Case study · Real estate · AI agents on GoHighLevel
Open DealFlow: every new lead called, texted and emailed by AI, and old leads worked every 30 minutes
Open DealFlow is a platform for real estate wholesalers and acquisition teams that run on GoHighLevel. We built three things for it: an AI system that follows up new leads and reactivates cold ones, a portal that shows a sales team what its CRM does not, and a team of three AI agents for acquisitions.
- Client
- Open DealFlow, a platform for real estate wholesalers
- Industry
- Real estate
- What we built
- AI lead follow-up, cold lead reactivation, the Pulse portal and an AI agent team
- Tools
- GoHighLevel, VAPI, Twilio, n8n, Supabase, OpenAI and Claude
- AI in the product
- Claude and OpenAI models
- Status
- Follow-up, reactivation and the portal are live. The agent team is configured, and its production deployment is in progress
- 3channels on every new lead: an AI voice call, an SMS and an email
- 24 hoursuntil an unanswered lead is tried again
- 5measures used to score every cold lead, every day
- 30 minbetween runs that work the highest-scoring cold leads
The problem
A wholesaling team lives on speed. A lead who is not called quickly goes to a competitor, and a lead who did not answer last month is rarely called again. GoHighLevel holds the contacts and the pipeline, but it does not make the calls, decide which old leads are worth another try, or give an owner one view across several accounts.
What we built
AI follow-up and cold lead reactivation
New leads are worked at once, and old ones are worked again, with no person involved until a deal is warm.
- A new lead in GoHighLevel triggers an outbound AI voice call, an SMS and an email.
- If nobody answers, the lead is queued and tried again 24 hours later.
- The outcome of each call moves the deal to the right pipeline stage.
- A daily scan scores every lead sitting in the nurture stage on five measures: stage, how recent it is, earlier responses, how complete its data is and how many attempts it has had.
- Every 30 minutes, the highest-scoring leads get an SMS, a voice call and an email. Each is then marked re-engaged or dead from the result of the call.
- A dashboard shows every run and lets the team change the settings without opening the workflow tool.
The Pulse portal
A white-label portal for sales teams on GoHighLevel. A user switches between several GoHighLevel accounts and sees one view of the pipeline.
- Sales dashboards with live figures, pipeline speed and lead tables.
- Executive views: revenue, source performance, weekly activity and each rep's performance, including speed to lead for AI against people.
- Marketing attribution: lead sources with conversion rate and return.
- Deal views: a pipeline board, a revenue forecast and a map.
- Deliverability monitoring across SMS, email and voice, with recommendations written by AI.
- An AI chat assistant for managing the CRM, paid for per use from a built-in wallet with automatic top-up.
- Role-based access and a feedback board where users vote on features.
An AI agent team for acquisitions
Three agents, each with one job, built on an agent platform with Claude.
- Scout, chief of staff: watches the other agents, posts a daily briefing and keeps the operator informed.
- Alex, lead manager: turns inbound leads into booked appointments.
- Jordan, acquisition manager: takes warm leads from appointment to signed contract.
GoHighLevel stays the system of record. The Pulse portal supplies the figures, VAPI handles voice and Twilio handles SMS. All three agents are configured and validated in a local build. The production deployment is in progress.
What happens to a lead
Control and visibility
- Settings in one place. The team changes how the automation behaves from a dashboard, with no need to edit the workflows.
- Every run is recorded. Runs and their results are listed, so a failure is seen.
- The CRM stays the record. Every call outcome and stage change is written back to GoHighLevel.
- People take over when a deal is warm. The AI works leads up to that point.
Under the hood: stack
Follow-up and reactivation: n8n workflows, VAPI voice agents, GoHighLevel, Twilio and Supabase PostgreSQL. The dashboard is Next.js and TypeScript.
Pulse portal: React and TypeScript on the front end. Node.js, Express, Drizzle and PostgreSQL on Supabase behind it. OpenAI for the assistant and the recommendations. n8n for background jobs such as activity logging and nightly snapshots. Incoming GoHighLevel webhooks are validated before they are accepted.
Agent team: Claude (Opus and Sonnet tiers), Docker, GoHighLevel, VAPI, Twilio, Supabase and ClickUp for team messages.
The portal took five months: 811 commits, 121 database tables, 498 route handlers, 26 scheduled jobs and 17 integrations. It replaced a 138-node low-code workflow that could not refresh its own credentials. A daily health check caught three jobs that were failing without any error.
Are leads waiting in your CRM for someone to call them?
Tell us how a new lead is handled today on a 30-minute call. You leave with a plan and a cost, in writing.
Book a discovery call