One Misconception in Early AI Adoption

It was great to catch up with some of you at the NFPA Conference & Expo in Las Vegas last week. We had a booth, and I had a small speaking event for the first time at NFPA. Quite the experience. 

The MeyerFire Booth at NFPA Conference & Expo 2026

TRENDS AND CHANGES WITH AI ADOPTION 
If you’re looking to get your hands on Trends and Changes with AI Adoption, where I talked about real changes we’re seeing today in fire protection with entry-level Large Language Models (LLMs) – you can see the recording if you attended the NFPA Conference here (https://nfpa.confex.com/nfpa/2026/meetingapp.cgi/Session/15945). 

If you weren’t at NFPA, be part of MeyerFire University and get our entire 2026 Leader Insights Series, where I talk on the concepts presented in more depth, as the course releases at the end of August. 

Trends and Changes with AI Adoption | NFPA Conference & Expo 2026

​SUPRISING TAKEAWAY
In the talk, I present practical use cases for entry-level AI (ChatGPT, Gemini, Claude) that we’re seeing actively used in industry today.

A key part of the material is looking specifically at the best use cases not being one-for-one replacement for what we do today with AI.

I think instinctively, we look at our processes we do today, and we tiral chat-based AI as a substitute for our thinking or our process. It doesn’t do that well. Part of that is AI doesn’t have our unique worldview, expertise, or experience.
But the other part of it is that noone actually wants AI to be a substitute for our judgment, for our communication, for our work, or for the things that make us most human (advocacy, trust, empathy, relationships). Clients don’t want us to spit out AI – they can just get AI from their AI. They want us and our experience and expertise to own our work and support them through a project. 

The best thing I feel we can do today is lean into the areas where we provide the most value, and that’s increasingly the human-side of what we do. 

What if we could be more responsive to clients?

More present in meetings? 

Have significantly better work output? 

Those are all things that are improvements over what we do today. 
 
THE MISCONCEPTION
The general misconception, and I hadn’t entirely wrapped my head around it until I spoke with many people last week before and after the talk – is that we naturally want to take AI today and simply swap out something we do with AI. It doesn’t work well, and it’s not what our clients want. 

What’s far, far better, is to have AI help us QC our own work. It’s instant, nearly-free, and it can give us a grade on our own work. 

AI-Generated work often feels inauthentic, less trustworthy, less human. Using our work, but instantly graded by AI in ways that we want to be graded, can help bring out our best authentic work that's human-decided, human-driven.
​

Here’s the key – we tell AI what it is we value. We tell it what to grade us on. What matters to us. That’s the key. Essentially, we create our own grading rubrics, and AI can help give feedback on what matters to us. 

If the report needs to be clearly communicated – if the change order proposal needs substantiation – if a proposal needs to be void of gaps – if the plan review needs code references, have AI grade us and tell us where we’re falling short. 

Not what the code reference should be (that’s a common AI hallucination and source of error today), but where in our work we don’t provide a reference (that’s really easy for AI to pick off). 

Large Language Models today work really well with two things – language and data. 

Despite being in a technical field, we have language everywhere – reports, emails, RFIs, plan reviews, meeting notes. We also have lots and lots and lots of data, though we usually don’t think of it that way. 

The last 10 years of the work we’ve done? The plan reviews we’ve sent? Our own boss’s review comments? Emails? RFI logs? Outside comments received? 

LLMs can make great use of that information, all internally, into our own QC checklists, grading rubrics, internal spreadsheets, internal dashboards, or internal tools that help us do far better work. 

Not replace a process, but build it better. 

There’s a huge difference there and the best applications we’ve seen thus far for fire protection and these entry-level LLMs are usually in that arena. 
 
Plenty more to come on this – if it’s a topic that interests you at all, I would greatly encourage you (that’s a midwest polite way to say seriously, check this out) to evaluate MeyerFire University for your team. We have so much technical and upstart coursework for new learners, but now we're NICET- and ICC-accredited CEUs and CPDs for senior-level teammates, 100+ PDF flowcharts, tables, cheat sheets, and checklists, and we’re adding our 2026 Leader Insights Series at the end of August. 

Most people I talk with are floored at how much has been added since they last looked. Come help your team get significantly better and be part of the community there – it costs less than you might think. 
 
Thanks, as always, for being part of what we do here! On the up -
​
– Joe