What Small Firms Taught Us About AI Adoption in AEC

Notes from the AIA Los Angeles "Intelligent Practice" panel

Last week, we joined the AIA Los Angeles panel Intelligent Practice: AI & Technology for Small Firms, a room full of principals, practitioners, and technologists asking one question in different ways: how should small and emerging firms actually adopt AI?

The AI conversation in AEC tends to split into two camps. Enterprise firms bring budgets, IT teams, and structured pilots. Small firms bring speed and agility. Both perspectives are valuable, and the firms moving fastest in either camp share one thing in common: they rethink their day-to-day workflow and power it with AI.

Here's what we took away from the conversation.

Four takeaways

1. Tooling breaks first, but workflow is the real issue. The mistake isn't picking the wrong tool, it's chasing the hype and stacking tools that never get leveraged in day-to-day work. Start from the workflow that eats your hours, then find the tool built for it. That's where the ROI lives: manual drafting, code research, design-to-construction coordination.

2. Adoption has to be led from the top. Junior staff can and should be part of testing, that's not the point. The point is that leadership has to drive the conversation and lock the tool into how the organization actually works. Most resistance comes from fear of being replaced. Good leaders help the team see the true trend: AI isn't here to replace your team, it's here to free them from the manual, tedious work so they can focus on the high-leverage parts of the job.

3. Small firms have a structural edge worth using. Enterprise AI investment will shape the industry, but small firms have their own edge: fewer layers, shorter decision cycles, no committee to convene. A eight-person studio can decide on Monday, test on Tuesday, and run production on Wednesday. That kind of agility is hard to replicate at scale.

4. The tools that work solve specific, real problems. Most AI in AEC still feels imported from SaaS: generic tools in hard hats. The ones that earn their place solve narrow problems: "generate permit-ready electrical plans," "automate code research," "turn a sketch into a buildable drawing set." That's where the ROI shows up in the first week, and where the work itself starts to feel different: less time on the parts of the job nobody enjoys, more time on the parts that drew you into the industry in the first place.

Six questions from the panel

Q1. What AI tools are having the biggest impact right now? Two types, both needed. 1) General AI (Claude, Gemini, ChatGPT, Grok) handles the everyday work: proposals, RFIs, meeting notes. 2) Vertical AI is purpose-built for a specific industry, and designed to take on the domain-specific work: building codes, drawing standards, construction documentation.

Q2. Is AI adoption realistic for a small firm, and how do you measure ROI? Yes. Teams under 5 start with low-cost general tools. Teams of 5+ move into heavier, domain-specific tools that change production capacity. The simplest ROI metric: the people-to-projects ratio. If AI lets a 5-person firm carry the workload of 8, that's real ROI here.

Q3. What happens to firms that don't adopt? Is there a real competitive gap? Yes, and it compounds. Deloitte's 2026 AI report shows two-thirds of organizations using AI report productivity gains. On the AEC side, Bluebeam's AEC Technology Outlook 2026 found and the 27% of AEC firms already using AI plan to increase usage. Early movers are pulling ahead, and clients notice when they can get the same job done 5x faster somewhere else.

Q4. What single leadership decision most accelerates AI adoption? Leadership committing to drive the change themselves, not handing it off to a curious junior. Once a tool gets used on a real project, the question flips from "should we?" to "why are we still doing it the old way?" That shift has to come from the top.

Q5. What major changes should we expect from AI tools in the next 6 to 12 months? Three shifts worth watching.

1. Agentic workflows: tools that take multi-step tasks instead of hand-holding each step.

2. Multimodal: sketches, photos, 3D scan, voice, and design intent in one conversation. Far less friction between idea and output.

3. Ecosystem integration: tools that connect into the larger ecosystem, like Revit, AutoCAD, and Bluebeam, instead of forcing teams into separate environments.

Q6. One ethical red line we won't cross with AI? Removing the licensed professional from the review loop. AI generates, humans validate. The stamp means something, and it always will.

Where we go from here

A few people in the room asked about the specific tool stack we walked through: what's worth testing, and where to start. That's a longer post, coming next one.

In the meantime, if you're a principal, developer, or project lead thinking about where to start with AI, our honest advice is this: don't start with the tool. Start with the workflow. Find the three tasks in your projects that eat the most hours, stall the progress, and deliver the least joy. Those are your candidates. Everything else is noise.

Thanks to AIA Los Angeles for convening the conversation, and to the fellow panelists and attendees who made it a substantive one.