Software development has changed faster than almost anyone in the industry predicted. Google reported in April 2026 that 75% of all new code at the company was AI-generated and approved by engineers. Microsoft put its own figure at 20 to 30% the year before. Across the broader market, DX’s Q2 2026 report found that more than nine in ten engineering organizations now use AI in development work, with roughly 30% of all merged code written by AI.

The more consequential shift is who can participate. Writing functional software used to require years of training, which meant the gap between having an idea and having a product was measured in hiring cycles. That gap has narrowed considerably. People who understand a problem deeply but have never written code can now produce working applications, which changes the economics of software creation in ways the real estate industry is only beginning to absorb.

For real estate companies, the practical effect lands on the buy or build decision, and it pushes in both directions at once.

On the buy side, there is simply more to choose from. Products that would have taken a funded team a year to develop can now be built by two people in a few months, and the proptech market has filled accordingly. More options is generally good. More options of uneven quality is more complicated.

“We have noticed in the last few years during the AI boom is that there are more and more startups that think they can build a PropTech product,” said Zaid Tirmizi, Product Manager at TechnBrains. The barrier that used to filter out weaker entrants was technical capability. That filter has largely dissolved, which means the market now contains a much wider range of products built by teams with a much wider range of understanding about the problems they claim to solve.

That shift relocates where the real advantage sits. When building software was hard, knowing how to build it was valuable. When building is comparatively easy, knowing what to build becomes the scarce input.

“Real estate companies have an advantage, they know what problems that they need to have solved,” Tirmizi said. That knowledge is accumulated through operating buildings, managing portfolios, and working through the specific failures that show up in daily practice. It is not available to a founder who read a market report and decided property management looked inefficient, and it cannot be acquired quickly.

Which is the argument for building rather than buying, at least in certain circumstances.

“If a real estate company has a lot of proprietary data but hasn’t consolidated it yet, they should consider building their own products,” Tirmizi said. The logic is specific. A company sitting on years of transaction history, maintenance records, or operational data has something no vendor can replicate, and a tool built around that data can do things a general-purpose platform cannot. Previously the development cost made that impractical for most operators. Now the calculation is different.

The calculation is not as different as the marketing suggests, though, and the gap between writing code and shipping software is where most of the cost still lives.

“You will need engineers to solve the technical debt,” Tirmizi said. “Writing the code is done by AI but comprehension of code is still a challenge. AI can write in a lot of garbage into code and going through it takes time.”

The data on that point is fairly consistent. LinearB’s 2026 benchmarks, drawn from 8.1 million pull requests, found that AI-generated code contains 1.7 times more issues per pull request than human-written code. Veracode’s 2026 security report tested more than 100 models and found an average security pass rate of 56% that has not improved across model generations. Stack Overflow’s 2025 survey found that 66% of developers spend more time fixing almost-right AI code than they would have spent writing it themselves, and 45% say debugging AI-generated code takes longer than debugging anything else.

Code churn tells a similar story. The share of code revised within two weeks of being written rose from 3.1% in 2020 to 5.7% in 2024, tracking AI adoption. Writing is faster. Everything downstream of writing is not.

What that means in practice is that a real estate company building internally still needs engineering capacity, just allocated differently. Less time producing code, more time reviewing it, maintaining it, and dealing with the accumulated consequences of decisions nobody made deliberately. Organizations that budget for the first and not the second end up with a working prototype and no path to running it in production.

Where the change is unambiguous is in the cost of finding out whether an idea works at all.

“This is a great time to test an idea,” Tirmizi said. “You don’t have to launch it to a huge market, just build it and get some users on it to see if the idea is working. This is a very good time to fail fast.” That applies to founders and to people inside large organizations equally. An asset manager with a theory about a better underwriting workflow can now build something testable without a budget request, a vendor evaluation, or a six-month development timeline. Most of those experiments will fail, which is the point. Failing in three weeks for very little money is a different proposition than failing in a year after committing real capital.

The labor market is adjusting to all of this, and the early movement is visible in outsourced development.

“We are already seeing the impact on outsourced developers. It has gone down by probably 30 percent already,” Tirmizi said. That tracks with broader labor data, which shows the effect concentrated rather than uniform. Stanford’s payroll analysis through June 2026 found no widespread displacement overall but a 19% employment gap for workers aged 22 to 25 in occupations where AI automates rather than assists. Demand has moved up the seniority curve. Junior work that involved implementing well-specified requirements is exactly what AI handles best.

For real estate companies, cheaper development labor compounds with cheaper development tooling, which makes internal builds more viable than they have been. Whether that produces better software is a separate question. The companies that will get the most out of this are the ones that understand their own problems well enough to specify what they actually need, and that have the engineering judgment to tell whether what comes back is sound. Neither of those is something AI provides.

The post What Happens to Proptech When Anyone Can Build Software? appeared first on Propmodo.

​