Field Notes
Mid-market manufacturers are told they need a “smart factory,” usually by someone selling sensors. What they actually need is simpler and harder: for the plant floor and the business systems to stop being strangers. Everything else — the downtime surprises, the quality escapes, the spreadsheet schedule, the AI pilots that never leave the demo — is downstream of that one disconnect.
Two worlds, one company
Every manufacturer we walk into runs as two loosely-coupled worlds. The OT world — PLCs, HMIs, machines, the people who keep them running — operates in seconds and shifts. The IT world — ERP, quality systems, finance — operates in transactions and month-ends. Between them sits a human integration layer: supervisors re-keying production counts, quality techs transcribing measurements, a planner maintaining the real schedule in Excel because the system’s schedule stopped being believable years ago.
The fix is not a monolithic platform. It’s a thin, honest industrial data layer that lets each world see the other in near-real-time, deployed line by line, starting at the constraint. When downtime on the bottleneck line hits the planner’s screen in minutes instead of at tomorrow’s production meeting, behavior changes — and every subsequent investment gets easier to justify because its impact is finally measurable.
AI that survives the floor
Manufacturing has the highest-yield AI plays in the mid-market — predictive maintenance, vision QA, demand-to-schedule forecasting — and the highest kill rate for pilots. The difference is almost never the model; it’s whether the data spine exists and whether operators trust the tool. We sequence AI after instrumentation, target the asset or inspection point where pain is felt every day, and design override-and-feedback into everything. Trust compounds exactly like OEE does.
The Sherpa difference
This work needs a leader who is comfortable in steel-toes at 6 a.m. and in the board meeting at 2 p.m. — someone who has run plants’ technology before and knows which vendor promises evaporate on contact with a running line. That’s who takes the engagement: embedded a few days a week, measured on downtime, yield, and schedule trust, gone in about a year, leaving a plant that sees itself clearly.