Generative AI has become impossible to ignore, with tools like ChatGPT now dominating conversations across industries. Yet despite the hype, many companies are struggling to see results. A recent MIT study found that 95% of organizations reported getting “zero return” on their AI investments. The problem, according to Mark McDonald, president of Visual Lease and CoStar Real Estate Manager, is often not the technology itself, but how it’s adopted.
“How do you understand what the technology is that's in place?” McDonald tells GlobeSt.com. The question, he says, goes beyond technical diagrams or software architecture. It’s about understanding at a high level what’s actually being used and how it fits into existing systems. “People may just assume that it's in there and that it's going to work, as opposed to what is actually in there,” he explains. “How does it work? How does it interact with my current workflow?”
Another essential step is to evaluate how the software is trained. Unlike traditional applications that rely on fixed code to perform specific functions, generative AI systems require training on data to produce results. In some cases, an off-the-shelf model might be sufficient. But for most companies, McDonald says, effectiveness depends on training the system with their own data.
That data relationship also raises important security considerations. McDonald notes that more companies are now negotiating restrictions on how their software providers can access and use proprietary data. These limitations, he says, are becoming standard to prevent sensitive company information from being shared or exposed to other users.
Before making a major purchase, McDonald advises that companies run a pilot program to see how the software performs in real use. “Make sure that the vendor is very specific about where the AI is and acts throughout the individual customer workflow and the application,” he says. Buyers should know precisely what part of the product relies on AI and how that component engages with their data.
Pricing is another area that requires attention. “It can get expensive, especially if the data sets are large,” McDonald tells GlobeSt.com. “What’s the cost implication of me as a user adopting this AI that is now embedded within the technology?”
Even though generative AI often promises greater efficiency, McDonald cautions that people remain critical to the process. Companies should not assume that new tools will automatically replace staff. “AI technology needs people in the loop, even when the goal is to need fewer people to perform certain tasks,” he says.
Ultimately, ensuring that data is clean, consistent, and unified across departments may be one of the most important steps of all. Without that foundation, McDonald warns, even the most advanced AI systems will fail to deliver measurable value.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.