Tracks of Change Part 5: The AI Paradigm Shift

Across this series we have watched a single railway pass through a long line of custodians, and we have read each handover as an asset management problem. Part one traced the public era: Victorian Railways, the 1919 electrification, the integrated authorities of the 1980s and the Public Transport Corporation built the asset and held it as a public good. Part two examined the franchising experiment that began in 1999, when the network was split into Bayside and Hillside and the State learned, abruptly, what it meant to contract an asset rather than run it. Part three covered the reunification under Connex and the arrival of Metro Trains Melbourne in November 2009 under the round we now correctly label MR3, renewed as MR4 in 2017, an era defined by performance regimes, punctuality targets and penalty payments. Part four tracked the transformation MR4 has carried: the Metro Tunnel, the High Capacity Metro Train fleet, and Australia's first brownfield high capacity signalling.

Each era set the asset management challenge for the next. The public era built a complex physical asset but left open the question of how to extract value from it. Franchising answered that with contracts, but contracts created a new problem: how to hold a private operator accountable for an asset the State still owned. Performance regimes answered that with measurement. And measurement, pursued through a decade of capital investment, has now produced something none of the earlier eras possessed: a data rich, instrumented, technically sophisticated asset base. This finale is about what that makes possible, and about a question the people of Melbourne have just started asking out loud.

A taste of free public transport

In autumn 2026 Victorians experienced something the network had never offered. From late March, in response to a fuel price shock driven by conflict in the Middle East, the Allan Labor Government made all public transport free across the state, suspending Myki fares on every metropolitan and V/Line train, tram and bus. The measure was framed as temporary cost of living relief, a fast domestic lever to take cars off the road while petrol pushed past two dollars fifty a litre. Take up was strong enough that the government extended the free period by a further month, and then stepped down to half price fares from 1 June through to the end of 2026, roughly halving a typical daily cost from $11.40 to $5.70.

The Premier was careful to call it temporary: free public transport "was always a temporary measure to help Victorians right now". But the genie is out. Melburnians have now lived a few months of turning up to a station and simply travelling, and the public appetite for keeping fares low, or removing them altogether, has not gone away with the fuel price. That appetite is the new political fact this finale has to reckon with.

The funding question, framed honestly

If the conversation is going to mature beyond a slogan, it has to start from how the network is actually paid for. Free public transport is not magic, and it is not really a fares question. It is a subsidy question.

Under the franchise model the State already pays the operator a large operating subsidy every year. Connex, the predecessor operator, received on average about $345 million a year; the current arrangements are larger again. Fares only ever covered part of the cost of running the railway. So "free PT" does not mean conjuring a service from nothing. It means the State deciding to fund the share that passengers currently pay at the gate, on top of the subsidy it already pays, and doing so permanently rather than for a crisis month.

That reframes the whole debate. The real question for government is not "can we afford to be generous". It is: how do you cost out of the Melbourne Train Refranchising model an appropriate level of subsidy, sized to the actual cost of serving each passenger, that could make public transport free or near free without the bill running away? Free PT is fundamentally a question of subsidy level and cost to serve, not a question of will. And the lever that decides whether that level is affordable is one this series has been circling all along: how well the asset is run.

Why AI-enabled asset management is the lever

Here is where the AI paradigm shift stops being a technology story and becomes a public finance one. The size of the subsidy needed for any given level of service is set by the whole of life cost base of the network. Drive that cost base down and the same service needs a smaller subsidy. A smaller subsidy is what makes free, or near free, fares fiscally plausible. AI-enabled asset management is the most credible route to driving it down.

The network the next operator inherits is, for the first time, deeply legible. The Metro Tunnel opened on 30 November 2025, with the full timetable switch on 1 February 2026. The 70 strong High Capacity Metro Train fleet streams telemetry from traction, braking, doors, HVAC and bogies. The moving block high capacity signalling on the Cranbourne and Pakenham lines continuously knows where every train is and how it is behaving. New tunnels and stations report the state of their own ventilation, drainage, fire and power systems. Layered over decades of maintenance and condition records, this is more data, at finer resolution, than any earlier era of this railway could have imagined.

Turn that data into intelligence and the cost base moves in four concrete ways. Predictive and condition based maintenance triggers intervention from an asset's actual state and trajectory rather than the calendar, cutting in service failures and the unplanned downtime that is always the most expensive kind. AI-optimised renewal prioritisation ranks competing interventions by risk and whole of life cost, extending asset life and deferring or avoiding capital that would otherwise be spent early. Digital twins and remaining useful life modelling cut over maintenance, the quiet waste of servicing things that did not yet need it. And better, governed asset information lowers the cost to serve per passenger kilometre across the whole portfolio. Lower cost base, smaller subsidy for the same service, free PT becomes something a treasury can actually model.

