Joined-Up Smart Cities: Why AI Is the Missing Link

A “smart city” used to mean sensors, dashboards, and a handful of pilot projects that looked great in a press release. In 2025, that definition feels outdated. The real test isn’t whether a city can deploy technology—it’s whether it can coordinate decisions across transport, housing, energy, public safety, and social services without tripping over its own silos.

That’s why the idea raised in the SmartCitiesWorld podcast episode on re-evaluating smart cities for a joined-up future matters. The argument is basically this: smart city strategy can’t be “tech first” anymore. It has to be tech + environment + people, at the same time. I agree—and I’d go one step further: without AI in the public sector, most “joined-up” ambitions stay stuck at the workshop-and-slide-deck stage.

This post is part of our “Mākslīgais intelekts publiskajā sektorā un viedajās pilsētās” series, where we focus on practical ways AI improves e-pārvaldes pakalpojumi, infrastructure management, traffic analysis, and data-driven decision-making. Here, we’ll translate the podcast’s themes into a clear approach cities can actually implement.

The new definition of a smart city: coordination beats gadgets

A smart city in 2025 is best understood as a city that can reliably coordinate services using shared data, shared goals, and shared accountability. Not a city with the most sensors.

The podcast points to a shifting taxonomy: you can’t treat technology, the environment, and people as separate tracks. City systems are tightly coupled:

• A traffic policy changes emissions.
• Emissions policies change logistics and business activity.
• Business activity changes housing pressures.
• Housing pressures change service demand (schools, healthcare, benefits).

Why the “project mindset” keeps failing

Most cities still operate with a project mindset: one department runs a pilot, procures a platform, builds a dashboard, and calls it transformation. The reality is that this approach fails for predictable reasons:

1. Data fragmentation: each system stores data differently, with different identifiers, and different access rules.
2. Conflicting incentives: departments optimize their own KPIs even when it harms citywide outcomes.
3. Procurement lock-in: new tools get bought to solve narrow problems, then can’t integrate.
4. Change fatigue: staff see “digital transformation” as extra work, not better work.

Here’s the stance I’ll defend: a smart city strategy that can’t specify how decisions will be coordinated across departments isn’t a strategy—it’s a shopping list.

AI as the “joined-up layer” for city operations

AI helps smart cities when it’s used as a coordination layer—connecting data, predicting outcomes, and guiding decisions across functions. It’s not about replacing city staff; it’s about reducing the friction that makes cross-department work painfully slow.

What “joined-up” looks like in practice

Joined-up operations show up in small, unglamorous moments:

• A roadworks plan automatically checks impacts on bus reliability, emergency response times, and school commute peaks.
• A housing inspection backlog is triaged based on risk signals, not first-in-first-out.
• Citizen support is routed using intent detection so residents don’t bounce between call queues.

AI makes these moments possible because it can:

• Unify signals from many systems (even if the city can’t fully modernize them yet).
• Forecast second-order impacts (e.g., “this change reduces congestion but increases emissions in this corridor”).
• Prioritize work when resources are limited (which they always are).

The three AI capabilities cities should focus on (not 30)

Cities get distracted by shiny demos. I’ve found it’s more effective to focus on three capabilities that map directly to public sector outcomes:

1. Entity resolution and master data: linking people, places, assets, and cases across systems.
2. Decision intelligence: models that recommend actions with constraints (budgets, staffing, legal limits).
3. Natural language AI for service delivery: summarization, routing, document handling, and multilingual support.

If your smart city roadmap doesn’t include these, you’ll keep building isolated tools that can’t talk to each other.

Breaking down silos: start with data, then governance, then automation

Everyone says “break down silos.” Fewer cities explain how. The workable sequence is:

1. Data interoperability (minimum viable integration)
2. Governance and accountability (who can decide what, using which data)
3. Automation and optimization (only after 1 and 2)

Minimum viable interoperability (MVI): the faster alternative to big-bang integration

Cities often stall because they think integration must be perfect. It doesn’t. A better approach is minimum viable interoperability:

• Pick 2–3 priority outcomes (for example: winter road safety, permit turnaround time, or bus punctuality).
• Identify the 5–10 datasets needed.
• Create shared identifiers (assets, locations, cases).
• Build APIs or event feeds only for what’s necessary.

This is where AI can help early: use machine learning to match inconsistent records (addresses, asset IDs, organization names) and surface confidence scores for human review.

Governance isn’t paperwork—it’s how you prevent “model chaos”

Once data starts flowing, the next risk is uncontrolled AI: different departments deploying models that conflict, degrade over time, or quietly introduce bias.

A practical AI governance setup for municipalities includes:

• Model registry: what models exist, what they do, who owns them.
• Data access rules: purpose limitation and role-based controls.
• Monitoring: drift, error rates, and complaints.
• Human override: clear rules for when staff can ignore the recommendation.

One-liner worth remembering: If nobody is accountable for an AI decision, the city is accountable for the fallout.

