gemini, NotebookLM and Google Sheets
- Evidence grade BCheck Meta's Built-In AI Agent Against Your Real Numbers FirstStrong evidence, using Manus and Meta Ads Manager
- Evidence grade CA Revenue Team That Had Never Shipped Code Built Its Own Call Review Tools in Claude CodeVendor or secondhand, using Claude Code, n8n and Clay
- 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
- AI sales call prepDepartment hub: AI for sales, 34 examples
Similar tools
Other tools for this job, and how they differ. Chosen by our editors, never paid for.
- Gemini
Deep Research with a custom Gem to research and score leads, a lighter option without code.
See it in use: A Consultant Turns a Company Name Into a Scored Lead in Under 20 Minutes
- OpenClaw
Another self-hosted, always-on agent framework that can run scheduled outreach; fully yours to maintain.
- Clay
Data enrichment and scoring across many sources at once.
“Light evidence” (grade D) means: A brief, early or in-progress claim. Treat it as an idea to test, not proof. How we grade. Spotted a mistake? Suggest a correction.