Designing the Future of Our Work (with AI), Today

By Joe Meyer, PE | Fire Protection Engineer 

We’re at a point with AI today where I think it’s worth saying out loud what we want the future of design and engineering to be.
 
For a period of time since ChatGPT surfaced in 2022 to today, much of what we, as an industry, have created with AI’s help has been creation that has happened because we now can do things.
 
The international dialogue about AI is talking daily about what we can create but also what we should create.
 
And before you think this is a moral lecture, it’s not. You’ll see in a minute where I’m going with this.
 
My core question today – where do we want to place our human touch in the future of our practice?
 
I don’t want us all casually watching AI happen before us, arriving at a future of design where we're mostly bystanders. I think our ideal future of work is when people initiate key decisions, not simply review them.

INCOMPREHENSIBLE OPPORTUNITY
For those who aren’t regularly working with AI multiple times a day, I’m not going to fear-monger and say you’re behind. You’re not. You’re the norm, not the exception.
 
For those who are in the throes of AI, you already know that our ceiling is completely blown out.
 
If we were walking a metaphorical hallway of opportunity, a few years ago we’d open a door (of AI opportunity) and see that there were a few more doors available to us. We open a door here or there and find closets, but then we’d open one and a giant room would open up. Today, working with any of your pick of LLMs, we’re opening one door and finding a Monster’s-Inc-style gigantic warehouse of doors.
 
If you can think of it today, it’s probably possible to make it. Today.
 
I’m not even talking some $100k worth of tokens or senior software engineer stuff – just using ChatGPT, Claude or Gemini. It’s unbelievably wild what can be created right now. It’s something I can’t fully wrap my head around.
 
We recently created a new subsidiary of MeyerFire called MeyerFire Labs, LLC, where we’ve been exploring and tackling a wide variety of opportunities related to improving how work gets done (software, tools, workflows). In doing so, we keep coming back to can vs. should, and what we want our field to be in the near future.
 
CAN VS. SHOULD
Once we come to a realization of the scale of the opportunity today – our instinct is to build-build-build.
 
What do I do that can be automated?
 
What do I do that can be replaced by my AI?
 
Many of us have tried this. Many of us have been disappointed with putting in something straightforward (to us) and getting a response that’s less than our skill. There’s an entire mentality to this that I can (and do) spend hours talking about.
 
But I think the more interesting question at this stage, and something we all will need to decide for our individual selves in time – is what are things that we can automate versus what are things that we want to automate?
 
In other words, where do we want to place our human touch in the future of our practice?
 
WHAT MATTERS, VERSUS WHAT DOESN’T
Some say that everything we do today, as designers, plan reviewers, engineers or whatever our role – that everything we do matters, and none of it should be automated.
 
In my opinion, that’s too rosy a view and gives us too much credit for our own time effectiveness.
 
Setting up job folders. Project kickoffs. Digesting specifications. Editing specifications. Setting up projects, or titleblocks, or detail sheets. Spellchecking. Exporting PDFs. Naming files. Combining files. Saving files. Sending files.
 
Did you know that in your line of project-based work – however you like to set up your job when you first start – you could probably have it 60% done just by literally saying a few words into your computer?
 
To me, that’s partially exciting, and mostly terrifying.
 
It keeps coming back to that very-specific question that’s been top of mind since early summer.
 
In our own future, what is it that we want to be doing?
 
DECISIONS
Thus far, I’ve seen a few AI startups that have taken the ball and run. Promises (likely true?) that you could have a full sprinkler layout in just a few minutes.
 
How does it work? AI does it.
 
I’m not going to make an argument here that design shouldn’t be automated because of a moral imperative that only a human should lay out sprinklers or orient branch lines.
 
We’ve seen versions of this argument throughout history whenever technology changes how jobs get done – from automation in banking to self-service gasoline to countless others.
 
Now we have autonomous cars.
 
So while you and I value our craft and seek to protect it, there’s not a natural law that says a line representing a sprinkler branch pipe or a circle that represents a sprinkler has to be physically positioned by a real person.
 
It doesn’t.
 
