Artificial intelligence and local authorities: useful AI is AI that knows its place

What if the value of AI in major cities were primarily determined by access to services that already exist?

"By 2025, 9 out of 10 major cities already had AI projects underway." (source: Data Publica 2025 Barometer / La Poste Group). However, the majority of these projects come up against the same obstacle: not a lack of data, but the difficulty in accessing it.

Major cities have never had access to so many digital services. Transport, urban planning, waste management, water, energy, housing, public participation, administrative procedures… Each area of public policy has gradually developed its own tools, platforms and databases.

This transformation has significantly improved the services provided to residents. But it has also created a paradox: whilst information has never been more readily available, it has sometimes never been harder to find.

The challenge in the coming years may not be to create more digital services. It will be to make those that already exist more accessible, easier to understand and simpler to use. This is precisely where artificial intelligence can play a role, provided it remains in its proper place.

The real problem: access to information, not a lack of it

When a resident contacts their local authority, it is not usually because the information does not exist. It is because they do not know where to look for it.

Here are a few examples that would be familiar to any local public service:

• “Can I build an extension on my plot?”

• “What’s the best way to get to the airport tomorrow morning?”

• “What financial assistance am I eligible for to renovate my home?”

• “Who should I report this road problem to?”

In each of these cases, the answer already exists: in a regulation, a database, a decision or an existing portal. The problem is not a lack of information. The problem is accessing that information.

Conversational AI therefore acts as a layer of simplification between the complexity of the public sector and the user’s immediate need. It does not create anything new. It reveals what was already there.

An opportunity for both staff and residents

The subject is often discussed from the perspective of the relationship with citizens. However, one of the most significant impacts of AI in local authorities concerns the staff themselves.

In many departments, a significant proportion of time is spent on essential but unrewarding tasks: searching for information in a fragmented database, directing a user to the right person, rephrasing technical content, and producing summaries of regulatory documents.

Recent research, notably a study carried out in the City of Lyon, estimates that "25 per cent of tasks carried out by local government staff can be performed, in whole or in part, by generative AI” (source: INET/CNFPT, April 2024). This figure does not mean that staff are replaceable. It means that they could devote more time to situations that genuinely require their expertise.

A RAG (Retrieval Augmented Generation) architecture, for example, makes it possible to connect an AI model to internal knowledge bases (decisions, procedures, business repositories) without altering the existing information system. This enables agents to access the information they need instantly. Users receive reliable answers, grounded in the local authority’s actual data.

Well-designed AI acts as an extension of metropolitan services, not as a substitute for them.

25 per cent of a local authority’s business processes could be handled, at least in part, by AI. This is not a threat to staff; it frees up time to focus on what really matters.

The three practical ways in which AI contributes to a major city

Experience shows that the most useful applications are based on three simple functions.

Making services more accessible through a chatbot

Major cities cater to an extremely diverse range of people: older people, students, tourists, new arrivals, professionals and voluntary organisations. Not all of them are equally familiar with digital tools or administrative terminology.

A conversational assistant allows anyone to phrase their enquiry in their own words and receive a clear response, without needing to be familiar with the local authority’s organisational structure. This is precisely what BeTomorrow has achieved with Ally, a conversational AI solution designed for transport operators and local authorities, which simplifies access to passenger information, benefiting several hundred thousand users.

MOBILITY AND AGENT-BASED AI: WHERE CAN VALUE BE CREATED?

• Where does AI really stand in 2026?
• Conversational AI as a catalyst for mobility services
• Agent-based AI for automating business processes
• Designing a mobile AI project in 5 steps, from scoping to optimisation
• Meeting the requirements of AOMs and operators
• ALLY: from passenger information to travel assistant

Download the white paper

Making services more accessible: the chatbot as the first point of contact

Public services cannot be open round the clock. Yet residents’ needs do not stop at 5 pm.

A well-configured public service chatbot provides an initial response to thousands of recurring enquiries: opening hours, civil registry procedures, information on roadworks, and enrolment in after-school care. It thus improves service availability without automatically increasing the resources required. In Issy-les-Moulineaux, IssyGPT handled nearly 20,000 citizen enquiries in less than a year (source: issy.com, January 2026), significantly reducing the pressure on in-person and telephone support services.

Making services more useful through the use of spatial data

This is undoubtedly the most strategic contribution. Cities have a wealth of data, but this is often scattered across various departments and systems.

AI makes it possible to combine this information and present it in a form that can be used directly: by a resident seeking an answer, by a staff member compiling a file, or by a decision-maker steering public policy.

Where exactly should AI be implemented within a local authority?

The key question now is one of discernment: identifying the functions where AI delivers real value.

Not all functions should be entrusted to a language model. Reference data must remain reliable and under control. Regulatory calculations must remain deterministic. Business rules must continue to be guided by human experts.

AI adds value where:

• the request is ambiguous and needs to be interpreted;

• the content is complex and needs to be rephrased;

• several sources need to be summarised quickly;

• directing the user to the right service reduces their uncertainty;

• 24-hour availability makes a real difference to the service experience.

In other words: AI does not replace the foundations of public service. It improves access to it.

AI in local authorities: a public service issue, not a technological one

The debate surrounding artificial intelligence is often presented as a technological issue. For major cities, however, it is first and foremost a matter of public service design.

How can we make it easier for residents to get answers? How can we help staff quickly find the information they need? How can we ensure consistency across digital ecosystems that have been built up in successive layers?

AI can address these questions, provided it is deployed pragmatically, without making unrealistic promises, and with a detailed understanding of the real constraints faced by public organisations: data governance, GDPR compliance, cybersecurity, service continuity and staff engagement.

It is this approach that BeTomorrow’s Cities Division advocates and implements, from strategic vision through to industrialisation, for regions wishing to transform AI into a driver of sustainable public services.

From access to action: the next step for AI in local authorities

The three previous contributions share the same underlying principle: AI helps to find the right information and present it clearly. This is the first step, and it is enough to transform the service experience. A second step is now emerging, in which AI also helps to achieve results.

In practical terms, an assistant capable of understanding a request can now guide the user through the entire process: pre-filling an application form using information already held by the local authority, checking which documents are required, highlighting any missing information and preparing the next step. The same development applies to staff members, for whom the assistant classifies an incoming request, directs it to the correct department and triggers the corresponding action in the relevant system, whilst ensuring that the decision-maker remains involved in situations where this is warranted.

This marks the transition from conversational AI, which responds to queries, to agent-based AI, which coordinates multiple stages and systems to see a task through to completion. For a city whose services have been built up in successive layers, this ability to orchestrate processes is particularly valuable: it links systems designed separately to the user’s actual needs, without requiring a complete overhaul.

This increasing level of autonomy makes the issue of judgement even more crucial. The more an AI system acts, the more important it is to define precisely what it can initiate on its own, what requires human validation, and what must remain in the hands of experts and business rules. This is the prerequisite for an agent-based AI to find its place in a public service: genuine autonomy for tasks where this is appropriate, underpinned by clear governance where the stakes demand it.

Are you supporting the digital transformation of a city or a transport operator?

BeTomorrow’s Cities Division supports local authorities and transport organising bodies, from strategic vision through to industrialisation.