Based on McKinsey’s latest research on AI and operating models in real estate.
Most real estate companies have tried AI. Plenty have run pilots. Some have seen demos that looked genuinely impressive. Very few have seen any financial impact.
McKinsey calls this “pilot purgatory” — and according to their latest research, it’s where the vast majority of real estate firms are stuck right now.
So what separates the 20% that are actually getting value? It turns out to have very little to do with which AI model you’re using. It has everything to do with how your workflows are designed.
The problem isn’t the technology – it’s your operating model.
McKinsey makes a sharp observation early in their research: when a pipe bursts at 6am, your tenant doesn’t care whether you’re running Gemini or Claude. What shows up is your operating model — your workflows, your response systems, your ability to execute under pressure.
The AI is only as useful as the processes it runs inside.
Right now, most real estate workflows weren’t designed with AI in mind. Work orders arrive through fragmented channels. Maintenance coordination is manual. Tenants are left wondering whether anyone received their request. And there’s no learning loop — no way to capture what worked, who the reliable vendors were, or how to get better over time.
AI can fix all of this. But only if the workflows are built to let it.
Pick a domain, not a use-case list
One of McKinsey’s most practical insights is what they call the “Goldilocks mindset.” Most companies go either too small — collecting individual AI use cases like a checklist — or too big, launching vague enterprise transformation programs that never ship.
The right scope is a domain: a connected slice of work that’s big enough to matter and focused enough to actually change. Think leasing and renewals, maintenance and facilities, or asset management reporting. Each has a clear business owner, measurable KPIs, and messy handoffs that AI can meaningfully clean up.
One well-designed domain delivers more than twenty disconnected pilots.
Automate the steps. Protect the judgment
Once you’ve picked a domain, McKinsey recommends splitting every workflow into two categories:
- steps; repetitive coordination tasks like routing tickets, scheduling vendors, logging outcomes
- thoughts; human judgment, relationship-building, high-stakes decisions.
The goal isn’t to replace people. It’s to return human time to the moments that actually build trust and drive value, while AI handles the coordination overhead that shouldn’t require a person at all.
Trust has to be designed in from the start
McKinsey’s third insight is the most overlooked: trust isn’t a nice-to-have — it’s a product requirement. AI agents need traceability. Workflows need audit trails. Humans need to stay in the loop on exceptions, escalations, and anything involving significant money or relationships.
The firms seeing the most success build governance into the system from day one: clear automation tiers, defined guardrails, and feedback loops that make the system smarter with every decision.
Bolt trust on afterward and you’ll spend more time explaining failures than capturing value.
The compounding advantage
Once you’ve redesigned a domain well, you own something your competitors can’t easily copy: proprietary data, learning loops, and reinforcement from thousands of real decisions.
AI tools will eventually be commoditized. The organizations that built the right operating model early — connected data, intentional workflows, earned trust — will have a durable advantage that compounds over time.
McKinsey estimates the real estate industry could capture $430–550 billion in value from AI. The technology is capable. The only question is whether your workflows are ready to let it act.
See ProptechOS in action — book a 30-minute demo
