AI Integration for UK Property Agencies

Estate agencies run on communication: email threads, WhatsApp messages, calls, CRM notes. The information needed to move a sale forward is scattered across all of them, and keeping the CRM true requires constant manual typing that agents do not have time for. So the CRM goes stale, and decisions get made on bad data.

A production AI layer that connects an agency's CRM, email, and WhatsApp into one system that understands what is happening:

Every inbound message is read, classified, and linked to the right property transaction

The CRM updates itself from the conversation instead of waiting for an agent to type

Outbound chasing and follow-ups are drafted automatically, in the agency's voice

A live dashboard shows the true state of every deal, because the data underneath maintains itself

Production AI systems fail in ways demos do not. This build needed reliable handling of messy real-world input (typos, half-threads, two deals discussed in one message), guardrails so nothing reaches a client unless the rules are met, and full audit logging so the agency can always see why the system did what it did.

Live in production with a UK agency. Claude as the reasoning engine, a TypeScript/Next.js application layer, deployed on Vercel, integrated with the agency's CRM, email, and WhatsApp Business.