From Wishlist to Roadmap: What Transport AI Strategies Really Need
Key Takeaways
The Problems:
- AI has significant potential to improve transport across a wide scope of uses.
- Developing a Transport AI Strategy to think strategically about the risks and opportunities that AI provides is a good idea.
-
A good transport AI strategy will:
- Set out a clear vision.
- Set SMART objectives.
- Set clear priorities.
- Focus on outcomes, not processes.
- Manage risks appropriately.
- Know when to use pilots and when to deploy.
-
A good transport AI strategy will also develop enablers for success:
- Avoiding the innovation vetocracy.
- Accessing and developing the right skills.
- Learning from others and not reinventing the wheel.
- Open Data
What Next?
Do you need to develop or improve your Transport AI strategy?
Introduction
Transport organisations worldwide are grappling with the same challenge: how to harness artificial intelligence (AI) whilst managing the risks.
In my latest newsletter, I provided an overview of the UK’s recently published Transport AI action plan and expressed my disappointment with it. It reads more like a wishlist than a roadmap to success.
AI has genuine potential to transform how we move people and goods, from optimising bus and rail timetables, to enabling autonomous public transport, to revolutionising maintenance schedules. But realising this potential requires more than good intentions; it demands the same strategic rigour we should apply to any major strategy.
This post outlines what a robust Transport AI strategy should actually contain, the practical framework that will help transport organisations make smart decisions about where, when, and how to deploy AI technologies.
Many thanks to Muriel Demarcus, who is conducting extensive work on AI in transport around the world and provided extensive input into this post. If you want to learn more about transport and AI, she is a great person to talk to.
The Strategic Foundation
Every effective transport AI strategy needs to answer three fundamental questions: Where are we going? How will we get there? How will we know we've succeeded? The components below provide the framework for answering these questions systematically.
1. Vision: Your North Star
Your vision should guide decisions. The UK's "Responsible AI embedded in a resilient transport system delivering cheaper, cleaner, and safer journeys for all" has too many buzzwords for my liking.
Compare this to a simpler alternative: "Using AI to make transport work better."
The best visions are memorable enough that staff can recite them and specific enough to exclude initiatives that don't align with them.
2. Objectives
Vague objectives are strategy's biggest enemy. They allow organisations to claim success while achieving nothing meaningful. SMART objectives prompt uncomfortable conversations about timelines, resources, and accountability, exactly the conversations that lead to results.
Instead of "explore AI opportunities in maintenance," try "Implement predictive maintenance using AI on 50% of our rail network by December 2026, reducing unplanned failures by 30%."
These specific targets create urgency, enable resource planning, and provide clear success metrics. They also reveal whether you're rhetorical levels of ambition are real.
3. Prioritisation
AI can potentially improve everything in transport, which is precisely why you can't pursue everything simultaneously. Effective prioritisation means choosing which opportunities to pursue first, second, and not at all.
This requires a robust evaluation framework that assesses each potential AI initiative against consistent criteria:
- Problem significance: How much pain does this solve?
- Implementation complexity: What's the technical and organisational lift?
- Resource requirements: What will this actually cost?
- Success probability: How confident are we that this will work?
- Strategic alignment: Does this advance our broader goals?
The goal isn't to find perfect initiatives, it's to make transparent trade-offs based on your organisation's circumstances and risk appetite.
4. Outcomes Over Activities
Transport strategies often confuse motion with progress, listing activities such as "establish working groups" and "conduct feasibility studies" without connecting them to tangible, real-world improvements.
Effective AI strategies will focus relentlessly on outcomes, including reduced travel times, improved safety records, lower operating costs, and enhanced accessibility. Every activity should contribute directly to these measurable improvements.
This means being honest about whether an initiative will deliver tangible benefits or just create the appearance of progress.
5. Risk Management: Balancing Caution and Progress
AI presents both tremendous opportunities and significant risks, ranging from job displacement to cybersecurity vulnerabilities and algorithmic bias. The key is developing a risk framework that enables informed decisions rather than paralysis.
