Transport AI ROI: 2026 results

In 2024 and 2025, transport leaders were expected to move quickly on AI. Now, in 2026, expectations are more grounded.

The focus is on measurable return.

At the same time, operational pressure has not eased. Congestion, disruption, recruitment challenges and rising costs continue to test existing models. Margins remain tight. Regulatory scrutiny remains high. Add those together, and technology investment is assessed with the same discipline as fleet deployment, labour allocation and network planning.

Boards are reviewing cost per vehicle mile, fleet availability, first contact resolution, claims exposure and margin stability more closely than before.

Against that backdrop, new capability exists across the sector. The priority is whether it supports earlier decisions and better use of the data already generated inside operations. In some organisations, it is influencing maintenance planning, demand forecasting and service management. In others, it sits outside day-to-day workflow.

We spend our time with engineering directors, CIOs and COOs working through these pressures. Across those conversations, a consistent picture is emerging of where ROI is being realised and where progress remains uneven.

Here is how that is playing out across key parts of transport operations.

Predictive maintenance and fleet ROI

Fleet reliability sits directly on the P&L. When vehicles fail in service, the impact moves quickly through cost per vehicle mile, overtime, lost mileage and service penalties. Even a small drop in availability can affect contract performance.

You can see the problem here. Most operators already hold telematics, diagnostic and maintenance history data. The data isn’t new. Meanwhile, servicing is still largely mileage-based. Vehicles are taken off the road on schedule, whether they need it or not. At the same time, early warning signs on other vehicles are missed.

In other words, the information exists. The intervention often doesn’t.

So instead of relying purely on fixed intervals, more operators are moving toward condition-led decisions. Fault probability is assessed against Mean Time Between Failure and component lifecycle data. Engineering teams can then prioritise work during planned downtime rather than reacting to breakdowns in service.

As a result, the operational impact becomes measurable:

Fewer roadside recoveries
Higher fleet availability
Reduced emergency repair spend
Improved technician utilisation
Better alignment between parts ordering and actual vehicle condition

Across a large fleet, a one to two point improvement in availability can materially influence cost per vehicle mile over twelve months.

But there’s a caveat: modelling alone does not change outcomes. Alerts need to feed directly into planning routines. Depot leadership needs to review availability and utilisation against agreed thresholds. Without that integration, gains are uneven.

Then, and only then, are you ready to see reliability discussions move from explaining breakdowns to demonstrating control.

What we see in transport: For the transport operators we support, performance improves when predictive outputs are embedded into daily scheduling and parts management rather than treated as a separate reporting layer.

Contact centre automation and Cost Per Contact

For many operators, this is where operational pressure becomes most visible. Disruption drives spikes in demand. Customers expect immediate updates. Recruitment remains tight. At the same time, cost per contact is under sustained review.

The issue is not volume alone. It is repetition.

A large proportion of inbound calls relate to the same queries. Timetable updates. Delay explanations. Refund status. Ticket access. During disruption, identical questions arrive within minutes of each other. Skilled agents end up repeating information that already exists in live systems.

That creates cost without improving service.

A different approach starts by isolating those high-frequency interactions and redesigning how they are handled. When that work is done properly, you can see the benefits in the service data:

Average Handle Time (AHT) begins to fall
First Contact Resolution (FCR) improves
Containment rates increase
Abandonment rates stabilise
Cost per contact reduces

Across a large network, even incremental improvement in cost per contact carries weight over twelve months.

There is a practical constraint, however. Automation layered onto disconnected systems produces uneven results. Call drivers must be understood in detail. Real-time service data must be accessible. Escalation routes must be defined clearly so complex cases reach experienced staff quickly.

What we see in transport: In the organisations we work alongside, progress comes when automation is treated as part of service management rather than as an overlay. Operational leaders retain ownership of service levels. Financial thresholds are agreed before rollout. Reporting reflects both cost and quality.

Once those foundations are in place, repetitive demand declines, and experienced agents spend more time resolving cases that require judgement.

That is where cost control and service stability begin to reinforce each other.

Safety analytics and claims exposure

When incident rates rise, the financial impact follows quickly. Insurance premiums increase. Claims volumes climb. Regulatory attention intensifies. Public confidence becomes harder to protect.

The challenge is timing.

In many operations, safety reviews still follow incidents rather than precede them. Data is collected consistently — telematics, CCTV, driver reports — yet it is often reviewed retrospectively. By the time trends are formally examined, the opportunity to intervene early has passed.

Consider how minor collision frequency develops. A depot records a steady rise over several months. Harsh braking events increase on specific routes. Driving behaviour in poor weather shows deterioration. Each data point sits in isolation until someone connects them.

When that information is examined together and acted upon promptly, the response becomes more precise. Coaching can be scheduled earlier. Route conditions can be reassessed. Vehicle checks can be prioritised. Supervisors can intervene before claims accumulate.

The operational effect is measurable:

Reduced preventable collisions
Lower vehicle damage costs
Greater stability in claims ratios
Improved position in insurance negotiations
Fewer regulatory escalations

Financial exposure tightens when incident frequency falls.

There is also a cultural dimension. Surveillance without transparency erodes trust. Safety data must be handled with clear governance and communicated properly to drivers and front-line teams. Without that discipline, adoption weakens.

