AI-Powered MaaS: Practical Steps for Cities in 2026

A MaaS pilot can fail even when the city has buses, trains, bike share, and ride-hailing already on the streets. The usual culprit isn’t a lack of transport supply—it’s fragmentation: separate tickets, mismatched timetables, walled-off data, and incentives that push people back into private cars.

Mobility-as-a-Service (MaaS) promises something simple: plan, book, and pay for a whole trip across multiple operators in one place. But in practice, MaaS only becomes truly useful when the digital layer is strong enough to coordinate the physical layer. That’s where AI in smart cities stops being a buzzword and becomes infrastructure.

This post sits inside our series “Mākslīgais intelekts publiskajā sektorā un viedajās pilsētās” and focuses on one very practical question public-sector leaders keep asking: what does it actually take to make MaaS work at city scale—reliably, fairly, and sustainably?

MaaS succeeds when it’s defined as a service, not an app

MaaS works when it’s treated as a service model—with clear user promises and accountable governance—not a single “super-app” procurement.

The SmartCitiesWorld podcast discussion highlights a real issue: definitions of MaaS are fluid, and that ambiguity kills adoption. If residents don’t understand what MaaS offers (and what it doesn’t), they won’t change habits.

Here’s a practical definition I’ve found useful for public-sector planning:

MaaS is an integrated mobility service that lets a traveler plan, compare, book, and pay for multimodal trips, with consistent rules and support across operators.

That definition forces a city to answer uncomfortable questions early:

• Are fares and entitlement rules consistent across modes?
• What happens when a connection is missed—who provides support?
• Do we optimize for speed only, or also emissions, accessibility, and equity?

If those aren’t answered, you don’t have MaaS—you have a trip-planning interface.

Why AI matters at the definition stage

AI changes what “integration” can mean. Without AI, integration is often limited to static schedules, basic routing, and coarse disruptions.

With AI, MaaS can offer decision support that reflects real conditions:

• Predictive delays and connection risk (not just “scheduled arrival”)
• Personalized accessibility routing (elevators, curb cuts, safe crossings)
• Emissions-aware options (nudging toward lower-carbon choices)

That’s not window dressing. It’s how MaaS becomes trustworthy enough for commuters.

The real barrier: interoperability (and AI can’t fix politics)

MaaS implementations usually stall at interoperability—not because the algorithms are hard, but because the ecosystem is.

Public transport authorities, private operators, and platform vendors all have different incentives. The podcast emphasizes the role of public–private partnerships like the Urban Mobility Partnership, and I agree with the premise: MaaS is a coalition project.

Still, a hard stance is needed: cities should treat mobility data and ticketing interfaces as civic infrastructure. If the city can’t set baseline rules, MaaS becomes a marketplace optimized for whoever shouts loudest.

A workable interoperability stack

For a city or region planning MaaS in 2026, interoperability needs to be designed as a stack:

1. Data layer: shared formats for static and real-time service info (routes, stops, capacity signals).
2. Transaction layer: booking, payment, refunds, and entitlements across operators.
3. Identity and access layer: concessions, student/senior eligibility, residency-based benefits.
4. Operations layer: disruptions, incident workflows, customer support handoffs.

AI adds value across the stack, but only if the data rights and responsibilities are clear.

The governance rule that prevents “pilot purgatory”

Many MaaS pilots never graduate because nobody wants to own the hard parts: revenue allocation, service standards, and dispute resolution.

A simple governance move prevents this:

Define a single accountable MaaS service owner (usually the public authority), even if delivery is distributed across vendors and operators.

AI components—like demand forecasting or dynamic pricing—need a clear accountability chain, especially in the public sector.

What AI actually does inside a MaaS platform (beyond routing)

AI in MaaS is often described vaguely. Let’s get specific. The highest-value AI use cases cluster into four areas that directly support sustainable mobility.

1) Predictive operations: fewer “surprise failures”

The biggest trust-breaker in multimodal trips is uncertainty: “Will I make the transfer?”

AI models can predict:

• Bus bunching and headway instability
• Connection failure probability based on live speeds and dwell times
• Disruption propagation (one stalled tram line affecting feeder buses)

That lets MaaS present options like:

• “Fastest trip (high risk of missed connection)”
• “Slightly slower trip (high reliability)”

Reliability is what convinces car users to switch.

2) Demand shaping: nudges that reduce congestion and emissions

Sustainable mobility isn’t only about adding modes—it’s about shifting demand.

