Dr Sinan Yordem of Cambridge Consultants shares practical steps for successfully deploying AI in transportation 

Could you introduce yourself and share how you came to be in your current role? 

Absolutely. I’m Sinan Yordem, Head of Energy and Infrastructure Business in North America at Cambridge Consultants, where I help our clients in their transformational technology and innovation initiatives.  

My journey to this role has been shaped by over two decades of experience in technology and product development in electronics, optics, energy and infrastructure – sectors where I led start-up and corporate-wide priority growth initiatives delivering physical products and digital services to sector-leading government and industry clients. 

Today, I have the privilege of spearheading cutting-edge projects collaborating closely with start-ups, multinational corporations, investors and government agencies on technology-driven investment and implementation programs to drive operational efficiencies and strategic growth. 

What are the most pressing priorities transportation leaders must address today to effectively integrate AI into their operations? 

Leaders in transportation are aware that they can’t be left behind and are intent on investments in AI. However, the biggest challenge is showing business impact with AI and managing the scale of change it brings.  

To move beyond experimentation, organizations must align the use of AI tools with strategic priorities, define success metrics, and foster collaboration across departments. It’s not just about technology—it’s about mindset. Embracing AI requires cultural change, cross-functional teamwork, bridging the skills gap, and a strong understanding of where AI fits best. 

Developing trustable AI is crucial. Leaders need to have the appropriate assurance and governance to develop effective, unbiased, explainable and safe tools. Alongside this, they need to be thinking how to design and integrate AI tools into the existing systems and workflows to encourage trust and adoption by users so the benefits can be realized, and AI can augment and enrich human capability.  

There’s enormous potential for AI tools to deliver value, improving outcomes in safety, resilience, and effective and efficient operation of transportation networks, but the cyber and physical security elements must be considered carefully and with an eye to the future development of capability when deploying. 

Rethinking procurement and data strategy is another key priority for transportation leaders. Traditional procurement processes can’t keep pace with rapidly evolving AI tools. Forward‑looking organizations are experimenting with new, agile approaches to procurement and data sharing—supported by clear governance, privacy protections, and data ownership.  

Underpinning these efforts is a mature data strategy: maturity in managing the emerging datasets required to train AI systems – the communication systems to collect the data, arranging equitable access or shared data systems as explored in our previous report – How digital infrastructure will shape the future of transportation systems, privacy enhancing technologies, provenance tracking, assigning ownership and liability – this all helps enable successful assurance and reduced risk when scaling a deployment.  

How can AI be deployed in ways that uphold the highest standards of safety, while also maintaining public trust through transparency and accountability? 

Deploying AI in a safe and trusted way is about careful design of the system; utilizing assurance frameworks incorporating trust, explainability, safety and security, and impact to govern and guide the development of the algorithms, training sets and acceptance testing for the AI models.  

Key deployment considerations: 

• Scalable oversight – ensuring that the human has a chance to monitor, intervene or act if necessary. Humans will always be users, creators and authorizers. 
• Understanding where AI is a good fit, and how to blend its development and deployment with more traditional engineering practices (such as conventional V&V methods) and how to integrate with current processes smoothly to ensure ongoing compliance with evolving standards and regulations.
• Understanding what the AI is, how it is being integrated – the trusted aspect should also be evidenced with the appropriate tests and tooling that are AI specific (e.g. tests for data poisoning, unwanted bias etc.)
• Monitoring systems – particularly for adaptive AI or where real-world data will gradually shift away from the training set.
• Identifying mismatches between the level of trust in a system and its trustworthiness (whether it merits that trust) and addressing them.
• Using structured assurance trees that cover sources of risk, hazards, undesirable conditions, and mitigation strategies, using an evidence-based approach to make safety arguments within a clearly specified operational design domain. 
• Getting buy-in throughout the organization on AI and ensuring there is a diversity of thought when it comes to risks and concerns.

What does a “robust infrastructure” look like when it comes to enabling AI in transportation, and what foundational elements are often underestimated or overlooked? 

Underpinning a robust infrastructure will consist of effective digital infrastructure where attributes like a common data layer will allow information to flow and applications to be enabled. Some of these will utilize AI and in the most safety critical cases secure assurance will be required.  

Alongside the digital infrastructure, information from sensing networks and a variety of sources (for example traffic flow monitoring, weather data, event data etc.) will need to be reliably integrated. This implies secure and robust sensing elements on our physical infrastructure, and effective communications across the network – not always easy when considering the scale of the transportation network. 

Data is not a big homogeneous thing – for example cybersecurity data is very sparse and signals can be hard to spot. Day-to-day operational data may be full of details that need to be parsed out for monitoring, and data to gather for training will likely need some pre-processing etc. It’s important to have the right processes in place for acquisition, curation, pre-processing, protection and post-processing of data and ensure all the necessary data workflows are in place. 

Backwards compatibility will likely be necessary, particularly for longer lived infrastructure – furthermore this will also play into interoperability between jurisdictions – this encompasses everything from data standards, to ensuring that AI models are free of drift (or at least measuring it), to compatibility of AI decisions with the cyber-physical systems that infrastructure is made of (e.g. ensuring sensors are still accurate, that any environmental markings such as road signs are well maintained).  

