For the next few weeks I’m going to look at AI from a lot of different angles. The hope is that you and your team can learn from our journey, and maybe take a few larger leaps because of it. 

Let’s start at the beginning of my own journey, which is inextricably linked to the Roth IAMS (my professional services team) and SLAM Technologies (my asset management software team) corporate journeys. 

I’ve always been an early adopter. I just find it fascinating to see what can be done that couldn’t be done before. So when AI leapt onto the scene, you’d think I’d have been first in line. Instead, my reaction was “not on my watch!” 

My concern was our intellectual property. I’ve been in the facility and asset management consulting business a long time. Since Roth IAMS started in 2014 and SLAM in 2018, we have invested heavily in developing IP that helps us do our work more efficiently and effectively. 

In the early days of AI, at least as I understood it, anything you entered into a model became fair game for training that model. I could see the potential benefits, but publishing our IP to get them was not a trade I was willing to make. So we put out a policy: no staff were to enter any of our IP into an AI model. In practice, that forbade its use for any material business purpose. 

The tech nerd in me couldn’t ignore AI entirely, so I kept playing with it on my own time. Recipes. Vacation ideas. Nothing that mattered. 

That held for the better part of two years. Two years where we got nothing out of AI, because I had made sure we couldn’t. While I can’t say exactly what that cost us, I do regret not learning the lesson I’m about to share a lot earlier. 

I can’t tell you exactly when the technology caught up with my concerns, but I can tell you exactly when I found out. It was a presentation I attended through my TEC group (Vistage in the US) in early 2025, where I learned you could run a model in an environment where your inputs are never used for training. 

This was the light bulb moment for me! All the benefits of the tools, without giving away IP we’ve spent millions of dollars developing. I came back from that session, scheduled a series of AI strategy sessions with my leadership team, and we were off to the races. 

Let’s revisit something I said last week. A no today often means not yet. My 2023 “no” was the right answer in 2023. It was the wrong answer by 2025, and I held onto it longer than I should have. 

So if you or your organization have concerns about protecting your data or your IP, know this. You can solve it. Your IT team can set up an environment where your data is protected; whether that’s your own hosted instance or an enterprise agreement with the right terms in it. The two options look very different on the invoice, so find out which one you actually need before you go asking for budget. 

Here’s the part I got wrong, though, and it’s the real lesson of this post. 

My first instinct was to write a policy. Policies depend on people reading them, remembering them, and following them under deadline pressure and the weakest link in any chain is always people. Don’t leave your IP protection to chance based on whether individuals remember to choose the right setting. 

Build the environment so that the safe thing is the only thing available. Then let everyone loose to play in it. 

Build the sandbox first. Then hand out the shovels. 

Next week I’ll turn to AI policy. That isn’t a contradiction, the sandbox sets the walls, the policy tells people how to behave inside them. You need both, and I’d argue you need them in that order.