AUTONEST VCDesign principles · Handoff

What becomes more necessary as robots multiply

2026-09-12 · 11 min read

Human-sized robots are reportedly down to a few thousand dollars. Cutting machines and welding equipment get cheaper every year. A question naturally follows — "Doesn't that mean design work disappears?"
The answer runs the other way. The cheaper the hands get, the more the bottleneck becomes the side that tells those hands what to make.

deciding side shell 8T · head 10T blank 1250 × 2400 weld seam position · length people + design tools settled values executing side cutting · forming welding · assembly handling · inspection support machines · robots — getting cheaper
What gets cheaper is the right side. The left side stays

Machines execute; they do not decide

Before a robot arm can weld, someone has to have settled the values. What thickness the shell will be, which head type, what dimensions the plate gets cut to, where the weld seam runs. The machine takes those values and moves. It does not produce them.

So as equipment gets cheaper and more plentiful, what disappears is not design but the slack between design and the machine. It used to be that you handed over a drawing and parts appeared days later. Now, once the values are settled, it is cut that day. When the machine is the one waiting, the late side is always upstream.


What actually happened — equipment arrived and design work grew

A fabrication shop brought in an automated cutting machine for hundreds of millions of won to improve quality and lead time. More accurate and faster than hand cutting, so surely less work.

What happened was the opposite. What shrank was cutting time; what grew was everything before it.

Before the machineAfter the machine
Draw the 2D drawing.
Done.
Draw the 2D drawing.
→ Build a 3D model from it.
→ Hand over the model data.
→ Produce the machining data.
→ Transfer that data into the machine.
→ Set the plate on the machine and align it.

Bigger than the extra steps was that responsibility moved upstream. A slightly ambiguous drawing used to get caught on the floor. Now whatever was settled upstream is cut exactly as settled.

And the chance to check disappeared

This is the heart of it. When it was done by hand there was open marking. Before cutting, lines were actually scribed on the plate and a person looked at them. Someone put a tape measure from the centre to the nozzle, and if a hole position looked off, it stopped right there. Formally part of the process, in practice an inspection.

Machine cutting has no such step. Feed the data and it cuts straight away. The machine does not ask whether a value looks wrong. So one error and that is that. It is not just one plate lost — material lead time and the downstream processes slip with it.

"Just run a simulation," people say. We do. But checking it on the screen means opening every 3D sketch plane one by one, or dimensioning each point yourself. It does not replace looking with your eyes and measuring with your hands.

Here is the part that rarely gets discussed when automation comes up. When a machine takes a process, it takes the inspection attached to that process too. The things a passing pair of hands used to filter out for free simply stop happening. So where does that load go? All of it comes upstream.

Which is why the more automated a shop is, the more decisive it becomes whether the design data agrees with itself. The thickness in the report, the plate sizes in the material list and the geometry in the drawing have to be the same values. Discrepancies used to be found by somebody at the marking stage; now there is nobody to find them.

So the check was turned back into values

That is exactly what we did at that shop. We built a program to cut the added steps down. The idea was simple. 3D data is not visible to a person. So rather than struggling to make it visible, pull out precisely the values that need checking and put them in a table.

Look again at what open marking actually did — confirm the distance from centre to nozzle, confirm the hole positions match the drawing. That was not work that required seeing the plate; it was work that could be confirmed with numbers. Scribing lines on steel was not the essence. Looking at those numbers one more time was.

So those numbers were pulled out into a familiar table format. Three things changed at once.

The inspection the machine took away was not abolished — it was handed back as values a person can read.

AUTONEST VC is the continuation of that. The same idea, widened from one cutting process to the whole vessel design. Thickness, sizes, plate blanks, weld seams — all come out as values. A drawing comes out too, but the checking is done on the values. And whether those values contradict each other is the program's responsibility.


Automation is not technology; it is decomposing the work

One thing has to be said here. What decides whether a job moves to a machine is not the machine's capability. It is whether that job has been decomposed.

Split shop work into pieces and roughly four kinds appear.

PieceExampleWhere it goes
① Repetition Trying thickness candidates one by one against a chart; rewriting the drawing, material list and weld lengths every time a thickness changes The machine's place. A person does not produce a better answer, and tires into mistakes.
② Looks hard but
has rules
Settling a thickness that meets the conditions; finding the minimum thickness that holds under external pressure; judging whether reinforcement is needed This is the crux. Rules existed, nobody wrote them down, so it looked like experience.
③ Decomposable but
different per company
Title blocks, drawing-number schemes, line conventions — in-house drawing formats Not a technical problem but a problem of not becoming a product.
④ Stays because of
responsibility
Fixing design conditions; final review and approval The person's place. It stays not because a machine cannot but because a machine cannot be accountable.

Where the line between ② and ④ falls is the point of this article. That line is not moved by technology — it is moved by whether someone wrote the work down.

Settling thickness looked like ④ until it was written down. "That's experience," people said. Decomposed and written as rules, it turned out to be ②. Conversely, work that has rules which nobody has written down still behaves like ④ today.

So more machines is not bad news for this field. The faster machines take ①, the more sharply ② comes into view. Only after the repetition is cleared away can you see what was underneath it. And moving that into rules requires someone who knows the work. Equipment is everywhere; people who can decompose are the bottleneck.


