Operating the built world

The built world is the places where people live, work and gather. It is a cornerstone of American life, and it drives up to 18% of the country's GDP. Developers shape our skylines. Operators run the buildings we wake up in every day.

For decades, the industry has kept the lights on with people. The operating model hasn't changed much since the mid-1900s. Property management companies hire large teams, who push through complex workflows and endless edge cases inside software built in the '70s and '80s.

That model is expensive, and it's breaking.

The operating model is breaking

Running a property costs more every year. Roughly 40 cents of every rental dollar goes to operating expenses. In residential real estate alone, payroll and administrative costs add up to about $100 billion a year, and operating costs regularly grow faster than rent.

Hiring is getting harder too. Open roles in real estate are growing at three times the rate of other industries. Once roles are filled, people don't stay: attrition runs around 16%, against a national average of 6%. The job means living near the asset, mastering bespoke software and putting out fires all day, for limited praise and pay. And more than a quarter of properties change management or ownership every year, which brings new systems, processes and people each time.

The median property manager is barely breaking even. Operators have responded by centralizing: consolidating admin work across properties and creating specialized roles. That relieves the symptoms. It doesn't fix the problem, because people can't keep being the band-aid for day-to-day operations.

Half the job is paperwork

Property teams split their time between two kinds of work:

• Work that needs a person there: leasing tours, inspections, repairs, community events.
• Work that doesn't: AP data entry, budget reporting, work order triage, lease administration.

The second bucket is high-volume and repetitive, and it follows standard procedures. Yet staff spend up to 40% of their week on it.

Today's property management software is, in practice, a digital wrapper around human labor. There are three-hour training videos on how to create a single purchase order in a popular system. At one of our first customers, accounting policies lived in spreadsheets, and AP staff worked in cubicles covered in sticky notes reminding them of vendor codes and invoice follow-ups.

Software couldn't do this work before because it happens everywhere at once: across browsers, spreadsheets, phone calls and inboxes, and it takes human context and judgment to finish. That's exactly what has changed.

AI can finally do the work

Large language models mean software can now do the work itself, not just record that a person did it. Real estate is unusually well suited to this, because so much of running a portfolio is administrative rather than physical.

But better software alone won't change how operators run. Most tools ask teams to change how they work to fit the product. We do the opposite.

How we work

Unitbase works alongside operators as a deployment partner. We embed with their teams, map how the back office actually runs, and build an agentic system around the way people work today, in their inboxes, browsers and systems of record. Then we standardize: every property follows the same process, every invoice gets the same checks, and nothing depends on one person's sticky notes.

Three things make this work:

• Context. Operational knowledge lives far beyond the system of record, in email threads, vendor contracts, spreadsheets and people's heads. Our agents read all of it, structured and unstructured, and act in the systems where the work gets recorded.
• Trust. AI only creates value if people use it. Early on, agents work with a human in the loop. Teams review and approve the work, correct it when needed, and the agents learn from every correction.
• Better decisions. When the work runs through agents, every invoice, vendor and dollar is visible across the portfolio. Operators can see where spend is drifting, which vendors charge more for the same work, and what to act on next, instead of finding out at month-end.

Why we started with the back office

We started with financial operations because it's mission-critical, and because AI makes it dramatically better on day one.

Financial operations touch nearly everyone. Accountants aren't the only ones processing invoices: every property manager does, and maintenance supervisors pick vendors and manage purchasing. Meanwhile, owners and asset managers are under real pressure to control budgets as operating costs climb. Some operators pay property managers cash bonuses of up to $2 per unit per month just for hitting budget targets, which is more than they spend on most software.

An AI back office changes the math. It handles the vast majority of the time staff spend on data entry and GL coding. And because it reads every line, it can do what a busy team can't. It checks every invoice against contract pricing, flags junk fees and makes sure early-pay discounts are captured. Every invoice becomes a chance to spend less.

What comes next

The back office is where we start, not where we stop.

The operators we work with run every core workflow across their portfolios: maintenance, tenant support, marketing, leasing. That work happens the same way finance does, across email, spreadsheets, browsers and text messages. Every agent we build and every integration we add carries over to the next workflow.

This path isn't new. Decades ago, the platforms that run most of real estate today started with accounting before expanding into the rest of operations. We think AI makes that path faster, and the result better.

The end state is an AI back office for real estate. It grows with the portfolio, builds context every day, never falls behind, and works around the clock. Property teams get their time back for the work that needs a person: residents, buildings and vendors.

That's what we mean by operating the built world.