Data serves as the fundamental building block of AI and the ultimate value that an AI system provides. Foundational success depends on robust data governance and management to ensure information is accurate, aligned, and secure. Data storage and access protocols are key to ensuring high-quality data is accessible at times and locations it is needed and available to staff based on identified roles. AI systems can offer more sophisticated insights if available data can be effectively and efficiently integrated across data types, collection methods and programs. The Intelligent Transportation Systems Joint Program Office (ITS JPO) is currently researching the state of data integration for AI & Transportation Systems Management and Operations (TSMO) applications. This effort will identify critical interoperability requirements to improve data readiness and test innovative, data-centric AI approaches for prioritized TSMO use cases.

Effective AI implementation in transportation is less about the "black box" of technology and more about the foundation of trust public agencies build around it. AI Risk Management acts as an all-encompassing umbrella, providing a continuous strategy to identify and mitigate potential hazards before they impact multimodal transportation systems.  This risk “umbrella” is held up by two critical pillars: AI Standards, which provide the rigorous, nationally recognized benchmarks (like the National Institute of Standards and Technology (NIST) AI Risk Management Framework) to ensure AI systems are defensible and consistent; and AI and TEVV (Test, Evaluation, Verification, and Validation), the structured process of proving a system is safe and reliable in the real world. Together, this framework ensures that whether the focus is on managing simple assets or complex traffic flows, AI remains a transparent, safe and trustworthy solution. The ITS JPO has a suite of forthcoming AI risk, standards and TEVV resources under development to support transportation agencies as they implement AI solutions.  These will be tailored by staff roles, program areas and agency type.

AI-driven solutions—ranging from advanced signal control to predictive maintenance—promise data-driven decision-making that can address real-time transportation challenges. However, the complexity and criticality of these systems demand rigorous methods to perform TEVV over AI components and their integration into legacy systems across their lifecycle, ensuring they are both reliable and aligned with stakeholder needs.

To tackle this complexity, USDOT is pursuing a structured approach to a comprehensive AI TEVV Framework. This effort is intended to guide state and local agencies—as well as their consultants and vendors—in systematically designing, implementing, and monitoring AI solutions using common definitions, artifacts, and acceptance criteria.

In an era of rapid technological shifts, transportation agencies face the daunting task of moving from curiosity to being "road-ready". A Capability Maturity Model (CMM) can serve as a guiding “compass” for this journey, providing a structured approach to assess an agency's readiness for AI implementation at different scales. For transportation agencies of different types and sizes, an AI CMM offers a customizable, structured approach to self-assessment that bridges the gap between different program areas or roles. Agency leadership and staff can leverage the results of a CMM assessment to collaboratively identify key focus areas for advancement, establish high-priority AI goals, and establish specific priorities that align available strategies and resources to move purposefully between levels of AI maturity. The ITS JPO is updating its 2022 AI CMM framework and will provide new AI readiness resources. These will include an enhanced AI CMM framework and online tool (with new institutional and technical assessment dimensions) along with curated reports, fact sheets, quick start guides, and supporting workshops and webinars.

As transportation agencies increasingly procure AI systems, shifting from a "buy and deploy" mindset to a lifecycle approach is essential for long-term success. Fundamentally, this means treating AI technology as a living asset that requires active management rather than a one-time purchase. Procurement begins with a rigorous planning phase where agencies define the problem before selecting the tool. At this stage, agencies should evaluate if their existing data is high-quality and ready to "feed" an AI and determine if their current IT infrastructure can actually support it. It is also critical to establish clear performance metrics and accountability protocols. During selection and award, the focus shifts to verification. Key considerations should center on how vendors will demonstrate transparency and prove their systems meet safety parameters through rigorous testing. Post-award, agencies must plan for continuous monitoring, as AI systems can "drift" or lose accuracy when they encounter new real-world conditions. By managing AI through every procurement stage, transportation agencies ensure their investments remain a safe, dependable, and trustworthy part of their decision-making infrastructure. The ITS JPO will deploy a national webinar and peer exchange series on AI procurement in late summer/early fall 2026.