What you’ll learn from this episode
- “Don’t start by buying a new system.” That’s Robert Grelowski’s first piece of advice for operations directors when it comes to digitizing a production plant. 80% of the cost and time of a rollout goes into cleaning up data the company has had for years.
- There’s nothing wrong with using Excel as long as it stays a helper tool. The trouble starts when the spreadsheet becomes “an ERP in its own right” and the company decides on numbers nobody verifies.
- Automation doesn’t push the operator out, it changes their role into supervising the system. In aerospace, every stage of a product’s life — from production through engine overhauls — is signed off by a person anyway.
- Before you buy a machine or a system, measure: data can show you the investment isn’t needed at all. At JPB, a KEYPROD vibration sensor showed that queues at the measuring machines only formed at certain hours. Instead of buying another machine, spreading out the schedule was enough.
- You retain knowledge from someone’s first day at work, not when they hand in their notice. JPB does it with five mechanisms: an assigned mentor for three months, cross-training on machines, paper shift notebooks, internal promotions, and showing people where their work ends up.
- A failed rollout also looks like this: the system sits unused and the work goes back to spreadsheets kept on the side. That’s how projects end when nobody asked production, quality, the toolroom, and the warehouse how the system should work in their daily job.
- In aerospace manufacturing, every part stays traceable even after 30 years in service. The quality regime turns out to be a ready-made data infrastructure that digitalization can simply stand on.
Watch with English dubbing or subtitles
Where do you start digitizing production — with a system or with data?
Asked what he’d advise operations directors who want to digitize production, Grelowski doesn’t hesitate: “Whatever you do, don’t start by buying a new system. Because that’s hitting a wall, and feeding any new system the same data won’t improve anything, won’t change practically anything.” The new system gets the same “version 15, final, final” files, and only their author knows what’s inside them. Sometimes, five years on, even the author doesn’t. This isn’t Robert’s first ERP rollout, and he estimates that as much as 80% of a rollout’s cost is standardizing, cleaning, and getting data into a state you can actually work with. Without that, he says, you can’t move efficiently or pick a system that fits what you need.
JPB Système has spent more than a year preparing to choose a new ERP. The project is mapped out across several years, and the company created a Digitalization Manager role with a seven-person team whose main job is organizing and standardizing data. Only with that homework done does the company want to pick a system tailored to itself, because — as Grelowski puts it — “it’s very easy to give in to the salespeople and buy a system that’s going to cure the company.” Then it’s the vendor telling the company what it needs, and the system fits the offer rather than the plant’s processes. So a company first has to know what it needs, and what question to ask its data to get the answer it’s actually after.
When does Excel stop helping and start doing damage?
Excel starts doing damage the moment it stops being a helper tool, something to calculate on the side or pull a working summary from. The problem shows up when the spreadsheet gets promoted to the official version of events: “sometimes Excel becomes an ERP in its own right […] And that’s terrifying, because we’re working on numbers that aren’t real,” Grelowski says. Not real, because every department keeps its own copy of the file and decisions land on figures nobody verifies. In that role, a spreadsheet can be the most expensive system in the company — except nobody counts what it actually costs in the budget.
The opposite pole is JPB’s own plant. “There’s one source of truth and that’s our intranet […] which holds the current version of a given procedure or drawing.” Every work order has to be done to the documentation that’s in the system right now, not to a printout from a month ago. The second part of the same discipline is about people. Rather than pulling “a million different kinds of data” out of systems, JPB trains people to ask the data specific questions so they get a specific answer.
Will automation push operators out of production?
“I don’t see a risk of the human being displaced, at least at the stage we’re at now,” Grelowski answers. Automation changes the operator’s role, but it doesn’t eliminate it. You can see it best on JPB’s automated line in France: six machines linked by a robot produce six different parts, a second robot handles quality control, and the system runs statistical process control on the results. When a measurement drifts toward the tolerance limit, the system sends the machine a correction on its own. Human intervention is kept to a minimum, but it doesn’t disappear. As Grelowski puts it, “we’re scaling their skills up to the level of someone who programs and supervises the system, rather than strictly an operator.” The factory of the future isn’t the one without people.
The line where automation stops is a hard one, and it’s drawn by responsibility, not by technology. Every stage of a product’s life — production, assembly, engine overhauls — is signed off by a person: “That stamp matters enormously. The price of that stamp and the responsibility that sits on it are huge, and we’re not going to eliminate that.”
Grelowski talks about this as someone who has already lived through one revolution of this kind. He came into the trade when complex engine components were measured with a dividing head, dial indicators, and trigonometry, and he rolled out the first coordinate measuring machines himself, watching metrologists with fifty years of experience be afraid of them. Today’s fear of artificial intelligence is, to him, the same fear pointed at something new. And it carries the same lesson: the advantage goes to whoever understands the process underneath the tool.
How do you monitor old machines without replacing your equipment?
