Skip to content

Expertise that stays in the plant

Nobody is sat down for an interview.

Every plant has a few people who hold the exceptions: the manifold that needs re-torquing after the first hour, the reset switch nobody wrote down. Most tools propose to capture that in sessions, before those people leave. Norlan captures it the only time it reliably appears — while they are fixing the machine — and shows, by model, where the plant still depends on one person.

Who closes what

By model · last 12 months

8

people who close

15

models with history

3

models with one expert

ModelWalkerBermúdezCostaWeberRuizNowak

Labeller LB-200

1 unit · 33 closures

67677

Slicer SL-7

11 units · 9 closures

23121

Checkweigher CW-500

8 units · 7 closures

22111

Metal detector MD-3000 1 person

7 units · 6 closures

6

Rotary filler RF-880

2 units · 2 closures

11

Vision camera VC-12 1 person

1 unit · 1 closures

1

Tray packer PK-40

3 units · 2 closures

11

Case labeller LB-40 1 person

1 unit · 1 closures

1

How it is captured

Out of the incident, not out of a session

An expert asked to explain what they know gives the manual back. The same expert, in the thread at 03:00 with a torch in one hand, says the one thing that is not in it: the half turn on the front bolts, the relay to check before the sensor. Norlan is there for the second version. When the thread closes, what was worth keeping is extracted from what was actually said, attached to the machine and to its model, and proposed as a field note for a tap to confirm.

  • Extracted at closure from the thread, the voice note and the photo
  • Attached to the unit and to every unit of the model
  • Confirmed with a tap by the person who found it, never typed up afterwards

What a machine thread holds

Field note, kept from the closure

Metal detector MD-3000 · applies to 8 units

Reusable
0:16

After a format change, check belt tension before touching the calibration. The K4 relay works loose on this model and the fault comes back within the hour if it is not reseated.

From the thread that closed itT. Walker · Maintenance

Recorded in one language. Read by each technician in their own.

With their name on it

Attributed to the person who found it, and kept when they go

The AI proposes the wording, never the facts, and the technician stays the author. The note carries their name, and so does the guide that grows out of it. When that person goes on leave, changes contractor or retires, the note does not: the next technician who asks the machine gets it cited as an earlier intervention, with who found it and when.

  • The technician is the author; Norlan does the typing
  • Cited by the assistant as an earlier intervention, by name
  • A guide that grows out of it keeps the attribution through every version

How a field note becomes a guide

Rotary filler RF-880 · Berlin

Note · R. Ortega · 12 Mar

Note
0:09

“Manifold needs re-torquing after the first hour at full speed. Every time, not in the manual. Half a turn on the two front bolts.”

Attached to the assetCited by the assistant

Does not open a thread.Turn into an incident

The new technician

They ask the machine, and the senior answers without being phoned

A new hire does not learn a plant by shadowing whoever is free. They learn it by asking, and most of what they ask has already been answered on that machine. Norlan gives them the senior's fix, cited, at the moment they need it — and says plainly when the record holds nothing, so they know when to ask a person instead of guessing.

  • The manual page and the earlier intervention, in one answer
  • The scope follows the question: this unit, this line, every unit of the model
  • When the record has nothing, it says so instead of inventing

How the assistant answers, and from where

Metal detector MD-3000

Berlin · 03:14

This unit
False rejects after the format change again. Where do I start?

Norlan

Check belt tension before the calibration: on this model it slackens after a format change. The K4 relay works loose too, and the fault comes back within the hour if it is not reseated.

Manual · p. 42Berlin · 29 JulField note · T. Walker

Is that written down anywhere?

Norlan

Yes — it became a guide in August, after the third time. Two steps, with the tension range.

Guide v2 · 8 units

Measured, not assumed

Answers that come, more and more, from the plant's own people

The proof that expertise stayed is not a folder of interviews. It is the share of answers the assistant grounds in the team's own interventions rather than in the manual, and how it grows as the record does. Next to it, the models that still depend on one person to close their incidents: that is the list to act on before the sick leave, not after.

  • The share of answers grounded in the team's own work, month by month
  • The models with a single expert, marked while there is time
  • The questions asked outside office hours: the calls that did not happen

The whole panel, for maintenance managers

Where the answers come from

453 questions · last 12 months

66 % of answers cite the plant’s own interventions — from 18 % in the first month.

AugSepOctNovDecJanFebMarAprMayJunJul
  • The plant’s own interventions
  • The manufacturer’s documentation
  • Had to ask outside

When questions are asked

39 % outside office hours — the calls that did not happen

M
T
W
T
F
S
S
0006121824

Where the line is today

Norlan does not run capture sessions.

No interviews, no workshops, no conversion of the old binders into training modules. Documents are loaded as they are, and what is in people's heads enters when they use it on a machine, not before. That has a consequence worth knowing: a machine nobody has touched in a year has a thin record, and the assistant will say so rather than pretend.

How a pilot starts, and what we ask for

Questions about keeping expertise

Do we have to interview our senior technicians?

No, and we would advise against building a plan on it. An expert asked in a meeting room what they know gives back a tidier version of the manual. What matters is what they say at the machine when it is broken, and that is what the thread keeps. The only deliberate act is confirming a field note with a tap, by the person who found it.

The person who knows the line leaves next month. What does Norlan capture by then?

What they do on the equipment in that month, and nothing they do not. If they close incidents in Norlan, dictate notes and confirm what is proposed, that stays with their name and the assistant cites it afterwards. If they spend the month on equipment that does not fail, the record stays thin, and Norlan will show which models still depend on them so the plant can plan around it. We would rather say that than promise a download of someone's head.

Does it capture what is in the binders and the old PDFs?

It loads them as they are, proposes which model each belongs to, and cites the page when it answers. It does not rewrite them into training content or work instructions. What the documents lack — the exception on unit three, the reset nobody wrote down — comes from the threads, not from a conversion project.

How do we know it is working?

Three numbers on the manager's panel, agreed before the pilot starts: which models depend on a single expert, before and after; the share of the assistant's answers grounded in the team's own interventions rather than in the manual; and the interruptions to a senior technician that did not happen. None of them needs a survey.

Do our technicians have to document more than they do today?

Less, in the moment that matters. The thread where the incident is being solved is the record, and a voice note counts as much as typing. Writing up a guide is a deliberate act, but it happens once, by whoever solved it, and the AI does the tidying.

Book a demo

Bring the name of the one person everybody phones. We will show what the plant already knows of theirs.

We start with the operation — which failures repeat, who gets called when something breaks, what the CMMS does not hold — and then show how Norlan would work on it.

What we usually cover

  • How an incident flows today, from detection to fix
  • Where the conversation actually happens: the systems, or calls and WhatsApp
  • What takes longer — the diagnosis, or getting the right people and information together
  • How much of what a technician knows is written down, and what happens when they are away
  • Which tools are really in play: CMMS, ERP, mobile, paper
  • What a pilot would measure, and how

Pick a time that suits the plant. Bring one failure that keeps coming back.

The call is with the people who build the product, with no qualification call first. Norlan is pre-general-availability and shaped with maintenance teams in pharma and food processing.