AI Agents for Construction and Engineering Firms: What Autodesk's 2026 Survey of 2,500 Leaders Shows

Nearly every architecture, engineering, construction, and operations (AECO) leader now uses some form of AI at work. That part of the story is no longer interesting. What is interesting is what happens next: a fast, uneven scramble toward agentic AI — systems that don't just answer a prompt but coordinate tasks, move work across tools, and carry a project forward with less hand-holding.

That scramble is documented in unusual detail in Autodesk's 2026 State of Design & Make: AI Pulse report, compiled with Statista Plus Research from a survey of 2,500 global industry leaders across AECO, design and manufacturing, and media and entertainment, fielded in January and February 2026. It's one of the largest, most current looks at how the built-world industries are actually deploying AI agents — not how vendors say they should.

AI tools are table stakes. Agents are the new frontier

The baseline numbers are almost boring at this point: 98% of Design and Make leaders say they personally use at least one AI tool, and only 2% report using none at all. Adoption has moved from differentiator to default so fast that Brad Sara, digital practice lead at architecture and design firm Warren & Mahoney, put it bluntly in the report: "AI is already becoming the new expectation — if you're on the front of the wave you get an advantage, but very quickly everyone catches up and it just becomes part of how work gets done."

The real signal is one layer up, in agentic AI specifically. Per the survey:

• 59% of organizations already are, or will be within a year, using agentic AI.
• 43% of organizations plan to be using agentic AI within a year — a 26% increase year over year.
• 65% of "AI early adopters" (organizations that say they've genuinely integrated advanced systems like agentic AI or LLMs) are already using agentic AI, versus far fewer of everyone else.
• 44% of all leaders are actively investing in AI agents specifically, and that jumps to 57% among early adopters versus 41% among the rest of the field.

Mike Haley, SVP of Research at Autodesk, frames the confusion around what "agentic" even means in the report as three distinct tiers most firms are moving through: lightweight assistants that execute a sequence of steps on request; workflow-level agents that coordinate across tools and data sets with less human intervention; and — still mostly aspirational — fully autonomous systems that understand intent and adapt on their own. Most AECO firms today are somewhere in tier one, edging into tier two.

Where firms are actually putting agents to work

The report backs the adoption numbers with specific, named examples rather than abstractions. Andrew Stanford, director of digital technology at consulting engineering firm Haskoning, told Autodesk: "We've trained over 4,000 people on AI agents and now have over 200 agents actively in use. They're supporting everything from tender reviews to legal contract analysis, which shows how embedded AI is becoming in daily operations."

That's not a pilot program — that's 200 agents running inside the operations of a single engineering firm. Cillian Kelly, digital project delivery lead at construction and engineering firm John Sisk & Son, points to the coordination problem specifically: "AI has a huge opportunity to improve coordination across projects, particularly in complex environments where multiple teams need to align. And that creates real potential for better outcomes." Yusuke Okura, BIM manager at spatial design firm Semba Corp, adds a more everyday framing: "Leveraging AI has become essential for us, particularly in streamlining routine documentation and information gathering, which allows our teams to focus more time on higher-value creative and design work."

The productivity numbers back up why firms keep pushing further: 84% of leaders say AI has had a positive impact on productivity in their organization, 77% say it's increased innovation, and 73% say it's improved the quality of output. Decision-making saw the single largest year-over-year jump — up 10 points, with 65% of leaders now saying AI helps them make better decisions, and a fifth of those calling the impact "significant."

The gap: adoption is outrunning the data foundation

None of this comes free. The same 2,500 leaders who are racing toward agentic AI also flagged, in the same survey, exactly why most of them aren't there yet:

• 50% say integrating AI with existing systems is a top implementation challenge — the single most-cited barrier.
• 47% cite a lack of the talent or skills needed to implement AI properly.
• 43% point to data quality issues specifically.
• 60% are concerned about security when it comes to AI.

Swapnil Shrivastav, CEO of air-to-water generating systems company Uravu Labs, summed up the underlying problem in the report: "We need much more real-world data before building full AI systems. Without enough data the models will not be reliable enough for the kind of decisions we need to make." That's the same theme Autodesk's own analysis lands on: agentic AI's potential is "constrained by the same factors that affect AI more broadly: quality of data, structure of workflows, and the extent to which systems are integrated." An agent that's supposed to coordinate a tender review or a legal contract analysis across systems is only as good as the data and workflow structure it's plugged into — and half of the industry says that plumbing isn't there yet.

What this means for construction and engineering firms going forward

The firms already running 200 agents, like Haskoning, didn't get there by buying an agent and hoping — they got there by training thousands of employees and building the workflow structure agents need to actually coordinate work, not just respond to prompts. That's the gap between the 59% of organizations "using or about to use" agentic AI and the much smaller number actually seeing the outsized gains early adopters report (a full 18-percentage-point advantage in decision-making alone, per the survey).

For AECO leaders evaluating where to start, the report's own data points to a clear order of operations: fix the integration and data-quality problems first — the two most-cited blockers — before layering coordination-level agents on top of them. An agent that can move a tender review from one team's system to another's is a workflow problem before it's an AI problem. Firms that get that sequencing right are the ones positioned to close the gap Autodesk's early adopters have already opened up.