400+ qualified matches in three months: an AI property matching platform for a UK off-market brokerage
A UK off-market property brokerage connects vetted buyers with sellers whose properties never reach the open market. We built the platform that runs it: NDA onboarding, AI matching, AI-written teasers, a deal pipeline, AI valuations and a portal for users, partner agents and affiliates. It made 400+ qualified matches in three months.
- NDA signed
- AI matching
- Teaser sent
- Deal created
Snapshot
- Client type
- Off-market property brokerage
- Country
- United Kingdom
- Industry
- Real estate: residential, commercial, and land and development
- System
- AI Automations
- Use cases
- Matchmaking · client onboarding · lead nurture · AI teasers · deal pipeline · client and partner portal · referral tracking · AI valuation · internal comms
- Stack
- Airtable · Make.com · GoHighLevel · OpenRouter · Slack
- Timeline
- Built in phases, early 2025 to end of 2025
- Headline numbers
- 400+ qualified matches in three months · 67 deals in negotiation · < 60s from inbound lead to ranked matches
What did the client need?
Match vetted buyers to off-market properties fast, without exposing anyone's identity or the property's details too early.What did we build?
One platform in four phases: NDA onboarding and matching, an Airtable deal pipeline, a portal with logins, and AI valuations.How does it run day to day?
New buyer criteria match instantly; new listings match in one weekday-morning batch; every match lands in Slack as a deal.
What did the client need?
The brokerage doesn't list properties publicly. Every buyer and seller signs an NDA before seeing anything, and both sides stay anonymous until the deal is ready. Buyers include investors, developers and family offices, often with very specific criteria.
Doing this by hand meant chasing NDAs and criteria forms, keeping buyers and properties in spreadsheets, and searching those spreadsheets by eye every time something new arrived. Introductions were slow, matches were missed, and there was no single view of which deals were moving. As partner agents and affiliates joined the network, the brokerage also needed outsiders to add their own buyers and properties without seeing anyone else's.
What did we build?
We worked as the brokerage's technology team and built the platform in phases, adding each layer once the business needed it.
| Phase | When | What was added |
|---|---|---|
| 1. Lead nurture and matching | Early 2025 | NDA sign-up with a pre-filled NDA, buyer and seller criteria forms, the property database, AI matching with teaser emails, a daily batch run for new listings and reminder sequences |
| 2. Airtable CRM and deal pipeline | By mid-2025 | The pipeline moved from GoHighLevel into Airtable, every match became a deal, and partner agents were onboarded with intake templates |
| 3. Portal, partners and affiliates | By September 2025 | Logins for users, private contact management for partner agents, an affiliate dashboard with referral links, a support form, and sales logs and tasks on every deal |
| 4. AI valuations | End of 2025 | An AI valuation agent for every property, plus a form for one-off external valuations |
How the matching works. When a buyer's criteria arrive, one AI step turns them into an Airtable search formula. Airtable returns the candidates, and a second AI step checks each one against the buyer's brief and writes a teaser for every match. A match needs the property type, budget and location to fit, and the AI also weighs the buyer's free-text details.
Confidentiality in every view. Buyers see the area, category, price and teaser, never the address. Sellers don't see the buyer until a team member makes the introduction. Partner agents see only their own clients and deals, and matches for their buyers go to the agent, not the buyer.
How does it run day to day?
NDA sign-up ─► pre-filled NDA + criteria form ─► signed ─► AI matching ─► deal + teaser ─► Slack alert ─► relationship manager
- The NDA is the gate. Nothing is sent until the NDA is signed, and the NDA is pre-filled so the lead only has to sign.
- Reminders until they finish. NDA signatures and criteria forms each get reminders after 3, 7, 14 and 28 days, then 3 and 6 months. Before every reminder, the automation checks whether the step is already done.
- Instant for buyers, batched for listings. A new buyer's criteria are matched straight away. New listings are matched in one run on weekday mornings, around 9–10 AM UK time, so buyers get one round of matches rather than a flood of emails.
- Only verified records match. Only Active buyers and Available properties are matched automatically.
- People build the trust. Every new match posts to Slack. A relationship manager then qualifies the buyer on proof of funds and intent, makes the introduction and guides the deal to completion.
- AI valuations as a first pass. The valuation agent combines a property's data with live market research and returns a price range and its reasoning, which the team reviews.
What were the results?
The platform made 400+ qualified matches in three months, with 67 deals in negotiation, and matches an inbound buyer lead against every live listing in under 60 seconds. The brokerage reported that most AI valuations were accurate, and the finished platform became the template for a second UK brokerage.
- Qualified buyer-to-property matches in three monthsPhased build, 2025
- Deals in negotiationPhased build, 2025
- From inbound buyer lead to ranked matchesPlatform matching speed
What did we learn, and what are the limits?
- Build in phases, and let real use decide the next one. The first version ran entirely through GoHighLevel and Make.com. Real use showed the deal pipeline belonged next to the data, in Airtable, which made matching and deal creation a single step.
- Let one AI step write the search and a second judge the results. It kept matching flexible without hand-coding every combination of criteria.
- Every confidentiality rule needs its own view. A policy isn't enough. Each rule (no addresses for buyers, no buyer names for sellers, partners see only their own clients) needed a filtered interface.
- Batch new listings, match new criteria instantly. It stops buyers getting a burst of separate emails when several properties arrive at once.
- Users want visibility, not control. Letting buyers and sellers follow their deal's stage, without moving it, cut "where are we?" emails.