AI integration in the Canary Islands.

We don't sell chatbots or "hype AI". We work on top of the data the installation already produces — consumption history, occupancy, HVAC, process sensors, security video — and apply models that deliver measurable value: anticipating energy demand, detecting anomalies before failure, conditioning rooms based on real patterns, in-line quality control or automating first-level support. AI integrates on top of the existing KNX/BMS or SCADA, it doesn't replace them.

No. AI is built as a layer on top of what already exists: we read variables via KNX-IP, BACnet/IP, Modbus TCP or OPC UA, store them, model them and send setpoints or alerts back to the existing system. The installation keeps working even if the AI layer drops. That "non-intrusion" is deliberate — a KNX/BMS system running for 5-10 years is the foundation, not something to throw out.

Only if you decide so. All time-series models (consumption forecasting, anomaly detection, adaptive HVAC) run on the edge on a NAS or local mini-server in the building or plant. LLMs (Claude, GPT-4, Gemini) are used via API only when the case justifies it, and always with anonymised or conversational data. In sensitive installations (hospitals, institutional buildings, industry with IP) we work full-edge with no outbound traffic.

Three with clear ROI: 1) Energy demand prediction by occupancy + weather, to match it with the PV curve and the hourly PVPC price; 2) Adaptive HVAC based on real usage patterns of room or hall, instead of fixed schedules — typically 15-25% savings in HVAC; 3) Detection of water and energy consumption anomalies per circuit (leaks, faulty equipment, wrong staff settings). For hotel chains, also automatic CMMS ticketing with incident prioritisation.

Predictive maintenance from PLC variables (temperature, vibration, current, pressure) — we train models on the plant's real history and warn when the pattern deviates before failure. Process anomaly detection to avoid defective batches. Setpoint optimisation in lines with multiple degrees of freedom (flows, pressures, temperatures). Integration with SCADA and with the client's ERP/MES via OPC UA or REST.

A pilot always includes the same: audit of available data, definition of measurable use case, trained and validated model, integration with the existing system and a tracking dashboard. The scope and cost vary significantly with the complexity of the use case, the starting data and the integration needed. We deliver a custom study with no obligation — define the use case and we come back with a fixed proposal and a clear success metric. If it works, we scale with an operations contract.

Yes, but with caveats. Standard KNX IP gateways allow voice control from Alexa, Google Home or Apple HomeKit. For advanced cases (complex natural-language queries, building context, multi-language, integration with the maintenance manual or ERP) we build our own assistants on LLMs (Claude, GPT-4, Gemini) with access to KNX/BMS state. Voice is one more layer — useful when it adds value, dispensable when it doesn't.

Office in Santa Cruz de Tenerife. AI projects, by their nature, are the easiest to operate remotely: once deployed, the model and dashboard live on the building or plant's own server, and from Tenerife we monitor, tune and retrain without travel. For work that does require physical presence (new wiring, additional sensors) we cover Gran Canaria, Lanzarote, Fuerteventura, La Palma and La Gomera.