Agility – both in process, in adoption, in people and in operations (e.g. MLOps) is key since the AI landscape moves quickly, and there are many moving parts to the transport ecosystem, whether that be supply chain, construction, operations, vehicles etc. Being able to update against an emergent threat or hazard or change quickly for efficiency purposes will make all the difference. 

Monitoring systems that measure ‘dataset drift’ can be overlooked – these are key to understanding exactly when the real world has changed so much that an AI system is no longer operating within its training data – at which point the assurance argument needs to be re-established. These or similar design domain breach detection systems should be a part of any long-term deployment with safety implications 

Not “infrastructure” per se but effective training and adoption of tools by the key staff involved in managing the transportation routes and hubs is essential to build and maintain a trusted and effective infrastructure. 

How can public and private sector organizations close the technical skills gap and build internal AI literacy to ensure long-term success in AI adoption? 

Building a responsible AI culture is key to supporting long-term success of AI adoption. Skills gap analysis and role redefinition (e.g. introducing AI aspects to existing job descriptions) can help bridge the divide. This plus technical upskilling helps avoid a siloed ‘AI team’ which doesn’t necessarily have the buy-in from the rest of the organization and processes. 

Another part of achieving a responsible AI culture is creating and implementing an organization-wide AI literacy or learning program that can ensure the workforce has the capabilities to leverage AI responsibly and effectively in their roles. This should be tied to organizational strategy and linked to organizational goals and challenges. 

In designing a structured AI learning program, some examples of key areas of focus include assessing anticipated upskilling of the workforce and preparing a skills development plan, defining objectives and KPIs for AI in transportation that are linked to organizational goals, differentiating learning by role and seniority and embedding mechanisms for feedback that allow for continuous improvement. 

What role can AI play in mitigating the environmental impact of transportation systems? 

The possibilities are broad – from dedicated large data models discovering novel materials for batteries or sustainable fuel to route optimization for fleet vehicles, in-depot applications around maximizing efficiency or operation improvements at logistics hubs, through to system optimization across wider multimodal transport using new shared data systems for training data.  

As AI becomes more embedded in mobility systems, how can organizations ensure that ethical principles are built in from the start in terms of bias, access, and fairness? 

By the time an AI system is viable there may already be assumptions about what datasets will be used (or have already been used!) – so dataset bias review or bias audits for discriminatory outcomes are critical to establishing a foundation for the future.  

Explainable or interpretable models may be necessary to understand and adapt decision processes to make equitable decisions. Universal design principles, and affordability modelling can all play a part.   

Structured assurance frameworks can help identify ethical issues alongside making sure AI is effective, transparent, and safe. 

Many AI projects remain in the pilot phase. What practical steps can transportation agencies take to move from experimentation to full-scale, real-world deployment? 

In general terms, thinking about the intended aim of a pilot from the outset – if it’s successful what needs to be true for this to deploy at scale and be effective. That’s a combination of broad topics; regulatory, scalability, cost effectiveness, underlying infrastructure, IT and OT, sensing, data and communications. And scaling isn’t just about volume – it’s also the buy-in from a wider set of stakeholders to allow for integration and compatibility between systems, models and decision-making processes. 

Standards (or selecting from emerging standards) are key, enabling integration and allowing for improved competition between suppliers. They need to be tied into cohesive legislation that addresses the transport industry as a whole, and are potentially a suite of standards covering DataOps, SecOps, MLOps, quality and safety – all while integrating with existing workflows. 

Plans need to be in place to train people effectively, for the business case at scale – bearing in mind this may be ROI in terms of spend, but also safety or displacing tasks that are hard to resource.  

Finally, resilience is key for system scales, balancing cost and redundancy – requiring upfront implementation before being tuned in the optimization stage.   

These themes are the core of our ‘AI deployment guide’ for transportation organizations – in this we define a pathway to AI maturity including a ten-point action plan for success.  

Looking ahead, how do you see AI reshaping the transportation landscape and what should decision-makers be doing now to prepare for that future? 

Delivering safer, more efficient transportation hinges on our ability to embrace and responsibly deploy transformative technologies like artificial intelligence (AI). AI is poised to become a critical enabler of many shared goals within the transportation sector, such as safety, efficiency, economic competitiveness, and mobility. 

Successful AI implementation in transportation organizations relies on several critical building blocks. They are essential for creating a robust and scalable AI pipeline that can support current and future needs. These building blocks are Leadership, Processes, Data, Technology, People, and Effective Delivery. 

These building blocks are not standalone components; they are deeply interconnected. For example: high-quality Data feeds into Processes, ensuring accurate governance and model reliability.  A skilled People function supports the Technology building block by driving innovation and execution. 

In addition, some elements depend on external factors like system interoperability standards, digital infrastructure, connectivity, and evolving regulations. Achieving these requires collaboration with stakeholders, including industry partners, academia, and other transportation organizations.  

Leaders in transportation organizations must adopt a holistic approach, integrating these building blocks to create a unified, scalable AI pipeline. This will ensure readiness for both current AI demands and future innovations, enabling organizations to enhance safety, efficiency, economic competitiveness, and mobility in the transportation sector.   

www.cambridgeconsultants.com  

Dr Sinan Yordem is Senior Vice President, Head of North America Advisory, Energy & Smart Infrastructure at Cambridge Consultants. Sinan grows and develops businesses through the power of innovation. With more than 15 years’ experience of leading start-up and corporate growth initiatives, he’s delivered breakthrough physical products and digital services to government and industry clients across the world.