Upskilling without excluding: the public sector balancing act

The podcast raises a tricky reality: cities need to upskill workers for a digital future without making inclusivity worse. That’s not theoretical—public sector employers are often the largest local employers, and the way they modernize affects the whole labor market.

What upskilling should actually include in 2025–2026

Upskilling can’t just mean “learn to use the new platform.” Cities need role-based skill paths:

• Frontline service staff: prompt hygiene, verification habits, and safe handling of sensitive data.
• Supervisors: how to spot automation errors, manage exceptions, and measure service quality.
• Analysts and planners: causal thinking, model evaluation, and scenario planning.
• Procurement and legal: how to specify AI requirements, auditability, and data protection in contracts.

If you only train your “digital team,” you create a two-speed organization where a small group moves fast and everyone else becomes a bottleneck.

Digital inclusivity: AI can widen the gap—or close it

AI in e-pārvalde can improve access, but only if it’s designed for it.

Ways AI can widen the gap:

• Chatbots that assume high literacy or perfect Latvian/English
• Services that require smartphone-only authentication
• Automated decisions with no clear appeal route

Ways AI can close the gap:

• Multilingual support and plain-language rewriting
• Voice-based interfaces for residents who struggle with forms
• Proactive outreach (e.g., reminders for expiring benefits or permits)

A practical rule: every AI-enabled service should have a “non-digital equivalent” and a human escalation path—not because digital is bad, but because cities serve everyone.

Where AI delivers measurable value in smart city operations

The fastest wins come from areas with high volume, repeatable decisions, and clear service metrics. Here are examples that fit the “joined-up” framing.

AI for traffic flow analysis and incident response

Traffic isn’t just a transport problem; it touches air quality, emergency response, and economic activity. AI can:

• Predict congestion and suggest signal timing adjustments
• Detect incidents from fused inputs (cameras, sensors, citizen reports)
• Improve winter maintenance routing using forecasts and road priority

The joined-up move is to connect these outputs to other teams: public communications, schools, public safety, and maintenance crews.

AI for infrastructure management: from reactive to planned

Cities spend too much money reacting late: pipes burst, roads fail, assets get replaced too early or too late. AI supports:

• Predictive maintenance (asset failure probability)
• Risk-based inspection scheduling
• Work order clustering to reduce travel time

You don’t need perfection—just a consistent pipeline from asset data to prioritization to field execution.

AI in e-governance: faster service without lowering standards

AI can reduce administrative load while keeping legal responsibility with humans:

• Document triage and summarization for case workers
• Automated completeness checks for permit applications
• Intelligent routing to the right specialist team

A “joined-up” city connects these improvements to customer experience: fewer handoffs, fewer repeated questions, and transparent status updates.

A practical roadmap: 90 days to start, 12 months to scale

Cities often ask the wrong planning question: “Which AI tool should we buy?” Better question: “Which city outcome should we improve across departments?”

First 90 days: prove coordination, not just technology

• Choose one outcome that crosses silos (e.g., roadworks coordination, permit backlog, winter safety).
• Map the end-to-end process and identify delays and handoffs.
• Stand up minimum viable interoperability for the needed datasets.
• Deploy one AI component that reduces friction (routing, matching records, summarization).
• Define metrics: turnaround time, cost per case, complaints, on-time performance.

Months 4–12: standardize what worked

• Create a reusable data integration pattern (not one-off scripts).
• Establish AI governance (registry, monitoring, escalation routes).
• Expand training beyond specialists—make it role-based.
• Scale to 2–3 additional outcomes with the same playbook.

This is the core message of joined-up smart city strategy: repeatable coordination beats scattered innovation.

People also ask: smart cities, AI, and public sector reality checks

Can AI break down silos by itself?

No. AI can expose inconsistencies and recommend actions, but silos are also political and organizational. You still need shared goals, shared metrics, and a decision forum that can resolve conflicts.

Is AI in the public sector mainly about automation?

Automation is the side effect. The main value is decision quality at scale—faster triage, better prioritization, and clearer trade-offs across services.

How do you avoid “AI theatre” in smart city programs?

Tie every model to a city outcome and a measurable operational metric. If you can’t measure it, it’s probably a demo.

Where this leaves smart city strategy in 2025

The podcast’s point about re-evaluating smart cities is right on target: the future isn’t a collection of smart projects. It’s a city that can act coherently across technology, environment, and people.

For this topic series—Mākslīgais intelekts publiskajā sektorā un viedajās pilsētās—the implication is straightforward: AI becomes valuable when it strengthens e-pārvaldes pakalpojumi and city operations in a way residents can feel. Shorter queues. Fewer repeat forms. Faster fixes. Clearer decisions.

If you’re building a smart city roadmap for 2026, the question to sit with is this: what’s the first cross-department outcome where AI can act as your joined-up layer—and who will own it when the pilot ends?