AI can support:

• Time-of-day incentives (reward off-peak travel)
• Capacity-aware routing (avoid overcrowded lines)
• Micro-incentives for first/last-mile by walking or bike

If a city’s policy goal is lower emissions, the MaaS experience must reflect it:

• Default sorting by “lowest emissions” for certain trip types
• Carbon labels that are consistent and understandable
• City-backed bundles (e.g., transit + bike share) priced to compete with parking

3) Accessibility and equity: the public-sector advantage

Private mobility platforms optimize for convenience and revenue. Public-sector MaaS should optimize for accessibility and inclusion.

AI helps when it’s used to remove friction:

• Step-free routing with live elevator status
• Safer walking paths at night based on lighting and foot traffic
• Tailored guidance for cognitive accessibility (simpler instructions, fewer transfers)

But this is also where AI risk is highest. If you’re using historical data that reflects unequal service patterns, models can reinforce them.

A practical safeguard:

• Require equity KPIs in MaaS performance reporting (coverage, wait times, affordability by neighborhood)
• Audit recommendation outputs, not just model accuracy

4) Fraud detection and entitlement protection

In public mobility systems, concessions and subsidies are sensitive.

AI can detect:

• Unusual ticket use patterns
• Multi-account abuse
• Suspicious refund activity

Done right, this protects budgets for those who need support. Done wrong, it can wrongly flag vulnerable groups.

Rule of thumb: use AI for triage, keep humans for final decisions, and document appeal processes.

A 2026 MaaS playbook for municipalities (what to do first)

A city that wants MaaS outcomes—not just a MaaS announcement—should sequence the work.

Step 1: Pick the outcomes and measure them

Define 3–5 measurable outcomes tied to policy. Examples:

• Reduce single-occupancy car trips into the city center by X%
• Increase public transport + active travel mode share by X%
• Cut average door-to-door commute variability by X minutes
• Improve accessibility trip completion rate (step-free journeys)

Then ensure the MaaS platform can actually measure those (data agreements, analytics).

Step 2: Start with two integration “hard wins”

Trying to integrate everything at once is how MaaS timelines explode.

Two hard wins that create visible value:

1. One account + one payment method across core modes (even if settlement is behind the scenes)
2. Disruption-aware multimodal rebooking (what happens when a tram is canceled?)

If residents see these working, they trust the system.

Step 3: Build a public-private partnership that isn’t vague

Partnerships only work when responsibilities are explicit.

A partnership charter should state:

• Data sharing scope, frequency, and quality targets
• Service-level expectations (uptime, latency, disruption reporting)
• Revenue allocation principles
• Customer support handoff rules
• Who can change fares and under what constraints

This is where many pilots fail quietly.

Step 4: Make “digital mobility” a capability, not a vendor dependency

The podcast points to “leveling up the digital side of mobility services.” That’s the right direction, and I’ll sharpen it:

If the authority can’t evaluate algorithms, data quality, and platform lock-in risk, it can’t govern MaaS.

Practical internal capabilities to build:

• A small data governance function (privacy, retention, access)
• Procurement literacy for AI (model transparency, monitoring, exit clauses)
• Product ownership (someone who represents the traveler, not the vendor)

Common MaaS pitfalls (and how to avoid them)

MaaS has enough hype history that it’s worth calling out the predictable traps.

“We’ll launch the app and behavior will change”

Behavior changes when the offer beats the default. That usually means bundles and pricing that compete with parking and fuel.

“Real-time data is optional”

It isn’t. Without real-time and disruption data, MaaS is a trip planner that apologizes a lot.

“AI will optimize everything automatically”

AI optimizes toward what you measure. If you measure only speed, you’ll get speed. If you measure reliability, emissions, and accessibility, you’ll get a system aligned to public value.

“One vendor can do it all”

Cities should avoid single points of failure—technical and contractual. Modular architecture and clear interfaces keep options open.

Where MaaS fits in the AI public sector and smart city agenda

MaaS is one of the most visible ways residents experience AI in the public sector. When it works, it feels like the city is coordinated. When it fails, it feels like the city is fragmented.

For the broader viedā pilsēta agenda, MaaS also creates a feedback loop:

• Mobility data informs infrastructure planning (bus lanes, bike networks)
• Policy changes can be tested via digital incentives
• Service gaps become measurable, not anecdotal

If you’re leading digital transformation in a municipality, MaaS is a chance to prove that data-driven decision-making can improve everyday life—not just dashboards.

Public-sector leaders should treat 2026 as the year to move from pilots to durable operating models: clear governance, interoperable foundations, and AI that’s accountable.

If your city is planning MaaS right now, the most useful next step is to write down one sentence: what promise are we making to travelers that they can’t get today? If that promise is real, the tech choices become much easier.