In the future, who will be operating the software?
Four trillion is just the surface. What truly deserves attention is that industrial software is evolving from "providing information to people" to "getting tasks done on behalf of people." In the future, it may no longer be about humans operating MES, ERP, and equipment systems — instead, humans will direct agents, and the agents will operate the software, call on systems, and execute tasks. The n
What is really changing in the 4-trillion-yuan industrial agent market?
I increasingly feel that industrial software is about to be used in a completely different way.
In the past, people operated the software.
In the future, people may command agents, and agents will operate the software.
Once this happens, the industrial software business will fundamentally change.
Recently, the 2026 China Industrial Agent Market White Paper released at the 2026 Shanghai Frost & Sullivan Global Growth Summit showed that by 2030, China's industrial intelligence market size is expected to exceed 4 trillion yuan.
But I believe what truly deserves attention is not that 4 trillion figure.
It is a shift:
Industrial AI is moving from "assisting people" to "doing the work for people".
01 The Past: Software Delivered Information to People
In the past, our approach to industrial digitalisation followed this logic:
MES handled production management, ERP handled operations, SCADA handled monitoring, and equipment systems handled data collection.
When a problem occurred, the system told you:
"An alarm has been triggered here."
And then?
People still had to check data, find root causes, look through historical records, and decide what to do.
So, in essence, most industrial software in the past was doing one thing:
delivering information to people.
02 The Future: Agents Complete Tasks for People
But industrial agents aim to go one step further.
For example, when equipment has an anomaly:
It doesn't just tell you "equipment anomaly".
Instead, it retrieves historical data, checks maintenance records, and analyses in combination with process parameters and knowledge bases:
What is the root cause? What should be done? Are there similar cases?
It may even go further by invoking MES, EAM, or equipment systems to complete subsequent actions.
What you will find is that the real change here is not just adding an AI chat interface.
It is that:
Software is shifting from "providing information" to "completing tasks".
This is a truly significant change.
Because in the past, enterprises bought software for its features.
For example, "I want an MES", "I want an equipment management system", "I want a quality management system".
In the future, what enterprises increasingly care about will not be:
"What features does your software have?"
but rather:
"Can you help me get this job done?"
For example:
Help me optimise today's production scheduling.
Help me find the root cause of the yield drop in this batch.
Help me identify which equipment is most likely to fail.
Help me resolve this anomaly.
03 The Rules of Competition Have Changed
At this point, the competitive logic of industrial software has shifted.
The past was about:
who has more features, a better interface, and more comprehensive modules.
The future may be about:
whose agent understands industry better and can truly get things done.
So I believe what truly deserves attention about industrial agents is not that they have created yet another new AI sector.
Rather, they may become the interaction and execution model for the next generation of industrial software.
In the past, it was:
People → Software → Systems.
In the future, it may become:
People → Agents → Software → Equipment.
This is what truly matters behind the "4 trillion" figure.
Because once agents truly enter the production floor, industrial software will no longer be just a "tool" in the factory.
It may begin to become:
a "digital employee" that truly participates in production, management, and decision-making.
And this may be where the next round of competition in industrial software truly begins.