From Restriction to Unlimited Potential

2025 Year End Review

As 2025 ends, it’s tempting to measure the “year of AI” by what visibly changed in our daily work. And honestly, a lot of the day to day reality for architects and engineers still looks familiar. We still coordinate, we still search for information, we still struggle with handovers, we still deal with messy data and fragmented responsibilities.

But the most important change this year was quieter. It happened one layer deeper.

AI changed how we create software

The real shift in 2025 wasn’t that AI replaced architects or engineers. It was that AI changed who can build tools.

In the past, creating a script or automation was limited by two bottlenecks: access to knowledge and access to developer time. Even if you knew exactly what you needed, “making it real” often meant weeks of delay, budget discussions, or dependence on specialists. Many of the best workflow improvements never happened, not because they were impossible, but because they were too costly to build.

This year, that constraint loosened. AI made it dramatically cheaper and easier to create personalized software, small automations, and custom workflows. The barrier to turning an idea into a working tool dropped. That changes the game for AEC, because it means the people closest to the problem can increasingly shape the solution.

Over time, this will shift our industry from labor intensive execution toward knowledge and insight driven work. Not because effort disappears, but because the question changes. Instead of asking “can this be done,” we can spend more time asking “what is worth doing?”

Unlimited supply is coming, so we need to choose

When the cost of producing outputs collapses, something else becomes scarce: direction.

We are moving into a world where there are endless possibilities available to us. Endless design options. Endless simulations. Endless reports. Endless automations. That can feel empowering, but it also forces a deeper responsibility.

Which path should we choose? What does a good built environment actually mean? Is it low carbon, low cost, resilient, adaptable, beautiful, inclusive, all of it, and at what tradeoffs?

And then the uncomfortable follow ups: How do we make money from building “good” in a way that is not extractive? How do we make sure the value created by this new productivity does not only benefit a small group, but improves outcomes for everyone involved?

If AI increases our ability to generate, our job becomes choosing what deserves to exist.

Computing is no longer a geek’s game

Another shift is cultural. Computing is moving from “specialist skill” to “basic literacy.”

We are getting closer to a world where using data is not optional. And data can be counterintuitive. It can break the comfortable cultural ideas we carry. It can reveal inefficiencies, biases, waste, and tradeoffs we would rather not face.

But acknowledging that we don’t know is the beginning of progress. That humility is how every scientific field moves forward.

In AEC, this matters because the built environment is ultimately governed by physical laws: energy, materials, climate, time. Those laws don’t negotiate with our assumptions. We can’t hide from them anymore, not if we want the built environment to remain livable and sustainable.

Maybe this is the moment we grow up. We live on this planet and call it home, but “home” does not mean safety by default. Seeing reality clearly can feel lonely. Yet at the same time, there may be nothing more freeing than clarity. When we face the truth, the old limitations begin to dissolve.

A question for the year ahead

If the reason we meet each other is no longer to make other people do something for us, what would we say to each other instead?

Maybe we will meet less for delegation and more for alignment. Less for execution and more for intention. Less for control and more for shared responsibility. AI might not only change what we build, but how we relate while building it.

From restriction to unlimited potential.
Happy New Year. See you on the other side.

Shicong, DataDrivenAEC