When AI Relies On The Wrong Version

On today's show we're going to talk about a disease that affects almost every organization. It's a disease that can cost a ton of money, delay projects, create disputes between business partners, and undermine the quality of decisions. I've coined a new word to describe this disease. I'm calling it versionitis. Versionitis is what happens when a project becomes negatively impacted by someone relying on the wrong version of a document, a drawing, a spreadsheet, or some other source of information. The problem is not new. But artificial intelligence is making versionitis potentially much more dangerous. Imagine you're developing a 100-unit apartment building. The architect has prepared a set of drawings. Those drawings have gone through several revisions as the project moves through design development, permitting, and construction. At some point, the structural engineer receives a set of architectural drawings and designs the structural system around them. Meanwhile, the architect makes a change to the location of a mechanical shaft. The revised drawing gets distributed to some members of the design team, but not everyone. The structural engineer continues working from the earlier version. The discrepancy might not become apparent until the contractor discovers that a structural beam is occupying the same space as the mechanical ductwork. Now there's a problem. Someone needs to redesign the structure, reroute the mechanical system, or perhaps both. Depending on when the conflict is discovered, the cost could be relatively minor or extremely expensive. Nobody necessarily made a mistake in their engineering calculations. The problem was that two competent professionals were working from two different versions of reality. That's versionitis. If you give AI access to the directory structure on your computer, how do you know which files it relied upon for its analysis? ------------- **Real Estate Espresso Podcast:** Spotify: [The Real Estate Espresso Podcast](https://open.spotify.com/show/3GvtwRmTq4r3es8cbw8jW0?si=c75ea506a6694ef1) iTunes: [The Real Estate Espresso Podcast](https://podcasts.apple.com/ca/podcast/the-real-estate-espresso-podcast/id1340482613) Website: [www.victorjm.com http://www.victorjm.com/](http://www.victorjm.com http://www.victorjm.com/) LinkedIn: [Victor Menasce](http://www.linkedin.com/in/vmenasce) YouTube: [The Real Estate Espresso Podcast](http://www.youtube.com/@victorjmenasce6734) Facebook: [www.facebook.com/realestateespresso](http://www.facebook.com/realestateespresso) http://www.facebook.com/realestateespresso%5D(http://www.facebook.com/realestateespresso) Email: [[email protected] mailto:[email protected]](mailto:[email protected] mailto:[email protected]) **Y Street Capital:** Website: [www.ystreetcapital.com http://www.ystreetcapital.com/](http://www.ystreetcapital.com http://www.ystreetcapital.com/) Facebook: [www.facebook.com/YStreetCapital](https://www.facebook.com/YStreetCapital) http://www.facebook.com/YStreetCapital%5D(https://www.facebook.com/YStreetCapital) Instagram: [@ystreetcapital](http://www.instagram.com/ystreetcapital)

• 0:01Speaker 00

  Welcome to the Real Estate Espresso Podcast, your morning shot of what's new in the world of real estate investing. I'm your host Victor Manash. On today's show, we're going to talk about a disease that affects almost every organization. It's a disease that can cost a ton of money, it can delay projects, create disputes between business partners, and undermine the quality of decisions. I've coined a new word to describe this disease, and I'm calling it versionitis. Versionitis is what happens when a project becomes negatively impacted by someone relying on the wrong version of a document, a drawing, a spreadsheet, or some other source of information. Problem, of course, is not new, but artificial intelligence is making versionitis potentially more dangerous. Let me explain. Imagine you're developing

• 0:43Speaker 00

  a 100-unit apartment building. The architect has prepared a set of drawings, and those drawings have gone through several revisions as the project moves through design, development, permitting, and construction. At some point, the structural engineer receives a set of architectural drawings, and he designs a structural system around them. Meanwhile, the architect makes a change to the location of, say, a mechanical shaft. The revised drawing gets distributed to some members of the design team, but not everybody. The structural engineer continues working from the earlier version. The discrepancy might not become apparent until you're actually in the middle of construction, and you discover that you've got a structural beam occupying the same space as a mechanical duct. Now you've got a problem. Someone needs