This means categorising risks by probability and impact, identifying mitigation strategies, and establishing clear risk tolerances. An organisation with zero risk appetite will miss AI's benefits entirely, while one that ignores risks may face serious consequences later.
Your strategy should explicitly address how you'll handle the most significant risks specific to your context, not just acknowledge that risks exist.
6. From Pilots to Production
Pilots serve a purpose: testing feasibility and building internal confidence. But too many transport organisations treat pilots as destinations rather than stepping stones. The result is an endless series of small-scale experiments that never scale to create a meaningful impact.
Effective strategies distinguish between what needs piloting and what can be deployed directly. If a technology is proven elsewhere, spending two years on local pilots may be unnecessary caution.
Before launching any pilot, define success criteria and the pathway to full deployment. If you can't articulate how a successful pilot leads to organisation-wide implementation, question whether the pilot is worth pursuing.
Breaking Through Implementation Barriers
Having a solid strategic foundation is necessary but not sufficient. Transport organisations face challenges that can derail even well-planned AI initiatives. These enablers address the most common failure points.
1. Escaping the Innovation Trap
Transport organisations can become vetocracies, where many people can stop initiatives when it comes to innovation and change, and implementing AI is all about innovation and change. Therefore, you need to think carefully about how to avoid the vetocracy.
Successful organisations adopt a distributed approach: empower individual teams to experiment with AI within defined boundaries while maintaining central coordination to share learnings and avoid duplication. The central team becomes an enabler, not a gatekeeper, providing guidance, resources, and expertise rather than approval processes.
2. Solving the Talent Puzzle
The AI skills shortage is real, and transport organisations aren't typically top destinations for AI talent. The solution isn't trying to compete with tech companies; it's being strategic about which capabilities to build internally versus access externally.
You need enough internal expertise to be an intelligent customer: people who can evaluate AI vendors, interpret results, and evolve systems over time.
The most effective approach combines hiring a small core of AI-literate staff with partnerships that provide specialised capabilities.
Equally important is upskilling existing staff to work effectively with AI systems. This doesn't mean turning everyone into a data scientist, but rather ensuring they understand AI capabilities and limitations well enough to use these tools effectively.
3. Learning at Scale
The AI landscape evolves rapidly, making it impossible for any single organisation to stay current through internal efforts alone. Successful organisations will build systematic approaches to external learning.
This involves actively engaging with other transport organisations tackling similar challenges, following academic research relevant to your use cases, and monitoring private sector developments that may apply to your context.
Establish formal mechanisms for this learning, including regular benchmarking exercises, participation in industry working groups, partnerships with universities, and structured vendor engagement programs. The goal is making external learning routine.
4. Data as Infrastructure
Long before people were focused on AI, there was a desire to provide open data on transport that would build trust, spark innovation and encourage collaborative problem-solving. This becomes an even greater imperative with AI, and the transport organisations that can make data openly available will have an edge in harnessing the benefits from AI.
The key is moving from data as a byproduct of operations to data as a deliberately managed resource that enables better decision-making, both internally and across the broader transport ecosystem.
Conclusion
The potential of AI to transform transport is undeniable, but realising this potential demands strategic planning grounded in proven strategy fundamentals.
A well-crafted Transport AI strategy provides the roadmap for delivery. By establishing a clear vision, setting SMART objectives, and making tough prioritisation decisions, transport leaders can move beyond buzzword-heavy documents to actionable plans that deliver real transport benefits.
The enablers: avoiding an innovation vetocracy, building the right skills mix, learning from others, and embracing open data are also critical. These aren't just nice-to-haves; they're the foundation that will determine whether your AI initiatives succeed.
The question isn't whether AI will reshape transport, it’s whether you will proactively shape its impacts or find yourself closing the stable door after the horse has bolted.
The transport organisations that invest time now in developing robust AI strategies alongside deploying AI will be the ones that harness AI's full potential while managing its risks responsibly.