Agents With Eyes: Thesis Driven's Deep Dive on SurfaceAI

Thesis Driven's Brad Hargreaves just published a deep dive on SurfaceAI, the multifamily operations platform we built with founder Jason Wallis. His framing is worth sharing, because it captures something we believe about where vertical AI is headed.

Most AI in real estate has focused on text: chat tools, drafting, document summaries. But multifamily operations don't live in clean text. Leases run 130 pages with handwritten initials in the margins. Vendor contracts arrive as scanned faxes. Inspection reports are photos taken on an old phone. The information needed to run a property is real, but its format has always been hostile to software.

Computer vision is what makes that data legible. Hargreaves' point is that the technology has quietly matured to where it can pull structured fields out of a messy document, distinguish floorplan elements, or flag conditions on a camera feed. Pair that with language reasoning and an agentic layer that decides when to act, and you get a system that can see a document, understand it, and do something about it. SurfaceAI has been building on that stack since 2023, before any single piece was production-ready on its own.

The clearest near-term proof is lease audit. Across a large portfolio, small mismatches between what a lease says and what the property system charges add up: an escalation that never triggers, a concession that should have burned off, a waived fee that quietly persists. Operators using SurfaceAI for lease audit report a 6-7x return relative to software cost. The novelty isn't the audit itself; it's running it continuously and autonomously rather than as a once-a-quarter consulting engagement.

The deeper idea in the piece is a shift in what software is. Traditional real estate tools assume a human sits at the controls, navigating menus and exporting reports. SurfaceAI inverts that. Wallis describes the product as an ombudsman that inspects the operation around the clock and surfaces only what needs a person's judgment. "We want to be their watchdog," he says. The dashboard stops being an operating surface and becomes an exception-handling one.

That's the kind of company we set out to build at Stackpoint: agent-native and grounded in a specific industry's messiest workflows.

Read Brad's full deep dive on Thesis Driven →