AI and APIs in residential building management

Ilia Sotonin, founder of Unitify, ran a residential management company for about 20,000 residents and built doma.ai, the platform used by 10,000 management companies across more than 5 million apartments, before exiting in December 2025. He is now rebuilding the same kind of software from scratch for AI. His main argument: AI helps residents and managers only when the building's processes are digitized and its data is open through an API, so the resident-facing app should come last, not first.

Processes first, the app last

In 2013 we launched a resident app for the iPhone. It may well have been one of the first resident apps anywhere. It was also a mistake. As I wrote in 2018: "The plumbers and electricians didn't have iPhones." Behind the nice screen, "the same paperwork, or calls to a plumber from the dispatch office" remained.

In 2026 I put it more bluntly: "The resident app is the thing you should do last." Because "even if you have the most beautiful mobile app with the nicest buttons," a resident presses "Call a plumber" and nothing happens. The rule I took from my own mistake: "First digitize the processes, then the interface for talking to the resident."

What digitization means here I wrote in 2021: "Digitizing a management company is not a smart home and not a resident app. It is setting up business processes so the company can earn more money by becoming more efficient." Our book says the same: do not automate chaos. Run the process by hand, check that it works, then automate it. Automation does not fix bad management. It makes bad management visible.

The resident data layer and the "Plaid moment"

Today paying the bill is often the only digital contact a resident has with their building: open the banking app, pay, close it. But life in a building is much more than that. Requests, meetings and votes, access, meter readings, the building's money and its equipment all live in separate systems, or in nobody's system.

My view, in my own English words from 2026: "I believe residential real estate will face its own Plaid moment very soon." And: "Before AI can help residents and property managers, buildings need an open API layer." AI agents are starting to replace apps as the way people get things done. Whoever owns a clean, standard data layer for the building is the one those agents will talk to.

This is not a new idea for me. In the book's automation chapter, formalizing an API for outside systems sits in the first-priority tier, next to the resident account and ahead of the marketplace and sensors. doma.ai was built on open source for the same reason. I wrote in 2023: "Usually everyone hoards their product, like Gollum over the ring." Closed systems lose once agents need to read and write data.

The AI dispatcher

The first place AI pays off is the dispatcher's desk: taking requests from every channel, sorting them, routing them to the right person and keeping the resident informed. In 2026 I wrote: "AI doesn't replace the dispatcher, it complements him: AI removes the routine, so the person can do what really requires a person." And: "Our task is not to replace people with algorithms. Our task is to make working at a management company look like working at a normal technology company."

The economics follow from staffing. One community manager can handle about 1,000–1,500 units. That ceiling is set by routine contacts, not by hard decisions. Take the routine away and the same person can look after more homes without the service getting worse. At Ande Tower in Batumi, requests are handled three times faster since the building moved to Unitify.

The next step is the building itself. In 2025 I asked: "If a person can create a task, why can't a pump?" Each piece of equipment can report its state and open its own request before a resident notices anything.

Rebuilding the software with AI

For most of my career, code was a black box to me. In 2026 I wrote: "Since writing code was a black box for me, I couldn't verify the timeline, the cost or the stack." That has changed. I now write much of Unitify's code myself with AI tools. In January 2026 I noted: "Now working on AI dispatcher, alone." We did not adapt the old system. We rebuilt it from scratch, because software designed for humans clicking through screens is not what agents need.

Questions people ask

How is AI used in HOA and condo management? Most usefully as a dispatcher and concierge: taking requests and questions from every channel, routing them, answering routine ones and keeping residents updated. It also helps with billing reminders and reports. In my experience it works only on top of digitized processes; without them, AI just answers faster that nothing will happen.

Will AI replace community managers? No. It removes the routine that caps how many homes one manager can look after, today about 1,000–1,500 units. The manager keeps the work that needs judgment and a human voice. What changes is how many residents one good person can serve well.

What does a "Plaid moment" for real estate mean? Plaid gave apps one standard way to connect to bank accounts. I believe residential buildings need the same: an open API layer for requests, payments, access and building data, so AI agents acting for residents and managers can work with any building. Openness has been part of my approach since we built doma.ai on open source.

Should a management company start digitizing with a resident app? No, the app should come last. I made that mistake in 2013: residents had an app, but behind it the work still ran on paper and phone calls. Digitize requests, billing and dispatch first, and add the resident interface once the work behind it actually happens.

Ilia Sotonin is open to conversations on this topic: [email protected]