The handoff is not built; it is taken out of what already exists

So what should actually be handed to the machine? There is a surprise here. By the time the calculation is finished, the data to hand over has already been produced. There is nothing extra to build.

Finding thickness settles the sizes, drawing settles the geometry, and producing the material list settles the plate dimensions per part. What was missing was one layer that sends it out.

So AUTONEST VC has an Export button that writes the result as a single data file. It is enabled separately for companies connecting factory equipment or in-house systems, and the file is saved under the report folder's export directory.

ExportedWhat it contains
GeometryInside diameter, length, head type, jacket presence and dimensions
ThicknessSettled thickness per part — shell, top and bottom heads, jacket, closure
Plate blanksWidth, length and quantity of plate needed per part
Design conditionsMaterial, pressure, temperature, vacuum, corrosion allowance
NozzlesSize, position, settled thickness, reinforcement verdict and pad specification
Flanges · volumeFlange dimensions and thickness, total and working volume

Whether it is wired straight into cutting equipment, loaded into a production system, or passed to nesting software, the starting point is the same single file.

Why we do not convert to machine formats

A request comes up often here. "Can you export directly in the format our equipment reads?"

Technically it is not hard. But we decided not to. There are many equipment makers, many models per maker, and versions keep rising per model. Start matching one and it becomes a structure where maintenance never ends. And in the end every user shares that cost.

One data file is the contract. Moving that contract onto your own equipment is the receiving side's work.

This is exactly the same judgment as not taking on each company's drawing format. In practice, cutting equipment rarely receives a design file directly. There is software in between handling layout and machining, and it already has its own standard. Our job is not to bend to that standard but to put out values complete and accurate enough to be converted.

The same program differs in price by an order of magnitude

We once asked what it would cost to have that program built externally. 50 million won.

We do not think the quote was excessive. Build it for one company and that company pays the whole development cost alone. Add the time the builder needs to learn the field from scratch. Something covering the whole of vessel design has a wider scope, so the figure would rise accordingly. Bespoke work simply costs that.

The same thing costs far less as a finished product not because the thing is different. It is because more people share the cost. Take what one place used to carry entirely and split it among many, and the output stays the same while each share falls.

What separates the price is not the quality of the output but how many places split the development cost.

So matching an in-house drawing format or a specific machine still carries bespoke pricing. That cost cannot be shared by its nature. When we said we do not export to equipment formats, it is not that we cannot — it is that the pricing moves into that structure the moment we do. What many can share stays a product; what only one place needs is handled in the way that suits it. That is where we draw the line.


What still stays with people

The row marked ④ above does not move across however good machines get. It is a matter of structure, not capability.


The work did not disappear. It moved

Look at that case again and in fact no work disappeared. Only its position changed.

Where it used to happenWhere it is decided now
Production control — which plates to take and how many, how to source material Settled at the design stage. Plate sizes and quantities have to be inside the cutting data already.
Quality — checking positions against the marking Replaced by whether the design data agrees with itself. After cutting there is nothing left to check.
Production — judging and correcting on the floor when a drawing is ambiguous Has to be settled as values upstream. A machine does not work out what was ambiguous.

Decisions that sat in three places have gathered into one. This does not mean departments went away. It means the moment of decision was pulled forward. Things that used to be settled in turn as the work passed through now have to be settled before the first cut.

And this shift accelerates

This matters. It does not end with one machine.

Every time one process is automated, the judgment and inspection attached to it move one layer forward. Automate cutting and cutting's share arrives; automate forming and forming's arrives; automate welding and welding's arrives. The cheaper equipment gets, the more processes are automated, and the more piles up upstream.

More machines does not mean less work. Work gathers at the front.

So the question shops will face is not "will design become unnecessary". It is "can design carry the weight that is gathering at the front". If you cannot triple the headcount, the front end has to produce more, accurately, in one pass. Not just thickness, but plate sizes, quantities and weld seams — all in a state where none of them contradicts the others.

The point where a tool becomes necessary is exactly here. Not to reduce people, but to put the work that has gathered at the front into a shape people can carry. All the more so at a scale headcount cannot hold.


To sum up

Machines getting cheaper makes execution cheap, not decisions cheap.

The common expectationWhat actually happens
Automate the equipment and design headcount falls. The work does not fall; it gathers at the front.
Decisions split across production control, quality and production have to be finished before the first cut.
Better machines will handle it all. A machine only takes work that has been decomposed and written down.
Work nobody wrote down stays where it is, however good the technology gets.
Automation brings quality with it. When it takes a process it takes the inspection attached to it too.
With nobody filtering at the marking stage, whether the upstream values agree becomes decisive.

So what has to be done now is neither fearing machines nor handing everything to them. It is splitting the work people used to do into pieces, writing what can be handed over as exact values, and keeping clearly what has to stay.

We started this at a shop like that. Where the steps had grown and the inspection had vanished, the job was to turn the values that needed checking back into a table. What we do now is the same — the subject has widened from one cutting process to the whole of vessel design.

And this shift will only get faster. Each time a process moves to a machine, the judgment attached to it comes one layer forward. What has to be prepared is not the machine but the values you hand it.

Take the design result as data.
From thickness to plate blanks, exported as a file with every calculation. See AUTONEST VC → How far it was verified →