One deep engagement. No case-study theater.
The practice began inside a real estate firm. Names redacted. What follows is the before, the build, and what the team can do now that they couldn't do six months ago.
Real estate acquisitions & investor relations2026Ongoing since 2024Anthropic + Attio + GoHighLevel + Microsoft 365 + Supabase
One workbench for a firm that was running on five tools.
A growing real estate investment firm had stitched together GoHighLevel for investor relations, Attio for acquisitions, Microsoft 365 for everything else, and a pile of shared spreadsheets. We built a single workbench that pulls every deal document, email, and conversation into one place — drafting LOIs, running underwriting, flagging due-diligence risks — with a memory layer that compounds as the team grows.
A small team managing rapid deal flow across three systems that did not talk to each other. Context fragmented across GoHighLevel, Attio, and Microsoft 365. Junior team members re-reading entire email chains to reconstruct a deal's history. LOIs drafted by hand from a template; underwriting rerun in spreadsheets for every new property. No shared memory — when someone left, the context left with them.
A workbench inside their own stack — not a rented dashboard. Connections into GoHighLevel, Attio, and Microsoft 365, with deal documents and email threads indexed together so any team member can ask “what happened with this property” and get a real answer. Claude-backed agents draft LOIs from the underwriting inputs and surface due-diligence flags from the property documents. A custom task system — deliberately not Monday or ClickUp — because the tasks follow the deal, not the person. A memory layer that learns as the firm learns: every closed deal, every broker interaction, every “we won't touch that asset class again” decision gets woven into what the system expects next.
Sources
Process
Output
Underwriting turnaround cut from half a day to minutes. LOIs drafted in the time it takes to review them. Due-diligence flags caught before human review. New team members onboarding in days, not months — because the system has the context. The memory layer now answers questions the team didn't know how to ask six months ago.
In the six months since
- Memory layer now answers “did we ever look at this broker” with context.
- Underwriting turnaround cut from half a day to minutes.
- Task system adopted across acquisitions, IR, and asset management.