How Daniel Foch replaced a full-time Inside Sales Agent ($50K+/year) with an OpenClaw agent connected to GoHighLevel's API — for under $1K/year. The exact stack, model choices, five automations to build first, and a weekend implementation plan.
Daniel Foch — February 2026. Real estate professional running the entire inbound side of his business through an AI agent: lead response, qualification, follow-up sequences, showing prep, post-close check-ins. Stack: OpenClaw + GoHighLevel + MiniMax M2.5. The position that previously required a full-time hire at $50K+/year now costs under $1K/year with AI.
The average solo real estate agent or small team is paying for some combination of:
Add a part-time or full-time Inside Sales Agent (ISA) — someone who responds to inquiries, qualifies prospects, books showings, and follows up after closes — and you're looking at $35,000 to $50,000 per year in salary alone.
That ISA spends 80% of their time on structured, repeatable tasks. These are exactly the tasks AI agents handle well.
| Function | Traditional Tool | Monthly Cost | AI Replacement | Monthly Cost |
|---|---|---|---|---|
| CRM + contact management | Follow Up Boss | $69/user | Airtable + AI agent | $20 |
| Drip email sequences | Mailchimp Pro | $75 | AI agent + Resend | $10 |
| Lead scoring | BoldTrail / manual | $300+ | AI agent (rule-based + LLM) | $0 (included) |
| Lead qualification (ISA) | Part-time ISA | $2,000–$3,000 | AI agent (auto-respond + qualify) | $15–$40 (API costs) |
| Follow-up sequences | Ylopo / manual | $350 | AI agent (scheduled check-ins) | $0 (included) |
| Appointment scheduling | Calendly Pro | $16 | Cal.com (free) + AI agent | $0 |
| Post-close check-ins | Manual / forgotten | $0 (but costs referrals) | AI agent (30/60/90/365 sequence) | $0 (included) |
| Total | $2,810–$3,810/mo | $45–$70/mo |
| Model | Monthly cost at real estate ISA workload (~3,000 interactions/mo) |
|---|---|
| Claude Opus 4 | ~$800/mo |
| Claude Sonnet 4 | ~$120/mo |
| GPT-4o | ~$100/mo |
| MiniMax M2.5 | ~$30–$40/mo |
| Claude Haiku 4.5 | ~$25/mo |
For lead qualification and follow-up sequences, you don't need frontier-model intelligence. Lead qualification doesn't need Opus. Save Sonnet/Opus for complex negotiation support or unusual situations. Daniel Foch uses MiniMax M2.5 — $30–$40/month vs $800/month for the same workload.
Airtable (free tier works, Pro at $20/mo): Use it as your contact database, pipeline tracker, and interaction log. Airtable's API is clean and well-documented — your AI agent reads and writes to it directly.
GoHighLevel ($97/mo, but replaces 5+ tools): Has built-in email, SMS, pipeline management, and calendar. Daniel Foch's production system uses this. The API supports everything an ISA would do.
For solo agents just starting: Airtable. For teams or agents already using GoHighLevel: stay with GoHighLevel and connect the AI agent to its API.
OpenClaw runs your AI agent 24/7. It handles scheduled tasks (check for new leads every 5 minutes), responding to webhooks, managing conversation state across multiple leads simultaneously, and executing follow-up sequences on schedule.
n8n connects the pieces: new lead → trigger AI agent → agent qualifies → updates CRM → schedules follow-up. Self-hosted n8n is free. Make.com starts at $9/month.
Resend for email ($0/mo for 3,000 emails/month). Twilio for SMS ($0.0079 per message). Your AI agent composes the messages; these services deliver them.
Free, open-source Calendly alternative. Your agent sends prospects a booking link. When they book, Cal.com notifies your CRM, and the agent confirms and prepares a pre-showing briefing.
NAR data is clear: leads contacted within 5 minutes are 10x more likely to convert than leads contacted after 30 minutes. Most agents take hours. Your AI responds in seconds.
Trigger: New lead arrives (Zillow webhook, Facebook Lead Ad, website form)
Agent workflow:
Agent instruction:
When a new lead arrives:
- Greet them by first name
- Reference the specific property or area they inquired about
- Ask: "Are you currently pre-approved for a mortgage?"
- Ask: "What's your ideal timeline for buying/selling?"
- Say: "I'd love to learn more about what you're looking for.
Here's my calendar if you'd like to chat: [link]"
- Keep it under 100 words. Warm, professional, not salesy.
- Sign with the agent's name, not "AI assistant."| Signal | Points | How Agent Detects |
|---|---|---|
| Pre-approved for mortgage | +30 | Asks in initial outreach, parses response |
| Timeline under 90 days | +25 | Asks in qualification sequence |
| Responded within 24 hours | +15 | Timestamp comparison |
| Specific property interest (not "just browsing") | +10 | Analyzes initial inquiry text |
| Budget stated and matches your market | +10 | Parses response to budget question |
| Referral from past client | +20 | Source tracking in CRM |
| Clicked email link | +5 | Email tracking |
Score thresholds: 70+ = Hot (immediate personal call from you) · 40–69 = Warm (AI continues nurturing, you review weekly) · Under 40 = Cold (long-term drip, revisit in 90 days)
Most real estate leads take 6–18 months to convert. The agents who stay in touch win. Your AI never forgets.
Every message references something specific to their search. "Just checking in" alone is spam. "Three new listings hit [neighborhood] this week under $450K — want me to send details?" is service.
Before the showing: Agent pulls property details, comparable sales, neighborhood stats, compiles a one-page briefing (price history, days on market, tax assessment, school ratings, walkability score), sends it to you and the client 2 hours before the showing, confirms the appointment via text.
After the showing: Agent sends follow-up within 2 hours ("What did you think?"). Positive response → ask about next steps. Negative → ask what didn't work, update search criteria in CRM, suggest alternatives. Logs all feedback for future matching.
90% of agents never contact buyers again after closing. The 10% who maintain relationships get 80% of referrals.
At the 90-day and annual check-ins, include: "If you know anyone thinking about buying or selling, I'd love to help them the same way." This single automation, running indefinitely on every past client, generates referral business that most agents leave on the table because they simply forget to follow up.
This is one item from the Ask Patrick Library — battle-tested configs, prompt templates, and implementation guides for AI agents. New items added weekly.
Get Library Access — $9/mo →