I thought AI would make my job less technical. Instead, I became a builder.

When AI tools showed up, I had clear expectations. I spent much of my time building complex financial models. I assumed AI would take over that manual work, and I'd shift toward facilitation and the people side.

AI did unsurprisingly take on much of the manual build, but it also pulled me into more technical work, not less. This year my team has built agents, redefined workflows, and widened the scope of our work thanks to these new AI tools. The job description defining my role today looks very different from when I first joined the company almost 3 years ago. The tools made it possible, and the results were too good to ignore.

I'll admit, the start wasn't graceful. With new tools every week, no clear place to begin, and a quiet question I didn't say out loud: am I technical enough for the job this is becoming?

More technical and more capable

Tasks that used to sit outside my reach are now part of how I design. I can prototype a tool, test the logic, and put it in front of our clients while the idea is still fresh.

What a builder does differently

By builder, I mean someone who designs the simulation and also builds the tool it runs in, from backend to frontend. It fits our company priority, Run Simple. I no longer hand the front end to someone else and wait, but can take an idea from design to a working tool ony own and get it in front of clients in a fraction of the time.

Three habits changed first:

• Design and build in one loop - I used to mock up what I wanted and wait for someone else to build it. Now I change the logic and see the screen update in the same sitting.
• Test like an owner - When I build both the model and the screen, there's less risk of discrepancies between the front end and the back end, and the client gets one cohesive experience from start to finish.
• Change things live with the client - When a client asks to rework a decision, I can make the change and show them before the meeting is over.

That's the shift behind our AI-first simulation approach, seen from the person doing the building.

What didn't change

The hard part of a simulation was never the tool. It's choosing the decisions that matter, setting the tradeoffs so no option is obviously right, and running a room where people face the consequences of their choices. AI can draft options, but it can't tell which tension matters most to this client's business, or sense when a room needs pushing. That judgement still comes from us.

The tool matters in one way: a simulation is only as good as its logic, and a polished tool with wrong logic is worse than no tool, because it teaches the wrong lesson with confidence, Judging it well takes years of practice, so whoever builds with AI needs to understand the simulation model and the client business well enough to check that it's right.

The faster the build gets, the more that judgement is the job. It's what separates a simulation that looks right from one that changes how leaders make tradeoffs once they're back on the job.

What clients see now

This year, I put a working mockup of a simulation in front of a client much earlier than I previously could. A mockup used to take around a week to create. Now it takes hours. Seeing something real, clients can say what works and what doesn't, and they give us more feedback, and better feedback, than a description ever would have.

Where my work is heading

Back to the question I didn't say out loud: am I technical enough? I've stopped asking it that way. These tools are easy to start with. They are much harder to master. Every layer I added, from understanding how a tool works under the hood to building skills and agents, demanded a deeper grasp of the simulation logic and of how the AI tools themselves are built. I became technical enough one project at a time, and the bar keeps moving. That is now part of the job.

Looking ahead, there are three things I want to do more of. First, bring clients into the design earlier and more often, now that a working mockup is quick to build, so they shape the simulation while it is still easy to change. Second, get sharper at checking the logic, because the faster we build, the easier it is for a wrong model to hide behind a polished screen. Third, use agents and redesigned workflows to take more of the routine work around a simulation off our plate. I don't have a map beyond that. Nobody does yet.

I expected AI to move me toward facilitation and the people side of the work, and I was half right. Deciding which decisions matter and sensing when a room needs pushing is still the job. What changed is that I now build much more of the solution I facilitate.

If you design simulations and wonder whether you're technical enough, pick one small project, build the tool yourself, and put it in front of a client early. You'll find out quickly what you need to learn next.