SMS-first AI lead nurture with agent handoff

Deliverables

• Twilio SMS trigger and reply path
• Supabase message history with the last 10 messages as context
• OpenAI prompt that returns a reply and qualification data
• Lofty status update to Ready for Agent with an agent SMS alert
• Agent reply webhook with approve and ask-for-more paths
• Daily summary and quiet-lead re-ping branches

The challenge

Real estate leads expect a quick reply, often by text, but agents can't answer every message the moment it arrives. Leads that wait go cold, and agents spend time on people who are not ready to move. The team needed fast, consistent replies with a clear point where a person takes over.

Our approach

A Twilio trigger receives each inbound SMS. The workflow extracts the lead's message and number, saves it to a messages table in Supabase and fetches the last 10 messages for context. A prompt with system instructions, examples and that history goes to OpenAI, and the response is parsed into the reply text and qualification metadata. The reply is saved back to Supabase and sent to the lead through Twilio. Qualification logic then checks the lead: if qualified, its status in Lofty is set to Ready for Agent and the agent gets a hot-lead SMS. The agent answers by text to a separate webhook; the workflow finds the lead by the agent's phone number and either confirms the booking with the lead or, if the agent asked for more information, passes the question to the AI and texts the lead a follow-up. The design also includes an end-of-day summary texted to the agent, a check that re-pings leads who have gone quiet, and an optional media add-on that turns summaries into Gemini images and short videos for Facebook.