Plants Digitised the Machines. The Shift Still Runs on Word of Mouth

Most plants can now tell you a motor's temperature to a tenth of a degree, the output of every line by the hour and the stock position of every part in the stores. Ask them why the second shift slowed a filler down on Tuesday night, or which operator knows the trick that stops a sealer jamming, and the answer is usually a person rather than a record. A decade of investment has digitised the machines and the back office. The layer in between, where people notice problems, work around them and tell the next crew, still runs mostly on talk.

The Most Important Data in the Plant Is Spoken

Relay, which sells voice and messaging tools for frontline workers, commissioned Dynata to survey 500 full-time US manufacturing professionals at organisations with at least 500 employees between August 1 and 14. The results, published on October 29 as The Factory Floor's Missing Data Layer, are perception research from a vendor with an interest in the answer, and are worth reading with that in mind. They are still striking.

Verbal communication was the most commonly cited channel for critical operational information, at 57%. Paper logs and whiteboards came next at 39%, with CMMS, ERP and ticketing platforms behind them at 36%. 86% said sensors and machine data alone are not enough to understand the shop floor, and 41% said their organisation has digitised its operational infrastructure while frontline work and communication still happen largely offline or informally.

The result is a record that does not match the shift it describes. Only 20% of respondents said their formal systems very accurately reflect what happened on a typical shift. 40% said only some of what frontline workers discuss informally ends up in a system, log or report, and a third said safety guidance is often communicated verbally rather than formally documented. The views also split by rank: 83% of executives were confident that a highly automated plant can spot problems on its own, against 40% of frontline respondents, a 43-point gap.

Every Hand-Off Is a Chance to Lose It

Information that lives in conversation has to be carried by someone. 82% of manufacturers in the Relay survey rely on employees to manually move information between teams, systems or shifts. 56% said documenting an issue or sending a status update takes at least four minutes with their current tools, 41% said communication breakdowns happen weekly, and 36% said a disruption caused by one typically takes at least an hour to resolve. The report itself adds that 78% lose production to communication breakdowns at least a few times a month.

Maintenance shows the same pattern from a different angle. Limble's survey of 686 US maintenance and operations professionals, published on October 7, found that 82% of organisations own and operate a CMMS but only 39% report consistent use across the entire team. 71% have asked a colleague for information their system already holds. The earlier Limble findings cited in the same release go further: 74% of organisations have at least one critical asset that only one member of staff knows how to fix, and 47% have lost knowledge they could not recover when someone left.

That last pair of numbers is where the informal layer stops being an inconvenience and becomes an operating risk. A machine that only one technician understands is a single point of failure in the same way as a part with one supplier. The difference is that the part shows up on a bill of materials and the technician's knowledge shows up nowhere until they are on holiday, off sick or gone.

New Crews and AI Both Need It Written Down

The pressure on that informal layer is about to rise. In Wipfli's Q4 Manufacturing Pulse Survey of 342 facilities, released on November 4, about two-thirds of manufacturers expect revenue growth in 2027 and 70% expect to hire, either to fill open positions or to support that growth. Every new hire needs what the experienced crew knows, and in Relay's research only 23% of organisations have a formal way to retain what departing workers know. Asked which challenges will matter most over the next five years, respondents put knowledge transfer from experienced workers first at 44%, narrowly ahead of AI and automation at 43%, and 75% said they are worried about losing experienced workers and their knowledge.

The same gap limits the AI projects many plants are now funding. In Limble's survey, 55% of maintenance teams are already using or piloting AI, and 92% agree it needs good, complete maintenance data to work effectively. A model trained on work orders that capture only part of what happened will learn only part of the plant. "Most maintenance teams aren't short on technology. They're short on time and information they can trust," said Limble chief executive Gary Specter. Relay co-founder and president Jim Mulcahy put the next step this way: "The next stage of digitization is preserving what the floor knows, turning it into insight."

What Operations Leaders Should Do Now

• Audit one shift hand-over: sit in on a change of shift and list what was said that never reached a log, work order or system. That list is the size of the gap at your plant, measured in your own terms.
• Make capture faster than talking: if recording an issue takes four minutes, crews will keep telling each other instead. Use voice notes, photos and short structured forms at the line so the record takes seconds, not minutes.
• Find your single-person assets: identify every critical machine that only one person knows how to fix, and pair that person with a second technician to document the fixes before the next absence.
• Fix CMMS adoption before adding AI: with only 39% of teams using their CMMS consistently, the first return on an AI budget may come from getting every technician to close work orders properly so the model has something true to learn from.
• Put safety guidance in writing: any instruction that changes how a task is done safely should reach a controlled document and a sign-off, not only a toolbox talk, so the next crew and the next contractor get the same message.
• Build knowledge transfer into hiring plans: if your plant is among the 70% planning to hire, budget overlap time for experienced staff to train newcomers and record what they teach, rather than leaving it to whoever happens to be on shift.

The sensors already tell a plant what its machines are doing. The harder data to collect is what its people know about those machines, and right now most of it leaves the building at the end of every shift. The plants that capture it will hire faster, recover from breakdowns sooner and give their AI something worth learning from.