If you’re a building owner, the AI debate probably isn’t playing out inside your own organization. You’re not deciding whether to build a language model or buy a SaaS subscription.

But your design firms and contractors are making exactly that decision right now. And which way they go, and how thoughtfully, will affect the quality of work you receive.

Understanding the tradeoffs helps you ask the right questions.

The “not ready yet” misconception, revisited

Two years ago, building custom AI tools required engineering resources and months of development time. Today, a skilled practitioner with no coding background can build and deploy a custom AI workflow in an afternoon.

That shift has reframed the build-versus-buy decision entirely. Speed is no longer the decisive argument for buying. The real questions are different: What do you actually need the AI to do? How much proprietary data does it need to work with? How many people need to use it, and at what cost?

When buying makes sense

For most organizations at the beginning of their AI journey, buying is the right call, and not just for cost reasons.

Off-the-shelf AI tools get you to testing quickly. You find out whether your team will actually use it, what problems it solves well, and what it doesn’t touch. That information is valuable before you commit resources to building.

Think of early adoption as intelligence gathering. If only a small fraction of your team adopts the tool, you’ve learned something important: either the tool isn’t the right fit, or the behavior change required to use it is bigger than anticipated. Either way, you’ve lost less than you would have if you’d built a custom system first.

The platforms worth trying, general AI tools adapted for AEC workflows, are accessible and fast to onboard. They let you develop the organizational literacy to have a more sophisticated conversation about what you actually need.

When building makes sense

Scale changes the math. For a firm with hundreds of employees, per-seat software costs compound quickly. At that point, building a custom solution often becomes more economical, and more strategically valuable.

Custom-built tools offer something else that matters for AEC firms: control over your own data. Your project archives, your coordination history, your RFI patterns, your cost data. These are proprietary assets. An off-the-shelf tool may use that data to improve its own models in ways you don’t fully control. A tool you build keeps your data yours.

The practical barrier to building has dropped significantly. Firms with developers already on staff, hired originally for BIM plugins or automation tools, can redirect that talent toward AI applications. The learning curve is real but manageable.

The third path most firms miss

The most productive answer for many AEC firms is neither pure build nor pure buy: it’s partner.

Find a technology company whose capabilities complement yours, and build something together. The tech partner brings AI development expertise. The AEC firm brings domain knowledge and real project data. The resulting tool is more useful than a generic off-the-shelf product and more achievable than a pure custom build.

This approach also distributes the risk. If the tool doesn’t work the way you hoped, the investment was shared. If it does work, you’ve built a capability that competitors can’t easily replicate.

What this means for how you evaluate your partners

When your design or construction team says they’re “using AI,” the meaningful follow-up is: built, bought, or partnered?

Each answer carries different implications. A team that bought a general AI subscription may be using it productively or may have abandoned it after the trial period. A team that built custom tools invested in something specific, ask them what it does and why they built it rather than bought it. A team with a technology partnership is betting on a capability that’s evolving, ask who the partner is and what they’re building together.

The answers are more revealing than any marketing claim about being “AI-forward.” The question you’re ultimately asking is: have they thought carefully about what they actually need, or are they just following a trend?

The firms that can answer that question specifically are the ones doing real work. The others are still deciding which tools to demo.