
Industrial technicians work with their hands full: gloves on, instruments in hand, often on ladders or in tight spaces. Yet the information they need is spread across manuals, maintenance systems, telemetry screens and parts databases, and every lookup means stopping work. Readings get written down later, if at all, and specialist handovers are often incomplete. FieldSense AI is a voice-first operations assistant that removes that friction while keeping the technician in charge. A technician says "compressor AC-104 has stopped, the panel shows fault E27," and FieldSense retrieves the asset's details, live and baseline telemetry, maintenance history, and the relevant section of the approved manual for that exact model. It explains the evidence, proposes the next check, and records spoken measurements against a work order it creates automatically. Safety is built in through a five-tier risk model, from observation to dangerous condition. When a reading crosses a safe limit, the agent stops routine guidance, raises the risk level, and prepares a structured specialist handover with the asset, symptoms, readings, history, manual sources and actions taken. It never treats verbal confirmation as proof a site is safe and never performs physical actions itself. FieldSense runs on the AssemblyAI Voice Agent API for real-time speech-to-text, turn-taking, voice activity detection and spoken responses, and technicians can interrupt mid-sentence. Ten strictly validated JSON-Schema tools query Neon PostgreSQL, run model-filtered retrieval on Chroma Cloud, and manage work orders. Every tool call is audited and streamed to a live dashboard, so supervisors can follow the reasoning as it happens.
30 Sep 2026