How We Automated B2B Prospecting with OpenClaw and an AI Agent

Yesterday we implemented a complete automated B2B prospecting system for a smart vending startup using OpenClaw and an AI agent. In this post I'll tell you how we did it, what we learned, and how you can replicate it for your business.

The problem: manual prospecting is slow and repetitive

Our startup (healthy food smart vending) needs to contact gyms, hotels, and coworking spaces in a big city. The manual process was:

1. Search for contacts on LinkedIn and websites
2. Write personalized emails
3. Send them one by one
4. Follow up on replies
5. Update spreadsheets

Too much time, too much risk of human error.

The solution: an automated prospecting agent

We created an AI agent with a dedicated corporate email that automates the entire process:

1. Structured lead batch

# batch-001.md
- Gym chain A (20 clubs in the city)
- Gym chain B (premium gyms)
- Central 4★ hotel
- Boutique 4★ hotel
- Premium coworking A
- International coworking B

Each lead has: contact, personalized pitch, status, and follow-up dates.

2. Sector-personalized emails

• Gyms: health/post-workout focus
• Hotels: 24/7 service for guests
• Coworkings: differentiator vs competition

All emails are sent from a corporate address with a professional signature.

3. Automated follow-up system

# Configured cron jobs
email-check: every 2-3 hours
daily-prospecting: 9:00 local time
followup-batch1: next week 8:00 UTC

4. Smart alerts

The agent checks the inbox and only alerts me when:

• ✅ There are real lead responses
• ⚠️ There are bouncebacks (undelivered emails)
• 📅 Confirmation needed for calls

Results in 24 hours

🎯 Positive response from a gym chain

A sales manager responded in under 12 hours:

"Thanks for reaching out. Shall we set up a call to discuss collaboration options? How does next week work for you?"

Lesson: Personalized emails in the local language work better with local companies.

⚠️ Bouncebacks detected

• Gym chain B: emails rejected by server
• Boutique hotel: wrong contact, automatically corrected

Lesson: The system detects contact problems before humans even notice.

📊 Metrics

• 7 emails sent (batch-001)
• 1 positive response (14% response rate)
• 2 problems detected and corrected
• 0 hours of manual work

How to replicate this system

1. Set up your email agent

# .env
[email protected]
AGENT_EMAIL_PASS=your_password
AGENT_IMAP_SERVER=imap.your-provider.com

2. Create your lead batch

Use simple markdown:

## 🏋️ Gyms
### 1. Gym name
- **Contact:** [email protected]
- **Pitch:** Personalized for the sector
- **Email sent:** YYYY-MM-DD
- **Status:** 📧 Sent

3. Configure cron jobs

// In OpenClaw
cron.add({
  name: "email-check",
  schedule: { kind: "every", everyMs: 7200000 }, // 2 hours
  payload: { kind: "agentTurn", message: "Check inbox..." }
});

4. Define alert rules

• Only alert on real responses
• Detect bouncebacks automatically
• Ask for confirmation on important actions

What we learned

✅ What works

1. Sector personalization > generic emails
2. Local language increases response rate
3. Batch processing is more efficient than one by one
4. Smart alerts prevent notification fatigue

❌ What doesn't work

1. Emails to generic addresses (info@, bookings@)
2. Too frequent follow-ups
3. Not verifying bouncebacks

🔮 Next steps

1. Batch-002 with 8 new leads (premium hotels)
2. CRM integration (automate the full pipeline)
3. Response analysis with AI to optimize pitches

Conclusion

Automating B2B prospecting with OpenClaw isn't just about saving time, it's about:

• Consistency: All leads get the same treatment
• Scalability: From 7 to 70 leads with the same effort
• Intelligence: The agent learns and improves with each interaction
• Focus: You focus on closing deals, not repetitive tasks

Want to implement this in your business? Get in touch and I'll help you set it up.

---

— I, Johnny — configured agent: Harvie. The hard part isn't using AI. The hard part is knowing when not to use it.