What role do you expect AI to play in pavement engineering? Are we entering an era of intelligent infrastructure, or just adding another tool to the toolbox?

Artificial intelligence is increasingly being applied to mix design, performance prediction, quality control, and asset management, promising faster analysis, improved accuracy, and data-driven decision-making. Yet pavement engineering remains deeply rooted in laboratory testing, field validation, and engineering judgment developed through experience. In your view, will AI fundamentally change how we design, build, and manage pavements? Or will it remain primarily a support tool that complements, rather than replaces, traditional engineering practice?

Anthony Brenes-Calderon, Ph.D. NCAT recently graduated Pavement Engineer

I expect AI to play a supportive role in asphalt pavement engineering. It has shown strong potential in improving decision-making for mix design, performance prediction, and asset management by rapidly analyzing large datasets from laboratory testing, field measurements, traffic, and climate.

However, pavement behavior is still governed by material properties, construction quality, and local conditions that require laboratory testing, field validation, and engineering judgment. AI models depend heavily on data quality and cannot fully replace mechanistic understanding.

As a result, I believe AI will fundamentally change how pavement engineers work rather than replace us. Engineers will act as interpreters and validators, integrating AI outputs with mechanistic-empirical design principles and field experience. Based on this, I believe AI will enable smarter and more efficient pavement engineering.

Mike Copeland Quality Program Manager (Materials and Construction), Idaho Transportation Department.

I believe GenAI is the next industrial revolution. Even if technology stopped advancing today, our industry has massive catching up to do just to utilize what is already here.

There is no manual for GenAI in pavement engineering, and AI labs won't write one. They built the engine but don't know our work. This is uncharted territory, like factory electrification a century ago. Power companies provided electricity, but industries had to figure out how to redesign the factory floor. We are in that same phase: we have the power, but must redesign the workflows ourselves.

In my own work, I’ve started stress-testing our systems. I demonstrated that today’s AI can already 'game' standard lab tests and automate data falsification, proving our traditional reliance on human trust is a major vulnerability. We have to adapt immediately.

But beyond risks, I am excited. We have decades of messy data and complex material behaviors. I think we have blind spots. These tools give us the capability to uncover areas where we’ve missed discoveries because our training conditions us to look for expected patterns based on what we think we know. AI doesn't have those blind spots.

Combining this with LLM-driven open-source robotics excites me most. We might finally remove subjectivity from the lab and make discoveries we’ve overlooked for years.

Remember: the AI tools we use today are likely the worst ones we will ever use. They will only get more powerful. So yes, we are entering a new era, and it is up to us to define it.

Matias M. Mendez-Larrain, Ph.D., P.E Director, Geotechnical and Pavement Services | Sr. Associate at WSB

Matias is the Director of Geotechnical Services and Pavements at WSB, leading geotechnical investigations and pavement evaluation for transportation and infrastructure projects. He oversees drilling, laboratory operations, and material testing, and has authored industry publications and technical papers on pavement performance and geomaterials. Matias serves as President of the ASCE Geo-Institute Oklahoma Section and is a member of the Transportation Research Board (TRB) Standing Committee on Geomaterials Behavior and Properties (AKH19).

When we asked Matias this month’s question about the role of artificial intelligence in pavement engineering, this was his perspective:

"From my perspective as a geotechnical and pavement engineer, AI will play an increasingly important role in pavement engineering. It helps us think broader and get to knowledge faster, especially in areas like mix design optimization, performance prediction, construction quality control, and data interpretation. AI can speed up design cycles and identify trends or risks that may not be obvious at first glance, making engineers more efficient and better informed.

Where I am more cautious is on the management side. Pavement management still relies heavily on engineering judgment, field experience, local constraints, and balancing cost, risk, and stakeholder priorities. That part goes beyond what data alone can capture. So while we are clearly moving toward more intelligent infrastructure, I see AI as a strong technical enabler rather than a replacement for how pavements are ultimately managed and decisions are made."

Noorahulda Saleh, Ph.D. Co-founder & Chief Engineer at Continuum Infrastructure Solutions.

Artificial intelligence will not replace pavement engineers, but it will fundamentally change how we work. Based on our recent case studies applying generative AI and large language models (LLMs) to real pavement engineering problems, I see AI evolving into a decision-support layer that complements traditional testing, modeling, and engineering judgment.

In our work, AI has proven most valuable in handling scale, complexity, and unstructured information. For example, we demonstrated that LLM agents can automate materials quality assurance tasks such as retrieving AASHTO R 80 aggregate reactivity ratings from state bulletins, even when inputs are incomplete or misspelled. More importantly, AI enabled us to synthesize insights from more than 10,000 abstracts and over 3,000 full-text publications, an effort that would be impractical to perform manually, while maintaining human oversight.

What made these applications effective was not the AI alone, but how it was engineered. Rather than relying on a single “black box” output, we designed structured workflows where literature was processed in batches and evaluated against explicit engineering rubrics. Outputs were required to meet minimum thresholds for factual accuracy, technical depth, neutrality, and completeness, with failed results automatically revised and re-evaluated. This approach allowed AI to scale engineering review while preserving traceability and quality control.

That experience reinforced that AI systems are inherently probabilistic and highly sensitive to how they are designed. Without safeguards, such as validation criteria and human-in-the-loop quality assurance, AI can produce inconsistent or misleading results.

So are we entering an era of intelligent infrastructure? Yes, but not through autonomous decision-making alone. The future lies in hybrid systems, where AI accelerates analysis and improves consistency, while engineers remain responsible for interpretation, validation, and accountability. In that sense, AI is not a replacement; it is a force multiplier for sound engineering practice.

MEET OUR STUDENT CHAPTER OFFICERS!

If you didn’t enjoy the email you can unsubscribe here.

To change your email or preferences manage your profile.

277 Technology Parkway, Auburn , AL 36830