Industrial software — no longer selling licenses?

What was the most common model for buying MES, PLM, or CAD in the past?
Buying a license, right?
How much per account, how much per module, how much for a perpetual license.
In essence, traditional industrial software sold:
"the right to use the software."

But with the arrival of AI, this logic may be changing.
Because software used to be a tool.
People used the software.
Now with the emergence of Agents, it may become:
software doing the work for people.

For example, engineers used to check equipment data, analyse faults, and search maintenance records themselves.
In the future, they might simply tell an Agent:
"Help me analyse why this equipment has been experiencing abnormal vibration recently."
The Agent would retrieve data, search historical records, call knowledge bases, run models, and finally provide a diagnosis.
This raises the question:
If software starts doing work for you, how should licenses be calculated?
By number of users? Or by module?
Neither seems quite right.

Thus, a new concept is becoming important: Token.
Not to say that all industrial software will be charged by Token in the future.
But rather, Token is becoming a new cost and measurement unit for AI software.
Software companies used to ask:
"How many users do you have?"
In the future, they may ask:
"How many tasks has your Agent completed?"

In the past, they sold an MES module.
In the future, it may become:
How much for one scheduling optimisation?
How much for one equipment diagnosis?
How much for one process plan generation?
Even more:
You pay for the value it creates for you.

At this point, the business model for industrial software may shift from:
License → Subscription fee → Token usage.
Simply put,
we used to buy software.
In the future, we may be buying:
the productivity that software creates for us.

01FDE: A New Role Connecting Site and Product

Once this happens, it's not just the pricing model that changes.
The organisational structure of industrial software companies will need to change too.
Why is FDE (Field Development Engineer) gaining increasing attention?
Because software companies used to work like this:
R&D built products → implementation delivered products to customers → after go-live, the project was basically done.
But the AI era is different.
When an Agent reaches the factory, that is where it truly begins.
It needs to understand enterprise processes, equipment, processes, customer data, and continuously adjust and optimise.
So future industrial software companies may need a very special type of person:
someone who understands software, AI, and the field.
This is why FDE is truly important.
They are not just implementation engineers.
They bring customer site problems back to the product; and bring product capabilities back to the customer site.

This forms a new closed loop:

Site → FDE → Agent → Data → Product Iteration → Back to Site.

02Is Industrial Software Still "Software" at All?

You will notice that
what industrial software companies are truly selling has quietly changed.
In the past, they sold: a software package.
Later, they sold: a platform.
Even later, they sold: a solution.
In the AI era, they may start selling:
a team of digital workers.

Therefore, what truly deserves attention in the future of industrial software is not just which software uses AI,
but a more fundamental question:
Is industrial software even still "software" anymore?
If software starts doing work on its own,
if enterprises start paying per task, per usage, or even per outcome,
then "License" may just be a name left over from the previous generation of business models.
The next generation of industrial software companies will compete not on "who has more features."
Rather, on:
who can truly bring AI into factories, truly do work for enterprises, and truly create value.

From License to Token,on the surface, it's just a pricing model change.In reality, it's the industrial software business that is changing.