# MUNICIP-IA: from using AI tools to governing change

Author: Sofía García Morales
Original publication: 2026-03-18
Spanish original: https://estrategiabyaleph.substack.com/p/estrategia-129-la-inteligencia-artificial
English URL: https://elcontemplador.github.io/estrategia-english/essays/129/
Status: Published translation

This is a translation of the original Spanish essay published on 18 March 2026. Its claims, examples and forecasts retain that historical context.

English publication: 2026-09-29

> This week, estrategIA brings you a first-hand account of MUNICIP-IA, an event on artificial intelligence in the public sector. Sofía García Morales, a consultant at ALEPH Educational Institution, reports on its principal conclusions from Galapagar, in the Madrid region.

Last Friday, 13 March, Galapagar hosted MUNICIP-IA, a day devoted to AI in the public sector. More than a technology event, it was a very concrete conversation about management, administration and the future of institutions: what a council can do with AI today, how to start incorporating it intelligently, and why this conversation is no longer optional.

Before turning to the ideas and lessons, we should acknowledge Román Robles Valades, the Galapagar councillor responsible for Citizen Services and Digital Administration, for creating spaces like this and inviting ALEPH Educational Institution to collaborate. When debate about AI risks remaining at the level of headlines, such events help ground it in more important questions: how it is applied, what it is for and what it demands of public decision-makers.

[![MUNICIP-IA institutional closing session, 14:30–14:40.](https://elcontemplador.github.io/estrategia-english/assets/images/e47ba116-e573-4cff-8e06-ed6cd7b98ca2_1600x1200.png)](https://substackcdn.com/image/fetch/$s_!3xq-!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe47ba116-e573-4cff-8e06-ed6cd7b98ca2_1600x1200.png)

*The MUNICIP-IA stage during the institutional closing session. The screen displays “Institutional closing”, 14:30–14:40.*

**The day brought out three distinct but connected conversations: how AI can support municipal management; how an institution can begin to use it purposefully; and what political and social implications this change will have in the years ahead.**

## AI in municipal management

One of the day's first powerful ideas came from Juan Corro, head of Madrid City Council's autonomous IT agency: the machine does not “understand” as we do. It does not reason in the human sense. It works with patterns, sequences and probabilities. This explains why a system can be very useful and also confidently wrong. It also explains why AI “hallucinations” are not an oddity, but a natural consequence of a system that fills gaps in its knowledge with what it considers most probable.

That technical point matters for a practical reason. If AI operates on probability and context, using it is not enough. It must be trained, bounded and governed in the real circumstances of each institution. One of the event's most interesting examples was feeding systems with specific municipal rules to reduce mistakes and provide better guidance to residents. The task is not simply to acquire a tool, but to make it respond within the logic, limits and concrete needs of the council deploying it. Corro proposed an initial roadmap that any council could follow, even without a large city's size or Madrid's organisational resources: establish an AI group with clear leadership, identify the main management challenges, include at least one technical specialist and commit a reasonable initial investment. His presentation sets this out in more detail:

[![Juan Corro’s first steps for a council AI initiative.](https://elcontemplador.github.io/estrategia-english/assets/images/e7faf24b-595d-4be2-a407-106a8c3e6743_1536x1024.png)](https://substackcdn.com/image/fetch/$s_!yrQE!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe7faf24b-595d-4be2-a407-106a8c3e6743_1536x1024.png)

*“First steps”, translated from the slide:*

1. **Create an AI group.** Include a leader or responsible person with authority. Being able to get things done always requires… power.
2. **Identify the main challenges.** Then prioritise them according to capabilities and returns.
3. **Identify one technical specialist.** At least one person in the council must understand the subject. The speaker adds: “If you need help, send that person to me and we'll get things moving.”
4. **Put in a little money.** A little—not an enormous amount.

This shifts the question from “How do we get on board with AI?” to “What problem do we want to solve, and what minimum structure do we need to begin?”

The conversation about institutional productivity followed the same logic. Pablo Martín Diez, academic director of ALEPH Educational Institution, discussed the “10x consultant”: a professional who knows how to think, request, validate and guide AI can multiply their capacity to produce, analyse and respond in ways that seemed unimaginable just a few years ago. This is not limited to consulting or large organisations. Small teams and municipal election campaigns with fewer resources can also gain far greater capabilities if they use these tools well.

[![Roundtable with the 10x political consultant framework on screen.](https://elcontemplador.github.io/estrategia-english/assets/images/e1972711-ec97-4361-8646-3259db48bf41_990x915.png)](https://substackcdn.com/image/fetch/$s_!G5s-!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe1972711-ec97-4361-8646-3259db48bf41_990x915.png)

*The roundtable screen presents “The 10x political consultant: doing the work of many with fewer resources thanks to AI”. Its seven visible areas are automated analysis and processing of large data volumes; strategic creativity and narrative design; personalisation and message segmentation; multimedia content generation; knowledge management; no-code technology development; and AI agents. It contrasts the traditional approach with a 10x approach using AI. The smaller explanatory text is less legible in this photograph and is not reconstructed here.*

David Povedano Alonso, deputy secretary of Galapagar Council, likewise highlighted the sharp reduction in time brought by AI: processes that once consumed months can now be planned and set in motion in just 30 days. Ignacio Azorín, Director-General for Digital Strategy in the Madrid regional government, grounded this idea in a specific example: [SIDERAL](https://www.comunidad.madrid/gobierno/digitalizacion/digitalizacion-entidades-locales), the service catalogue through which the region supports digitalisation in local authorities with fewer than 20,000 residents. Again, the conclusion was clear: adopting these tools is not an option reserved for large councils.