What we see in transport: Safety analytics functions best as a risk management tool embedded into operational review cycles. Depot leadership remains accountable for outcomes. Insurance performance is tracked alongside service metrics.

When safety intervention moves earlier in the cycle, claims exposure reduces and regulatory conversations become more grounded.

Engineering optimisation and technician utilisation

Engineering capacity is under strain across much of the sector. Skilled technicians remain difficult to recruit. Backlogs accumulate quietly. Emergency repairs interrupt planned work. Parts delays extend downtime. Workshop productivity feeds directly into cost per vehicle mile and fleet availability.

What often goes unexamined is how work is allocated.

In many depots, scheduling still depends heavily on manual judgement. Jobs are assigned based on who is available rather than on estimated repair time, vehicle criticality or technician skill mix. That approach works when pressure is low. Under sustained demand, it begins to expose inefficiencies.

Take backlog growth. It rarely happens overnight. Instead, smaller delays compound. A vehicle stays off-road longer than planned. Overtime increases. Parts orders miss alignment with actual demand. Utilisation drifts without being noticed immediately.

Bringing scheduling data together with repair history and risk indicators changes the cadence of workshop planning. Vehicles with higher failure probability can be prioritised. Repair time estimates can inform allocation decisions. Parts forecasting can align with expected workload rather than historic averages.

When that discipline is applied consistently, several metrics respond:

Technician utilisation percentage
Average repair cycle time
Backlog volume
Overtime expenditure
Parts inventory turnover

Even modest movement in utilisation across a large fleet influences labour cost materially over a year.

There is a practical boundary, however. Tools alone do not stabilise workshops. Depot managers still need to review utilisation against defined thresholds. Scheduling outputs must feed directly into daily rosters. Inventory decisions must follow planned work.

What we see in transport: In the operators we work alongside, optimisation succeeds when it supports management oversight rather than attempting to replace it. Workshop leadership remains responsible for output and availability.

As backlog shortens and turnaround time improves, fleet availability follows. From there, cost control becomes more predictable.

Network forecasting and revenue stability

Network design affects financial performance long before it appears in monthly reporting. Frequency decisions, fleet allocation, depot capacity and franchise bids all rely on demand assumptions. When those assumptions are wrong, revenue and cost follow.

Recent years have made forecasting more difficult. Peak demand no longer follows historic curves. Off-peak travel moves around more than it once did. Subsidy structures and franchising arrangements continue to evolve. Planning cycles built on older assumptions now carry greater financial risk.

Most operators already hold the relevant data. Passenger counts, ticketing records, punctuality data and event calendars are available. The challenge is how regularly that information influences scheduling and fleet decisions.

Take route performance reviews. In many organisations, formal reassessment still happens once a year or after a sustained decline. By that stage, load factor has already weakened and revenue per vehicle mile has softened. Adjustments then happen under pressure.

More frequent forecasting changes the timing of those decisions. When projections are reviewed consistently and tested against current performance, capacity can be adjusted earlier. Fleet deployment can be aligned with current demand. Bid assumptions can reflect recent travel behaviour.

The commercial indicators respond:

Load factor stability improves
Revenue per vehicle mile strengthens
Underperforming routes are addressed sooner
Fleet deployment becomes more deliberate
Franchise submissions carry stronger demand evidence

Even modest movement in load factor across a network influences revenue materially over a financial year.

Forecasting only delivers value when it informs action. Scheduling must follow updated projections. Fleet investment must reference forward demand. Review cycles must be maintained.

What we see in transport: In the operators we work alongside, forecasting supports capital allocation as much as timetable planning. Fleet purchases, depot expansion and staffing levels are considered against projected demand, not past norms.

When deployment aligns more closely with demand, revenue volatility reduces and financial planning becomes steadier.

Where AI ROI remains inconsistent in transport

Some programmes are delivering measurable return. Others continue to struggle.

The difference often becomes visible during implementation.

Operational ownership is unclear

Initiatives frequently begin inside digital or innovation teams. Operational leaders are consulted, yet the performance target does not always sit directly with them.

Engineering may receive new systems while availability targets remain unchanged. Contact centres may introduce automation without adjusting service level accountability. Finance may review outcomes after deployment instead of agreeing thresholds beforehand.

When responsibility is shared too broadly, momentum slows.

In operators achieving consistent return, the person responsible for the metric also leads implementation. If cost per contact is expected to move, the service leader carries that responsibility. If fleet availability is under review, engineering leadership owns the target.

Clear ownership alters behaviour quickly.

Further reading: The 2026 operational mismatch in transport

Financial thresholds are not agreed early

Some pilots begin without documented starting points.

Current cost per vehicle mile.
Current backlog volume.
Current FCR.

Without agreed baselines and target movement, evaluation becomes difficult. Conversations drift toward opinion rather than measurable change.

Stronger programmes document financial and operational thresholds before deployment. Review cycles are scheduled. Scaling decisions reference agreed performance measures.

Process discipline is inconsistent

Introducing new systems into inconsistent processes produces inconsistent outcomes.

If maintenance planning varies by depot, optimisation cannot stabilise it. If contact centre workflows differ by shift, automation reflects that variation. If safety reporting lines are unclear, analytics does not resolve it.

Leadership teams seeing sustained return address operational discipline first. Technology then reinforces structured practice.