Operator onboarding and tribal knowledge
Operator onboarding that captures tribal knowledge
There is one person on every team who knows everything. The pump that trips on Mondays, the exact sequence to bring the furnace back after a trip, the thing that goes wrong every spring. That person is also the one everybody calls at 2am. Digel is built on a simple bet: if the software brings answers to you instead of waiting to be asked, institutional knowledge moves out of the heads and into the system.
Why every plant has a 2am phone call person
When software waits to be asked, the only model that works is the one where a handful of senior operators carry the plant in their heads. They know which asset is quirky, which shift left the furnace hot, which feedstock Line 3 hates, and the workaround nobody wrote down because everybody just knew. They are also the ones working the late shift, taking the 2am call, and training every new hire.
It scales until they leave, retire, or have a bad day. Onboarding a new operator is supposed to be a fixed-length process. In practice it is the slow transfer of knowledge that nobody ever managed to write down. Every shift handover is a small audit of what the last shift noticed, and most of what they noticed does not get captured.
The answer is not another training module. It is software that watches the same thing the senior operator watches, writes down what they would write down, and answers new-operator questions with the context the senior operator would have pulled up.
How Digel captures and shares tribal knowledge
Six concrete mechanisms that move knowledge from heads into the system, in production today at our pilot customers:
Asset-linked notes
A rich-text editor with '@' mentions. Operators type '@Pump 4 has been running hot since the firmware update last Tuesday' and the note becomes part of that asset's graph context. Hover cards show mentions inline.
@digel anywhere
The chat agent is woven into every surface of the product. You can '@digel' in any comment, on any issue, in any note, and it will answer using the full context of the thing you are looking at and everything it connects to.
Triage decisions as signal
Every accepted, dismissed, or argued-with triage item feeds back into the graph. Disagreements teach the agent what your team cares about. Comments become durable context.
Shift handovers that carry
Shift reports are generated and published automatically, attach to the assets they touch, and carry forward. The next shift reads a summary, not a stack of PDFs.
Plain language across the plant
Operators ask in the language they prefer. The agent responds in the same language. Our pilots run with Norwegian manufacturers and the chat threads are mostly in Norwegian.
Onboarding from day one
A new hire on day one talks to the same colleague the senior operator does. They get the same answers, the same context, and the same heads-up. Everyone on every shift starts with the plant's collective memory.
What changes for a new operator on day one
Asking in plain language
The new hire types "what should I watch on Line 3 during this run" and gets a real answer, with the open issues pulled in, the recent notes surfaced, and the procedure linked. No tool-switching, no asking around.
Scanning a machine
A QR code on every machine lands them on its current state, its open issues, and the notes the senior operators have left. The native mobile app works offline, so it works in the middle of the plant.
Arguing with the agent
When they disagree with a finding, they push back in chat and the agent re-investigates. The disagreement is not friction. It is how the system learns what the team cares about.
Knowledge management that feeds the rest of the product
Notes and decisions do not sit in a knowledge base that nobody reads. They live in the same context graph as your sensor data, your work orders, and your documentation. Every investigation, every maintenance decision, and every shift report draws on them.
See AI for manufacturing for the broader picture, AI maintenance management for the CMMS surface, and automated reporting for how shift handovers carry the knowledge forward.