The sixth AI at the AEC 2026 conference showcased the evolution of AI discussions. There were, naturally, many talks about software and technologies. But more than before, there were conversations about realizing AI’s business value.

Two themes appeared in nearly every session I attended. First, many companies struggle with AI adoption, not because they lack tools, but because their thinking isn’t right. Second, when AI works, it disrupts the business model that brought them there.

The adoption problem is not technical

70 to 80 percent of AI initiatives fail due to adoption challenges, not technology. The tools and data are increasingly available. The missing piece is an organizational mindset.

Several sessions highlighted the same root cause that Dr. Sam Zolfagharian discussed in her keynote. Companies are burning through AI tokens and running pilots, yet the ROI never shows up. That’s because they are deploying AI to systems and processes built around human limitations, rather than redesigning those processes to take advantage of what AI can really do.

One presenter from an engineering firm shared a powerful example. When they tried to implement an AI agent for piping design, they found that their 30 designers were following six different processes. The most valuable result of that project was not the AI itself; it was the final mapping and standardization of those workflows. The lesson: do not automate chaos. Standardize first, then automate.

Similarly, adoption rates increase significantly when AI is integrated directly into the tools engineers already use, instead of forcing them to switch to a different interface. Working within Plant 3D, Revit, or whatever the team already uses reduces friction and shows that AI is there to support the work, not to replace the workflow.

Culture beats governance

Several speakers argued that governance frameworks alone cannot guide responsible AI use in AEC, especially regarding data decisions and design choices. Culture must do the heavy lifting.

This means teaching people what AI actually is, not just how to use it. One company shared that they had stopped running prompt engineering workshops altogether. Instead, they focus on teaching the fundamentals of large language models and machine learning. When people understand the mechanics, they become more comfortable experimenting on their own and are better at selecting the right tool for the right problem.

How leadership frames the transition also matters. Instead of implementing AI out of fear of falling behind, the most effective leaders linked AI adoption to professional pride. The question is not “how do we deploy this tool?” but “how do we help each person maintain the craft they value?” That reframing turns resistance into genuine engagement.

One striking approach: during a firmwide Copilot rollout, employees were told they were now supervisors. The AI was their new colleague. What it produced, they were responsible for. That shift from task-doer to steward fundamentally changes the relationship with the technology.

Finally, organizations undergoing AI transformation must protect their ability to learn. Expecting employees to gain new skills on top of a full workload will lead to burnout, not progress. Allocating 15% of time for learning and innovation, as some companies have done for years, seems like a smart strategy.

When AI works, the business model breaks

Joaquin Arocena, EIT, MS, PMP, LEED AP of Proyectos Engineering reported last year as the first in which they generated more revenue with fewer billable hours. Nobody was let go; they just accomplished more in less time. This sounds like success. But it exposes a fundamental problem: in an industry built on selling hours, efficiency destroys the model.

Petra Svensson Gleisner and Ivana Kildsgaard put it starkly in their presentation. When machines produce value, architects still charge for human work by the hour. But machines cost money, hardware, software, API tokens, and the competence to run them. If those costs cannot be recovered through billing, the traditional model collapses.

AI forces us to rethink roles and business models. Should a consultant just do more of what they’ve traditionally done, or should they extend their scope to new areas with AI?

From super consultants to data hamsters

Petra and Ivana proposed four viable paths forward. The first, which most of the industry currently adopts, is the super consultant: leveraging AI to deliver better results, more quickly, with higher profit margins. It’s a strong position in the short term, but it offers limited differentiation as AI becomes widespread.

The second approach is to expand the value chain: the vertical integrator enters adjacent stages of the project lifecycle that were previously inaccessible. AI enables a design firm to confidently branch into procurement analysis, operations planning, or construction logistics.

The third is the tech shark: creating proprietary AI tools that become products on their own. Few companies have the appetite or ability for this, but those that do can generate completely new revenue streams.

The fourth, and perhaps the most overlooked, is the data hamster. Firms that have been gathering project data for decades are sitting on an asset most of them haven’t yet monetized. Proprietary data, used to develop custom AI models and insight-driven services, is where lasting competitive advantage resides. “Build on your own data,” or, as Guido Maciocci put it, “Own the intelligence.”

Aim for radical improvements

The “10x” factor was mentioned multiple times at the event, emphasizing that we must and can pursue radical improvements, not just small steps with AI. Bart Brink set a clear goal: our industry needs to increase its productivity by a factor of three by 2050 to meet customer demands.

One presenter described procurement reform as the key that could speed up everything. When buyers move from focusing on hours to focusing on outcomes, suppliers can invest in AI-driven efficiency, enabling transformation throughout the entire supply chain.

The takeaways

The attendees left Helsinki with much to consider and valuable ideas for future paths. They realize that AI shouldn’t be limited to IT or a single champion but should be owned by the C-suite. Their companies need to prioritize building culture and fundamentals, not just tools. As a result, they are beginning to ask tough questions about what they are really offering.

PS. It was a pleasure to be the content partner of AI in AEC for the second time. I look forward to the 2027 conference. Hopefully, by then, the ROI of AI will no longer be just theoretical but a tangible reality for most companies.