We watch the reviewer. Not the sensor.
Sovel is the cross-system decision-evidence layer for regulated industrial maintenance. We read the work-order narrative, the shift log, and the operator chat, then tell a named, qualified reviewer which closures don't yet have the evidence to hold up under audit. We help capture the missing reasoning, bind a reviewer's sign-off to it, and show the gap was closed.
What Sovel is
- CMMS-adjacent — the first wedge, not the whole category. We sit on top of Maximo, SAP PM, AiM, AMOS, and the spreadsheet your planner actually uses. We never replace them. CMMS is where many non-routine decisions first surface; the category is broader — the decision-evidence layer across maintenance history, controls data, procedures, compliance records, handoffs, and the reasoning that walks out the door at retirement.
- Detection-first. We surface closures that look fine in the CMMS but have no evidence anchor: repeat failures, prerequisite-readiness gaps, shift-handover holes, operator-overridden setpoints nobody captured. Find what's missing before you capture what people say.
- Knowledge-governance-first. Every governed decision record carries provenance, an evidence anchor, and an audit trail a PSM, NERC, or Part 11 auditor can read. We hold our records to a higher bar than any regulator asks.
- Reviewer-operated. The AI drafts; a qualified reviewer approves. No autonomous writes to governed truth, ever. The reviewer has the final word, and that architecture is the moat.
What Sovel is not
- Not a CMMS replacement.
- Not a generic chatbot or search-over-PDFs product.
- Not a connected-worker tool.
- Not an autonomous-agent vendor promising closed-loop control.
Ben Klement
Background spans operations research, industrial systems engineering, and hands-on time with reviewers in PSM-, NERC-, and FDA-regulated facilities. The product is shaped by a single rule: nothing ships if it would let an unsupervised model write a regulated record.
Reach out: ben@sovel.io · book 20 minutes on Cal.com
Where we are
Pre-revenue, and honest about it. The reviewer console and the detection engine are in active build with design partners; the first paid engagements open as the evaluation harness graduates from synthetic to customer data. If you run a regulated industrial facility and you've ever closed a work order you knew lacked the supporting evidence, we want to talk.