Unlocking city intelligence with LLM: Conversation as data analysis

Table of Contents

chatbot llm en iris360

Let’s be honest: data from a city, a factory, or a field are often quite unfriendly.

Until now, if you wanted to know if the air quality sensor in the main square was correctly configured, you had two options: either you were a database expert or you had to dive into an instruction manual longer than a Russian novel. At Libelium, we have spent 20 years seeing how data remains locked in charts that no one has time to interpret.

So we asked ourselves a question: What if, instead of looking at graphs on the platform, we simply ask it a question?

Goodbye to Hallucinations (and the "User Manual")

The new iris360 assistant is not a generic chatbot that “hallucinates” responses. It operates under a RAG (Retrieval-Augmented Generation) architecture. What does this mean? That the AI anchors every word to curated official documentation and real data.

In human language: we have given the AI the platform manual and the real data from your sensors.

  • If you ask it: “How do I configure an alert if NO2 exceeds x levels?”, the AI doesn’t search the internet; it searches its own documentation and guides you step-by-step.
  • You no longer search the tool (or the user manual you don’t know where you saved); the tool explains itself.

A Data Scientist Who Never Sleeps

When we have all the information on the internet at our fingertips, one thing becomes clear: what is important is not the information available, but knowing how to ask the right questions. This is why generative AIs in chatbot format have gained popularity.

The value of data lies in the ability to ask the right questions. The iris360 assistant acts as an expert in data analysis:

  • Advanced identification: Detects correlations, anomalies, and trends in real-time.
  • Context awareness: By “seeing” the same control panel as the user, the AI offers situated and relevant answers to what is being viewed.
  • Predictive models: Uses historical data to offer predictive models, helping managers anticipate environmental or operational challenges.

And the best part: it adapts its language. If you are the manager who needs a quick summary, it gives you a headline. If you are the technician who needs to get down to the details, it gives you the precise data. The LLM integrated into iris360 adapts its language, offering explanatory information for managers and deep technical data for researchers.

ALIA chatbot in iris360

Technological Sovereignty: iris360 as a key piece of the European Data Infrastructure

Libelium is positioned as a strategic actor in the European data infrastructure, ensuring that the data from Union countries does not leave their borders. We live in turbulent times, which is why sovereignty is in vogue. Companies no longer want the data from a Spanish city or a German car factory or an Italian olive grove to end up feeding a black box abroad that holds the key to access to the data itself.

We have connected iris360 with the core of ALIA, Spain’s AI infrastructure, and the power of the Barcelona Supercomputing Center (BSC). This means two things:

  1. Total control: Sovereign data hosted on local servers.
  2. Real context: You can speak to it in Spanish, Galician, Basque, or Catalan. It understands local nuances because it has been trained with models of each co-official language.

At Libelium, we have designed the AI infrastructure to guarantee total control of information and comply with European Union standards (AI Act Ready). Therefore, this chat integrated into iris360 is based on:

  • Self-hosted infrastructure: Unlike solutions based on external servers, iris360 is designed to be self-hosted, ensuring that entities maintain custody and privacy of their citizens’ data.
  • Traceability: Every AI response allows auditing the original source (data or document), guaranteeing complete factual accuracy.
  • Ethics: Active blocking of irrelevant or harmful queries before generating a response.
ALIA Spanish AI infraestructure
The integration of ALIA in iris360 was presented during MWC 2026 by the Minister of DIgitalisation of Spain Government.

Privacy by Default

We know that AI is a bit scary when we talk about public data. That is why iris360 comes with a “standard filter”: The system includes filters that eliminate any personally identifiable information (PII) before the information reaches the AI model, ensuring privacy by design. It deletes any personal data before the model processes it.

Furthermore, being a self-hosted technology, you are the key guardian. The data stays at home.

By integrating a chatbot, it’s not that we have made the platform smarter; it’s that we have made it more human. Because the challenges of mobility in cities, food sovereignty, or the efficiency of our industry should not be slowed down by not knowing how to correctly interpret the data.

Behind the Change.

Beyond the Challenge.