• 1:24Speaker 00

  to redesign the structure to reroute the mechanical system, or perhaps both. Depending on when the conflict is discovered, the cost could be relatively minor, or it could be huge. Nobody necessarily made a mistake in their engineering calculations. The problem was that two very competent individuals were working from two different versions of reality. That's versionitis. The same phenomenon can happen in financial underwriting. Maybe you're evaluating an apartment development. Your financial model assumes construction costs at, say, 300 a square foot. After receiving updated contractor pricing, the budget increases to 325 a square foot. The development team updates the financial model to reflect the new pricing, but somewhere along the way, an earlier spreadsheet gets attached to an

• 2:07Speaker 00

  email. Maybe an investor downloads the spreadsheet and starts evaluating the opportunity. Maybe the lender is working with the wrong numbers. The numbers look reasonable. The formulas work, the calculations reconcile. There's no obvious indication that there's something wrong. But if you're relying on a financial projection that's based on assumptions that are no longer valid, you've got a problem. A calculation can be perfectly accurate with the underlying information being completely wrong. So now introduce artificial intelligence into that environment. AI has the ability to process huge amounts of data. It can review hundreds of documents, compare financial models, analyze contracts, and produce sophisticated reports in a fraction of the time that it would take a human to do that same amount of work. That's an extraordinary capability. But if you're giving your AI agent access

• 2:52Speaker 00

  to a directory structure, how do you know which documents it's relying upon to do its analysis? Imagine asking an AI agent to review the feasibility of a development project. The agent searches through the directory. It contains drawings, construction estimates, zoning reports, environmental assessments, financial models. It finds different versions of drawings, different versions of budgets, and different versions of rent studies. Which one should it use? Unless the information's been organized, and the agent's been given very clear instructions about document authority, it could select the wrong version, and you would have no way of knowing that. Whereas still, it might combine information from several different versions. It could use the building area from one set of drawings, the construction

• 3:33Speaker 00

  cost from another estimate, and the unit mix from a different feasibility study. The resulting analysis is going to look very beautiful. It's going to be well-formatted. It's just going to be wrong. This is where versionitis becomes dangerous in an AI-enabled organization. Historically, human analysts working with conflicting documents might recognize that something didn't make sense. They might remember a meeting where the unit count changed, or notice that a particular construction estimate predates a major redesign. An AI agent operating without that organizational context would have no reason to question the document that it retrieved. In fact, the quality of the presentation can make the problem even harder to detect. A polished report creates an impression of authority.

• 4:13Speaker 00

  The reader sees precise numbers. They see coherent explanations. They see a lot of math. And naturally, the assumption is the underlying analysis must be sound. But precision and accuracy are not the same. So this problem isn't new. We had this with humans. It's just now more acute with AI. In many organizations, documents don't have simple linear revision history. One spreadsheet gets copied into a folder, someone creates another scenario, a person downloads a local copy, and then modifies several assumptions. A third person might rename the file and call it final. But just because it's the most recent document doesn't mean it's the official release version. That might be an experimental copy, where someone's doing some sensitivity analysis. Before long, you've got a directory with numerous

• 4:57Speaker 00

  files named final, or final revised, final updated, final version 2. The word final has become meaningless. The traditional solution involves document control, revision numbers, approval workflows, and a clearly designated source of truth. But AI introduces an additional requirement. It's no longer enough for the correct document to exist somewhere in the organization. The AI agent has to be able to identify which version is the official version, understand when it became effective, and distinguish that approved information from preliminary working material. It should also be able to identify conflicts between sources, rather than just quietly choosing one of them. That requires a lot more than a clever prompt. It requires both an information architecture, and it requires you to query

• 5:41Speaker 00

  your AI agent to find out where it got its sources of information. And you need to audit that as part of your quality assurance. AI could be one of the most powerful tools for detecting versionitis. It can compare revisions. It can identify inconsistencies. And it can flag situations where assumptions are no longer matching approved drawings. But that same technology can also amplify the consequences of poor document management. We're entering an era where organizations can produce sophisticated analysis faster than ever, and it's amplifying the really boring topic of document version control. As you think about that, have an awesome rest of your day. Go make some great things happen, and we'll talk to you again tomorrow.