On 15 October 2026 I gave a presentation at the SENTIATECH congress on prediction, monitoring and management of air quality, in Seville, for inBiot Monitoring SL. The presentation was part of the block From data to decision: artificial intelligence, digital twins and advanced tools for air quality management, chaired by Alex Borg (Lisbon Council). Each speaker had 15 minutes plus 5 for questions. There is no recording.
Content
The presentation covered Anne, an AI assistant I am building at inBiot to read air quality data from sensors and assess it against standards. A “harness” is the structure around the language model: the tools it can call, the steps it must follow, and the points where a person has to approve. Without that structure, a general chatbot given sensor data does the arithmetic itself, can invent thresholds, and can give a different answer each time.
In Anne, the language model does not calculate. Scores, statistics and charts come from separate software tools, so the same data always gives the same result, and every finding can be traced back to the tool and the data that produced it. A person reviews the findings before Anne suggests any action. I presented this design in June at HEALTHY INDOORS in Valladolid.
The new point in Seville is interoperability: other systems should be able to receive and understand what Anne produces. Without that, a finding about a classroom or an office stays inside the conversation where it was made. If the same finding is published in a shared standard format, a building management system, a city platform or a public dashboard can use it without a custom integration. For this I use FIWARE, an open-source platform used by many smart city projects. FIWARE stores the current state of things such as rooms, sensors or streets in a common format (NGSI) and describes them with shared schemas (Smart Data Models). Anne connects to FIWARE through an MCP server I published, which lets an AI assistant read and write data on a FIWARE platform. Adding it to Anne is one line of configuration. The assessment of one building then becomes information that a city can combine with outdoor air, traffic or weather data.
Publication
No separate publication; see Materials.