The short answer. Use AI for what it is good at, such as summarizing documents, extracting meaning and scoring risk, and keep clean, structured data and human judgment underneath it. Measure success by whether your organization becomes more aligned, not only by time saved. And point AI at the knowledge only you have: your own data and the people closest to your buildings.
In practice:
- Start with a specific problem, not a mandate to “do something with AI.”
- Put clean, structured data underneath before adding AI on top.
- Let AI handle probabilistic work, and keep people on judgment.
- Feed it your own ground truth: your numbers and your teams’ explanations.
Why so many AI projects in real estate fail
They skip the foundation
Boards are pressuring real estate companies to pursue AI, and many projects skip the foundation. AI is brilliant at probabilistic work: summarizing documents, extracting meaning and scoring risk. It is not a replacement for clean, structured data. AI projects usually fail not because AI doesn’t work, but because it is being used for the wrong things.
They aim too low
Saving a finance team two weeks on tax forms, or cutting rent roll analysis from hours to minutes, is useful. But if speed is the benchmark, the bar is too low. The bigger potential is alignment: connecting decisions across teams and sharing context about what is happening, why it matters and what needs to happen next.
General AI can’t see inside your buildings
AI tools learned commercial real estate from people who explained the industry in public. That makes them fluent on general topics like cap rates. But they cannot tell you why occupancy slipped at your property last quarter. That knowledge sits in your own data and with your property teams.
They don’t fit how people work
Under reporting pressure, people want help that fits the way they already work, with no disruption and nothing new to learn. Tools that ask for a big leap to a “better future” rarely get used.
What it looks like in practice
A Refined Data client acquired a significant portfolio with 10,000 documents needing due diligence review. AI extracted metadata and summarized the key findings for every document, and generated a risk score for each one. People reviewed and validated those scores. A relational database tracked everything, flagged outliers and kept the team on one source of truth.
- Review time was cut by over 90%.
- Every document had an accurate, complete risk score.
- The team worked from one source of truth.
- They had a production-ready dataset from day one of ownership.
How to get real value from AI, step by step
- Pick a specific problem. Name the decision or workflow you want to improve before choosing a tool.
- Put clean, structured data underneath. Use a database for what databases are good at: tracking, flagging outliers and keeping one source of truth.
- Let AI do probabilistic work. Summarizing, extracting meaning and scoring risk.
- Keep people on judgment. Have your team review and validate what AI produces.
- Point AI at your own ground truth. Your portfolio data, your variance explanations and what your property teams saw happen.
- Measure alignment, not only time saved. Ask whether teams are making better, more connected decisions.
How Refined Data helps
Refined Data puts AI inside the platform your team already uses, alongside structured data from Yardi or MRI. It is built to capture what general AI can’t: the explanation behind each number, from the people closest to your buildings.
Book a 30-minute consultation and bring your workflow and reports.
Frequently asked questions
Why are so many AI projects in real estate failing? Usually not because AI doesn’t work, but because it is used for the wrong things and the foundation of clean, structured data is skipped.
What is AI good at in real estate finance? Probabilistic work, such as summarizing documents, extracting meaning and scoring risk, with people reviewing and validating the results.
Can AI tell us why a property’s numbers moved? General AI tools can’t. They know the industry in general, but the reason a specific number moved sits in your own data and with the people closest to that building.
How should we measure the ROI of AI? Not only by time saved. The better test is whether the organization becomes more aligned, with decisions connected across teams.
Does AI replace a database? No. AI does the probabilistic work, and a database keeps clean, structured data and one source of truth. They work best together.
About the author
Carl Paulse is CEO of Refined Data and has spent over two decades in commercial real estate.