It turns out you don’t have to replace machines or interfere with their controls — you just have to read their vibrations. At JPB that’s KEYPROD, a device built by a company inside the JPB group: the box attaches to the machine housing, plugs into power, and that’s it. “It’s Shazam, but for production,” Grelowski says. “It doesn’t matter whether it’s a machine from 40 years ago, a huge machining center, or a coffee maker. It shows you what’s going on inside that machine right now.” The operator categorizes a stoppage with one click (tool change, first-article measurement, inspection), and the manager sees a live picture of production from their phone: the biggest losses, stoppages grouped together, places to react fast. For companies running decades-old equipment, that’s the answer to connecting old machines to new systems.
The best proof of what this approach is worth is a decision it stopped them from making. Production staff asked for a new coordinate measuring machine, because queues were forming at the existing ones and people couldn’t keep up. The sensors showed the queues only formed at certain hours, and spreading out the schedule was enough. The device’s software is also moving toward predictive maintenance — maintenance that prevents failures rather than reacting to them. The system detects anomalies in vibration and, once trained, recognizes what the machine is currently producing.
How do you retain the knowledge of experienced employees?
Robert answers with JPB’s own example. The company worked out five mechanisms that hold knowledge at different levels and stages: from a new hire’s first months, through the daily work of shifts, to promotion paths.
First, mentoring: a new employee gets one assigned trainer for the whole three-month onboarding period. Second, cross-training: everyone teaches everyone else to run a different machine, so anyone can cover for anyone else at any time. Third, shift handover notebooks, on paper, as Grelowski admits: “it’s a bit […] going against digitalization, but it’s a simple way of writing down the problems you ran into on a given shift and solved.” Fourth, internal promotions instead of hiring from outside — quality control and process engineering roles go to people from production, who bring with them the knowledge of how a part is made and what affects it. Fifth, meaning: when he trains people, Grelowski shows them where their work ends up, because “each of these parts is human lives, not just a piece of metal.”
The cross-training mechanism came out of the hardest moment. During the pandemic people worked three days a week at full pay, and the company put the freed-up time into having employees pass experience to each other. Grelowski sums up the math behind it in one sentence: “hiring an employee and not training them is far more expensive than hiring an employee, training them, and even losing them.”
Why does a production system end up unused after rollout?
Grelowski has seen rollouts launched with great fanfare, after which the new system sat there and the work went back to spreadsheets anyway. He lists three conditions for success: “people have to be convinced, people have to be trained, people should be asked what they think a system like this should look like, because they’re going to be its everyday users.” And he’s precise about how far that “asked” reaches: “If nobody asks production, nobody asks quality, nobody asks the toolroom or the warehouse guy what he needs and what data he wants to put in” — then the rollout simply won’t work.
For an operations leader, that flips a familiar frame. The employee resistance you’re afraid of before a rollout is the consequence of leaving people out of the design. Asking the toolroom and the warehouse what they need and what they want to get out of the system is the cheapest element of the entire project, and it decides whether the rest of the spend starts working at all.
What does quality mean in aerospace manufacturing — and what follows for other kinds of production?
In aerospace, quality “is in every smallest aspect of the company. From top management to the furthest corners and nooks of the plant,” Grelowski says. What surprises people from outside the industry most is how far that goes in practice, starting with part traceability: “Let’s say we have a quality problem on an aircraft, and the aircraft has 30 years of service behind it. We’re able to trace every part at every stage of production, who made it, to which process, to which revision, where the material was ordered from.” Documentation can go back 50 years. With engine parts, quality is a matter of life and death.
A quality regime isn’t a cost imposed from outside, it’s data discipline. A plant that can reconstruct the history of any part 30 years later already has the information infrastructure in place. Audit requirements turn out to be the foundation of digitalization rather than an obstacle to it.
What makes information circulate around a plant?
At JPB, two kinds of meetings do that job — the loose kind and the hard kind, each in a different role. The meeting without an agenda takes half an hour and happens every day before 7:00 a.m. in the company kitchen: everyone from administration shows up at the Coffee Corner, along with anyone who feels like walking over. “There’s nothing to prepare for it. There’s no output from it in the sense of minutes.” People simply agree among themselves who can help whom with today’s tasks.
That morning meeting “breaks down silos, it really does,” Grelowski says. Someone from finance gets interested in a quality problem and looks at it from a different angle, and sometimes a problem the team had been sitting on for days turns out to be simple to solve.
Grelowski adds that a loose coffee doesn’t replace discipline, though. “We also have serious, strictly hard, heavy meetings with real output, with minutes, with specific data.” Both serve the same purpose: information has to circulate between people — freely day to day, rigorously in reporting meetings. And the kitchen meeting is something you can start tomorrow, and it costs nothing.
About the guest
Robert Grelowski — Operations Manager, JPB Système Polska. Since 2019 he has been responsible for the development and operational performance of JPB Système’s Polish plant. It’s a French-owned family business whose self-locking fasteners go into aircraft engines from Safran, Rolls-Royce, and GE, among others. Before that he spent 11 years at Pratt & Whitney Kalisz, moving through every level of quality from cleaning V-blocks to managing the Special Measurements Lab. In the corporate rotation program he worked at PZL Mielec on the Black Hawk program, in procurement on consolidating the supplier base for the Airbus A320neo, and at Hamilton Sundstrand on the thrust reverser assembly line for the Boeing 787. He is a member of the Sectoral Council for Competences in Aviation.
