Will software products disappear?
AI is driving down the marginal cost of software development. The future may no longer be "one standard product sold to 100 factories," but 100 factories with 100 needs, each getting AI-generated software. Yet products won't disappear—only their form changes. What gets reused are "industrial building blocks": order models, equipment models, quality rules, scheduling algorithms. AI assembles them o
Why did industrial software have to be made into "products" in the past?
Because software development was too expensive.
A software company would spend years building an MES.
Then find a way to sell it to 100 or 1,000 factories.
A software product, in essence, is about spreading the cost of one development effort across many customers.
So the software industry in the past had a very important logic:
Standardisation enables scale.
01AI Is Changing the Marginal Cost of Software
But AI is changing one thing:
The marginal cost of software development may be getting lower and lower.
Previously, a feature took a programmer days to develop.
Now AI may complete it in hours.
Previously, when a factory raised a customisation requirement, the software company had to assess, develop, test, and deploy it.
In the future, you may simply tell AI:
"Modify this feature according to our factory's process."
If this truly becomes easier and easier,
then the question arises:
Why should we still make 100 factories use the same software?
Perhaps another model will emerge in the future:
not one software company developing one standard software and selling it to 100 factories.
But rather:
100 factories raise 100 kinds of requirements, and AI generates 100 applications that belong to each of them.
02Software Products Won't Disappear, but Their Form Will Change
At this point, many people might say:
So will software products disappear in the future?
No.It is just that the form of the "product" will change.
The software product of the past was a complete application:
MES, ERP, WMS, equipment management…
In the future, what can truly be reused may not be this complete application, but the individual capabilities within it:
order models, equipment models, quality rules, scheduling algorithms, process knowledge, data interfaces, industry knowledge…
plus the AI that can invoke these capabilities.
They are like blocks of "industrial building blocks."
Whatever the factory truly needs,
AI reassembles these blocks
to generate a set of software that fits the factory's actual business.
So the future may not be:
"Which MES are you buying?"
But rather:
"What capabilities does your factory need?"
And then AI is responsible for assembling these capabilities into software.
At this point, you will notice a very interesting shift:
In the past, the most important capability of a software company was developing a piece of software.
In the future, it may become:
accumulating a set of industrial capabilities that can be continuously combined and invoked.
03What Will Truly Endure in the Long Term?
So software products will not disappear, but "fixed software products" may become fewer and fewer.
What truly endures in the long term
may be the knowledge, models, rules, capabilities, and interfaces behind the software.
And the software itself
will increasingly resemble a real-time generated result.
If things truly reach this point,
then the competition among next-generation industrial software companies will no longer be about who has more code, more pages, or more features.
But rather:
who owns more high-quality industrial building blocks.
But there is an even bigger question here:
If all this software can be generated on-site by AI, then who defines what the factory needs?
This question may be the true core of the entire industrial software revolution.