[![Madrid regional presentation of SIDERAL for local digitalisation.](https://elcontemplador.github.io/estrategia-english/assets/images/de86770f-147d-4128-a356-47abcb776987_1536x1024.png)](https://substackcdn.com/image/fetch/$s_!ffLM!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fde86770f-147d-4128-a356-47abcb776987_1536x1024.png)

*The slide reads “Move forward with SIDERAL — A more digital municipality”, from the Madrid regional government's Department of Digitalisation. Its printed address ends digitalizacion-entidades-locales-locales; the article's linked address above has a different ending. Both are preserved as found, without silently resolving the discrepancy.*

## There is no single AI: different levels, uses and moments

Another important contribution came from AI communicator Xavier Mitjana, who organised uses of artificial intelligence into eight levels. His central point was clear: AI does not serve one uniform purpose, and organisations need not all begin in the same place. All eight levels can add value, but the key is understanding where each institution stands and what help it actually needs.

Sometimes a chatbot will be enough for specific queries or basic generic tasks. In other cases, targeted or advanced assistants will be more useful, accelerating recurring tasks, connecting to tools or working with more dynamic information. When a process needs greater autonomy, agents enter the picture, handling recurring processes or more complex tasks with fewer detailed instructions.

What matters is neither remaining at the simplest level nor racing towards the most advanced. It is understanding the complexity each process requires, the human oversight it needs and the institutional return it can generate. In public administration, progress will often come from an accumulation of well-chosen, well-governed small improvements rather than one great disruption.

[![Xavier Mitjana’s level 7 slide on general-purpose agents.](https://elcontemplador.github.io/estrategia-english/assets/images/d3aa2adf-f5be-4750-b1be-d1d87bf02bc3_1536x1024.png)](https://substackcdn.com/image/fetch/$s_!A90j!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd3aa2adf-f5be-4750-b1be-d1d87bf02bc3_1536x1024.png)

*The photographed slide shows only “Level 7 — General-purpose agents”, which “solve complex and general processes without detailed instructions”. Examples: Claude Cowork, Anthropic, January 2026; and Manus, 2025. The reference describes Manus as a general-purpose autonomous agent executing complex tasks in a sandbox with internet access and prints arxiv.org/abs/2305.20240. That is the slide's reference, not a newly checked attribution. The La Salle Technova mark is visible. The other seven levels are not shown in this photograph.*

## Digital sovereignty and a new division of work

The roundtable, with Professor Ángel de Antonio, David Povedano Alonso and Pablo Martín Diez, brought forward a discussion extending far beyond administrative efficiency: digital sovereignty. Europe remains heavily dependent on technologies, models and infrastructure developed in the United States and China. The issue is therefore not merely how to use AI, but where it is developed, who controls it and how much autonomy our institutions actually possess. In that context, developing European AI ceases to be a rhetorical aspiration and becomes a strategic question.

The conversation did not end there. Beyond who leads the technological race, another fundamental debate emerged: AI's own impact on work, education and how our societies will be governed.

If AI multiplies some professionals' capabilities, reduces the need for others and reorganises entire sectors, the question extends beyond incorporating it into public administration. It also becomes a matter of responding politically to a society whose assumptions about work, effort, social mobility and progress may be shaken. That brings ideas previously regarded as theoretical, such as universal basic income, into the discussion.

This is no small matter. We must not simply ask what purpose today's educational and employment models will serve amid growing automation. We must recognise that all of this will become citizen demands and new expectations about protection, opportunity and the future—and therefore political discourse, strategy and decisions.

In other words, AI will require institutions to work differently. It will also require politics to offer different answers.

## Judgement will make the real difference

MUNICIP-IA left a fairly clear message: AI is no longer confined to experiments, curiosity or conversations about the future. It has entered the local public agenda in full. It concerns how councils can serve, manage and decide better, and make better use of their capabilities. Increasingly, it also belongs to a broader, more uncomfortable discussion: how to govern a society that will change alongside it.

The councils that adapt best will not necessarily be those that talk most about AI, or even those that buy the most tools. They will be those that integrate it purposefully, with clear priorities, prudence where needed and ambition where warranted.

The question is no longer whether councils will work with artificial intelligence. It is which will learn to do so with sound judgement first.

[Sofía García Morales](http://www.linkedin.com/in/sofiagarciam)

Consultant, ALEPH Educational Institution