The tools are already in the design and construction industry. What comes next?

Artificial intelligence has moved from an experimental curiosity to a working part of the architecture, engineering, and construction (AEC) toolkit, reshaping the profession at every stage of a project’s life. Generative design systems can now explore thousands of structural layouts, facade configurations, and spatial arrangements against defined performance constraints in the time it once took a design team to sketch a handful of options by hand.

Optimization techniques are applied to structural sizing, material selection, energy performance, and construction sequencing to reduce cost, time, and waste before a shovel breaks ground. Machine learning models support load prediction, cost estimation, schedule forecasting, and predictive maintenance, tasks that used to depend entirely on an experienced engineer’s judgment.

A large-scale, peer-reviewed survey of 501 architecture, engineering, and construction professionals offers a picture of how AI is actually being used in practice, as opposed to how it is discussed in the trade press. The study found that AI adoption is widespread but uneven: generative AI assistants such as ChatGPT, Claude, and Copilot are the most common entry points, while specialized applications remain comparatively rare. Most firms use AI in their marketing and internal management processes; few have fully embraced it for the actual design.  AI familiarity and use remain uneven across the profession, raising questions about how firms develop competency across different levels of experience. The fast pace of adoption is why these questions cannot wait: the tools are already inside the design process, often ahead of any firm-wide policy governing how their outputs are checked, documented, or insured.

This is not just a story about the drafting table. Tools within platforms such as Autodesk explore design alternatives. OpenSpace uses site imagery to track construction progress and is developing AI “agents” and application programming interfaces (APIs) that synthesize outputs across software systems. Procore supports risk forecasting and document analysis, while firms such as Boston Dynamics have developed autonomous equipment and inspection robots for active job sites. The global AI construction market was valued at approximately $4.86 billion in 2025 and is projected to reach $35.53 billion by 2034. The professional-responsibility for design and construction follows these tools across the project’s lifecycle.

AI as a creative partner, not a substitute for professional responsibility

Autodesk’s description of its own AI design tool offers a useful starting point for thinking about the appropriate role of AI in professional practice:

“enabl[ing] designers to identify what works, detect patterns, and select the best-performing option. Think of AI as a creative partner that manages the heavy lifting, allowing designers to focus on vision and decision-making.”

This description distinguishes AI’s role in generating and evaluating options from the professional’s responsibility for vision, verification, and decision-making, emphasizing that the professional ultimately holds the reins.

Although AI may improve efficiency through speed, accuracy, cost savings, and scalability, its use does not shift professional responsibility. The licensed architect or engineer of record remains responsible for damages arising from a defective design, regardless of the extent to which AI contributed to the drawings, specifications, or calculations.

When a registered engineer, architect, or other registered professional places their state seal on a final drawing, survey, report, or other construction document, they assume liability by certifying that the design is made to “safeguard life, health, property, and promote the welfare of the public.” State requirements reinforce this accountability through rules governing competence, delegation, sealing, and responsible charge, for example:

For the licensed engineer or architect signing and sealing the work, an AI-generated output is an input to professional judgment, not the end of the process. Reviewing, verifying, and standing behind the finished product remain part of the job, just as if it had been prepared without AI. AI development companies targeting the AEC industry recognize this allocation of responsibility and rely on it to limit their own liability.

Reasonable professional care, not a guarantee of perfection

Knowing who remains responsible is one question; understanding the standard against which the work is judged is another. For engineers and architects, using AI or another new technology impacts  the professional standard of care. As with all work performed by architects and engineers, liability for negligence is based on whether the work complies with what a “reasonable” architect or engineer would have done under the same or similar circumstances. How engineers and architects use AI, what their review processes are, and the policies they adopt all will shape the standard of professional care with respect to AI use in the industry. Keeping apprised of how the industry is using AI will be critical to understanding how the standard is being shaped.

It is important to note that this standard does not mean perfection. The benchmark is the skill and knowledge normally possessed by members of the profession in good standing, unless the professional claims greater or lesser skill. Engineers and architects are not held to an unduly burdensome “utmost care” standard that would improperly turn them into insurers of the result. Nevertheless, whenever a failure occurs on a construction project, the engineer’s or architect’s compliance with the standard of care will almost always come into question.

The stakes of getting it wrong are illustrated by real design failures. For example, federal investigators concluded that the collapse of the I-35W bridge in Minneapolis, which killed 13 people and injured more than 100, resulted from inadequate load capacity caused by a design error in the bridge’s gusset plate calculations, compounded by inadequate quality-control review of those calculations. This was not an example of AI-assisted design, but it illustrates why the calculations, quality-control procedures, and professional judgment behind a design matter, not merely its final output. For AI-assisted work, the lesson is not that every error establishes negligence or that professionals must guarantee a flawless result. It is that the speed and apparent sophistication of an output cannot substitute for professional review and verification.

The consistent nationwide standard is captured in the seminal formulation from Coombs v. Beede (1896), still cited across jurisdictions: a design professional’s undertaking implies the skill to perform services ordinarily and reasonably well, but “the undertaking does not imply or warrant a satisfactory result,” and an error of judgment is not necessarily evidence of a want of skill or care.

For AI-assisted work, then, the question is not simply whether the tool produced the right answer, but whether the professional exercised reasonable care in evaluating and using it consistent with how other professionals would. In future updates, we will  explore this standard of care in practice and address loss prevention tools such as contracting, communication, insurance and verification/quality assurance and quality control.

This article does not constitute legal advice. For legal advice that applies to your situation, you should contact an attorney.

Sign up

Ideas & Insights