But how do we feel about deciding protection for a combustible overhang? Or a hazard classification for tote storage inside an assisted living facility? Or performance-based criteria for smoke control inside an amphitheater?
 
Those are pretty important decisions, and I think most of us would agree that we’ve missed the mark if that’s automated away.
 
LINE IN THE SAND
Just my take, but I think the line that should be drawn in the sand for what can be automated and what shouldn’t is by asking, ‘if our end-user fully understood the process, what would they actually want us spending human attention on? What decisions are important enough that they should originate from a person, not AI?’
 
If we were fully transparent about our process, what would our clients want to see?
 
Would they be thrilled, or horrified?

Would a client be annoyed if you had AI tools at your disposal, and you didn't use them?  
 
‘Well, Joe, my clients wouldn’t give a hoot if our design was good or bad as long as it didn’t cost them any time or money.’
 
I’m not talking about apathetic clients. It’s our job as experts to advocate for their safety, whether they care or not.
 
I’m talking about a knowledgeable, concerned end-user. What would they truly want to see us lay human eyes on? What would they be fine with us automating? What would they be upset that we weren't automating? 
 
I think that’s where the line should be drawn.
 
DESIGN EXAMPLE
If I’m an architect hiring a consultant, or I’m a general contractor hiring a sprinkler sub – do either of those parties care exactly how a line is drawn in AutoCAD?
 
Do they care how projects are set up and logged in the job folder?
 
Or, do they care about the quality, accuracy, and ability of the submittal to be code-compliant, effectively communicate what is supposed to happen, and have someone who is responsible and can attest to all those things for this specific project?
 
PLAN REVIEW EXAMPLE
If I’m a plan reviewer, I might not think of end-users as clients, but it’s similar. Most probably wouldn’t care if administrative pieces were largely automated.
 
Do responsible and cautious building owners care how projects are logged, tracked, fees assessed, and communicated?
 
Many would probably say that reviewing for correct occupancy, egress, required systems, applicable standards – at some point, a human needs to be directly involved in those things – even if they have tools supporting them.
 

Where do we need to have people actively making key decisions, versus just review output?

TOMORROW’S EXPERIENCE IS BEING BUILT TODAY
There are a thousand micro-decisions being made today that will impact how we work ten years from now.
 
Some of those decisions happen inside our organizations, some in traditional software providers, and some are happening in AI software startups.
 
My hope is that collectively, we intentionally think about and design a future where we place ourselves where we want to be.
 
In software circles, they say “human in the loop.” Is there a human in the loop in this process, or in this decision?
 
Personally, I think there are better ways to pull it off.
 
Maybe I’m being too technical but being “in the loop” sounds so ancillary to the process. I don’t want someone just casually watching things happen before them, I want key decisions made by that person. Not reviewed, but initiated by a person.
 
Can we speed up and support the things that we do really well?
 
Can human-driven decision-making compete at a speed that AI can, if done in the right environment and with the right systems?
 
Are we going to create AI output where the only human interaction is review? If so, that system is doomed from the start as one big miss, and there’s a whole cascade of downstream issues (not to mention liability).
 
Or, alternatively, can we design workflows where it’s not “someone should review this AI output” but rather deliberate points along the way where humans are the decision makers.
 
Hazard classifications?
 
A whole list of organizations say that’s an engineering responsibility.
 
If I’m designing the system, I don’t want the EOR ‘reviewing’ hazard classifications by AI or anything else. It’s a big-enough deal that I want the EOR initiating that decision.
 
Same goes for protected vs. omitted areas for sprinkler protection. Or using NFPA 13 versus 13R.
 
I don’t want AI taking a stab and allowing ourselves to ‘fall asleep at the wheel’ during a review – I’d want a deliberately designed system where the key decisions are made by an accountable person.
 
Force the decision to happen.
 
I think that’s the future we want for ourselves.
 
AI is going to change how our work gets done. That’s not really the question anymore.
 
The question is whether we intentionally design our new way of working where humans make decisions, or simply automate everything we can and leave ourselves responsible for reviewing the result.
 
I think we need to be deliberate about where we need to be involved now so that we arrive at a place of quality, transparent work and still be responsible stewards for the industry.