Weekly Briefing

article sponsor image
Partner Content

AI Can’t Fix Bad Data. These Ideas Can Get You on the Right Track.

May 29, 2025 | Carla Hinson, VP, Solutions & Innovation, MRI
5 min read
listen to article Listen to this article

Real estate visionaries constantly integrate innovative technology to make their organizations more efficient. The most transformative technology today is artificial intelligence (AI). However, AI success hinges on a critical factor: data quality.

AI can do many things, but fixing bad data isn’t one of them. AI uses your data as it exists today to populate models for anomaly detection, forecasting, and other business functions. AI is only as accurate as the underlying data from your lease agreements, occupancy information, maintenance records, financial statements, and other sources.


More concerning are the “hallucinations” when AI algorithms detect inconsistencies and conjure missing values to continue the analysis. Say you want to estimate rent income based on your portfolio’s square footage. If some properties don’t have that data, the AI might use an average instead of your actual total. Inaccurate and incomplete data yields AI-generated results that aren’t useful – or even worse, lead your teams to incorrect conclusions.


These questions prepare your data for AI

Successfully implementing real estate business AI depends on ensuring your data is up to the task. It should cover your operational, financial, and property management functions and be complete and high-quality. Only then will you get an accurate picture of what’s happening – and what could happen – across your portfolio.


If you’re not sure if your real estate data is ready for AI, these questions can help.


Do teams trust your data?

The first step is ensuring your AI can access all correct data sources. If the model can’t use the right inputs, you can’t trust the results. You must avoid the “someone has a spreadsheet somewhere” problem: when someone owns a critical dataset that must be updated regularly, but these updates happen outside the normal processes.


Take forecasting budget vs actuals. The rent runs powering this analysis might be exported to an Excel file. If someone managing that data finds an error, they could fix the error in the file before it’s sent on for payment. However, the error still exists in your original data where AI can find it, leading to unreliable forecasting.


You first need to determine why data fixes are happening outside your regular processes. A great place to start is identifying data gaps. You should define exactly what data is needed to fuel an AI-driven analysis, and then check all data fields for accuracy and completeness. (This is a typical step with organizations requesting an MRI Software Health Check to improve their PropTech efficiency.)


Once you know where gaps exist, you should determine how to optimize the data for AI. The best option is fixing the gaps by filling in any missing information. If that’s not possible, then specifying where to run your AI (e.g., avoiding known gaps) can help. Moving forward, you can prevent future gaps by requiring critical fields to be completed when people enter data.


Are your data formats standardized?

Next, you must ensure your AI can understand the real estate data across different systems consistently. Your data needs uniform naming conventions, date and value formats, and categories.


Inconsistently formatted data increases your risk of AI-powered anomalies. You may get nonsensical or incorrect recommendations that are easily spotted. But you could also get results that appear factual but aren’t based on your data – and you’d never know. Recognizing bad results means incurring the costs of fixing your data and re-starting your analysis. If these faulty outputs go undetected, you face not only the costs of poor decisions, but also the opportunity cost of missing the best choices for your business.


Try these tips to standardize your data for AI:

  • Clearly define data types. Establish each data type’s context as well as its source of record. For example, “rent” could be defined differently in your general ledger and property management system. The AI should analyze current numbers from only one source.
  • Use standard date formats across your systems.
  • Confirm correct currency and other number formatting. Sometimes numbers can be saved as text or the wrong data type, which can confuse AI models.
  • Use a common, unique identifier for each property in your portfolio.
  • Similarly, use common status definitions. To avoid data entry variations, codify the status options in a dropdown menu.
  • Use consistent nomenclature to group classifications correctly for analysis. For example, specify location types (office, lab, etc.), lease options (renewal, termination), and site characteristics (mixed use, residential, commercial, retail type, school proximity, etc.). Correct data mapping provides more accurate answers to questions like “What are my most successful store types?” Coding options in a dropdown menu also helps here.

Is your data current?

Acting on old data often means missed growth opportunities. So, your next step should be ensuring your data is current. First, find out how often the data you need is updated. Anything you download from a source is immediately stale. Reducing the gap between updates will yield more accurate results from your AI.


Note that “current data” doesn’t necessarily mean “real-time data.” The ideal update frequency depends on the context in which you use the data, and the analysis needed. For example, your daily workflow involves processes where you need to act immediately. Artificial intelligence support works better here with real-time data. Forecasting and other long-term analysis still work well with less frequent, “nightly load” updates.


Further refining your data

Implementing AI with data that isn’t ready risks data hallucinations and poor results. It also diverts your team’s time and focus to fix issues. You can avoid these costly mistakes by asking the questions above before your organization begins using AI. You’ll have a better understanding of where your data stands and the actions you need to take to gain the AI-driven insights your real estate vision deserves.


These are just a few ways to build a solid foundation for real estate artificial intelligence. Please read my recent blog post for more questions that will help you get your data ready.

Carla Hinson, VP, Solutions & Innovation, MRI
Carla Hinson is Vice President of Solutions and Innovation, North America at MRI Software. She leads go-to-market strategy and is responsible for product management and development of the North America product suite encompassing mid-market to enterprise solutions. Carla has over 25 years of commercial, tenant, and retail PropTech experience. Before MRI, she was Executive Managing Director, GCS Global Technology for Newmark and served as a COO and Services Director for multiple consulting and software companies.

This Week’s Sponsor

MRI Software is a leading provider of real estate software solutions that transform the way communities live, work, and play. MRI’s open and connected, AI-first platform empowers owners, operators, agents, and occupiers in commercial and residential property organizations to innovate in rapidly changing markets. For more information, please visit mrisoftware.com.