AI also helps government size and target the subsidy rather than write a blank cheque. Demand modelling and utilisation analytics show where patronage actually is, which corridors and times carry the load, and what an extra dollar of subsidy buys in travel. That is how you fund an appropriate level, calibrated to cost to serve and real demand, instead of a politically round number that balloons.

The challenge, and the win, for the next operator

The Melbourne Train Refranchising programme is live. With MR4 running to the end of 2026, the State has launched MR5, the fifth refranchising round. It is a competitive re-tender that has not yet been awarded. As of early 2026 the Department of Transport and Planning has shortlisted three consortia: Go Places Melbourne (Go-Ahead Australia, ACCIONA Rail and Tokyo Metro), RATP Dev JH Downer (RATP Dev, John Holland and Downer), and Melbourne Grow (MTR Corporation Australia, CRRC and DT Infrastructure). No operator has been selected. The term is reported as up to 15 years, the new contract is due to commence in late 2027, and the contract value is being established through the procurement now under way. What is being competed for is the right to run more than 2,000 daily services across roughly 1,000 km of track, an expanding fleet, seven maintenance depots, more than 220 stations and close to 7,000 employees.

Whoever wins inherits the richest asset dataset in the network's history, and a public that has tasted free travel. That is the win and the pressure at once. A custodian who turns telemetry, signalling data, condition monitoring and maintenance records into foresight can move maintenance from a cost centre to a value lever, lifting availability, deferring renewals safely, and shrinking the failures that punctuality regimes only ever measured after the fact. Every dollar taken out of the cost base is a dollar that does not have to be found in subsidy if the State chooses the low fare path.

But the same inheritance is the challenge, and it is an asset management challenge before it is a technology one. AI is only ever as good as the asset information beneath it. Telemetry is voluminous but not automatically trustworthy, and a model trained on dirty data forecasts confidently and wrongly. ISO 55001's discipline of line of sight, from organisational objectives down to the asset register and the data describing each asset, is the difference between a coherent information base and a thousand disconnected dashboards. A moving block signalling line and a 1919 era corridor sit in the same portfolio, and the intelligence has to span both. And there is the operating model: the skills, the governance and the willingness to act on a model's recommendation, responsibly and accountably, rather than collecting insight no one is empowered to use.

There is a further wrinkle peculiar to a recently transformed network. New assets are not the same as understood assets. A tunnel commissioned in late 2025 has barely begun to reveal its degradation behaviour, and a signalling system in its first years carries a regime built on design assumptions, not yet on lived failure data. AI thrives on history, and the newest, most valuable assets have the least of it. The discipline of the first few years, capturing a clean baseline, will shape every model that follows.

The SAS-AM thesis

Here is our point of view, and it is the thread running through all five parts. AI does not replace asset stewardship. It amplifies it. Good asset management, a sound register, honest condition data, clear line of sight from objectives to interventions, becomes dramatically more powerful when AI can reason over it at scale. And poor asset management is, if anything, more ruthlessly exposed. An AI layer over weak data does not hide the weakness. It industrialises it, producing precise, automated, confident error.

That is also, precisely, the asset management case for free public transport. Free or near free fares are only sustainable if the underlying asset base is run intelligently and cheaply. Run it well and the subsidy needed for a given service falls into a range a state can defend. Run it poorly, on dirty data and over maintenance and avoidable failures, and the cost to serve stays high, the subsidy required for free travel balloons, and the policy collapses under its own weight. The appetite Melburnians discovered this autumn is real. Whether it can ever be more than a crisis measure depends less on political generosity than on how cheaply and intelligently the railway is run.

A 15 year franchise is a long commitment placed against far longer asset lives, and the alignment problem we have returned to throughout this series does not disappear in an AI era. It sharpens. So our counsel to whoever wins MR5, and to the State that awards it, is consistent. Treat asset information as a first class asset in its own right, governed to ISO 55001 line of sight, not as exhaust from the operation. Invest early in a clean, shared condition baseline, because the next operator's models will only ever be as good as the data captured now, while the network is young. Build the operating model, the people and the governance, to use AI as decision support for stewardship, not a substitute for it. Do that, and the conversation about affordable, even free, public transport rests on something solid. The technology is arriving. Whether it pays off over the coming decades, and whether it can fund the railway Melburnians have just glimpsed, will be decided, as every transition in this series has been, less by who runs the trains than by how well they look after the railway.

Start the series again: Part 1 — The Public Era

SAS Asset Management. We provide advanced analytics, expert asset management services and maturity assessments to help asset owners realise their value. If your organisation is preparing to turn an instrumented asset base into asset intelligence, talk to us.