The intelligence you already own: preparing transport data for AI

Every ticketing tap, vehicle movement, sensor reading and maintenance record adds to a growing picture of how a transport network performs. Operators also hold information from GPS, traffic signals, cameras, asset inspections and work orders. Together, these sources can reveal patterns in demand, asset condition and network risk.

However, collecting more information does not automatically lead to better decisions. Data often remains within the system or team that created it. By the time decision makers need a network-wide view, they may be comparing spreadsheets, reconciling different definitions or relying on information that is no longer current.

The opportunity is to make better use of what already exists. When organisations can access consistent information and understand its context, they can direct attention and investment towards the decisions that matter most.  

AI can process large amounts of information quickly, but speed does not correct poor inputs. Incomplete, inconsistent or outdated data can produce outputs that appear convincing without being dependable. For transport owners and operators, that can affect investment choices, maintenance priorities and decisions with safety implications.

A practical information management approach starts by defining what trusted data means. Information should be accurate enough for its intended use, current, controlled and available to the right people. Its ownership should also be clear, including who maintains it, who can change it and who is accountable for its quality.

This work may feel less visible than an AI pilot, but it provides a reliable basis for analytical tools. AI should support professional judgement, not create false confidence in information that has not been checked.  

Reliable information can still have limited value when it stays isolated. A maintenance team may understand asset condition, a planning team may hold growth forecasts and an operations team may track daily performance. Each dataset can be useful on its own, yet many transport decisions cut across all three.

Shared definitions, common data models and connected systems help information retain its meaning as it moves between teams. This creates a clearer view of the network and makes it easier to test how one decision may affect other outcomes.

National Highways in England addressed fragmented asset information through a common data model and joined-up digital representation of the network. This supported applications including pavement assessment and flood and rainfall modelling. In Sydney, an AI engine developed by CSIRO’s Data61 with Transport for NSW uses existing sources, including Opal card taps, GPS and signal data, to predict congestion across the network.

Spatial digital twins in New South Wales and Queensland also demonstrate the value of bringing transport, infrastructure, planning, property and environmental information into a shared view. Technology matters, but the wider benefit comes from giving teams a common basis for decisions.  

Technology is only one part of AI adoption. Organisations also need to decide who owns the data, who reviews the output and who remains accountable for consequential decisions. These questions become more important as AI moves from analysis towards systems that can recommend or initiate actions. 

People must remain responsible for applying context, testing assumptions and considering whose experience is represented in the data. Transport information is not neutral. If some journeys, communities or network conditions are poorly captured, the resulting analysis may reinforce those gaps. 

Good governance sets boundaries for how information is used and gives decision makers a clear path for reviewing outputs. AI can connect information across disciplines, but transport professionals still make the decisions that shape safe, practical and community-focused outcomes.  

Long-term value comes from helping infrastructure owners manage their own information and tools. A solution that remains dependent on an external provider can limit an organisation’s ability to adapt, scale or challenge its outputs.

Our work with Tauranga City Council illustrates this approach. The council’s digital asset information had arrived late and in fragmented forms, creating reliance on consultants. We transferred ways of working and built capability progressively so the council could manage federated 3D models and run its own clash detection. This contributed to an increase in first-time-right asset data handover from 35 percent to 80 percent.

This approach shifts the focus from handing over a black box to building the capability to manage it internally. As that capability grows, external support can be reduced. The organisation retains the knowledge, governance and confidence needed to use its information in business-as-usual decisions.

Transport organisations do not need to begin with a large AI transformation programme. A practical first step is to focus on one decision-critical dataset: 

• Choose one dataset that already informs a difficult or high-value decision. 
• Assign accountability to a person or role, rather than to a system. 
• Record what accurate, current and complete enough means for that decision. 

From there, organisations can improve the dataset, connect it to related information and test how analytics or AI could support a defined decision. This keeps the work practical and makes progress easier to assess. 

The intelligence needed to improve transport decisions may already exist in the information organisations hold. Trusted, connected data and clear accountability give AI a stronger basis to extend specialist knowledge rather than replace it.