Industry Reflection | Where Is the Agent on the Shop Floor?

WAIC 2026 came to an end in Shanghai this week.

Over four days, the event gathered more than 1,100 companies and showcased 300+ global product launches. The word “Agent” became the hottest topic across the exhibition floor.

Some companies even built a “super factory of AI agents” inside the venue, demonstrating hundreds of role-based digital experts designed for different tasks.

But just 100 kilometers away, inside many machining workshops, the reality remains unchanged:

A night-shift operator is still holding a handwritten tool request form, standing in front of the tool room, searching for a φ12 end mill.

Two worlds.
One era.

01 | Why Are There Still No Agents on the Shop Floor?

It is not because factories do not need them.

Actually, factories need them the most.

A new part arrives:

Which tool should be selected?
How much tool lifespan remains?
Why did the production line stop again?

The answers already exist — but they are scattered everywhere:

• Paper drawings

• Experienced operators’ memories

• Excel spreadsheets

• Machine alarm records

Even the smartest Agent cannot make decisions without access to reliable industrial data.

Without data, an Agent remains only a smart conversation in a chatbot window.

02 | Before an Agent Goes to Work, Data Must Go to Work First

The Agents showcased at conferences are built on computing power and massive datasets.

But the Agent inside a factory must grow from something different:

Real, structured, and continuously generated production data.

This is not just our observation.

At WAIC 2026, embodied intelligence became one of the biggest themes. The number of participating companies in this field increased dramatically compared with last year.

Yet many industry leaders shared the same challenge:

“The biggest bottleneck is real-world data. Laboratory data cannot cover the complexity of actual operating conditions.”

The entire AI industry is facing a data challenge.

Shop-floor Agents face the same reality.

Before an Agent can truly work in manufacturing, three foundations must be built:

1. Data must be standardized

A single cutting tool may contain more than 140 data fields.

A single BOM may involve over 120 parameters.

Only when industrial languages are unified can machines truly understand them.

2. Data must be collected — not manually filled afterward

Historical records created after the fact cannot generate trustworthy decisions.

3. Data must flow through the entire process

Wherever the data chain breaks, the Agent loses visibility.

03 | Before Talking About Agents, Build the Soil First

This is exactly what Knowhy has been building over the years:

A complete cutting management ecosystem that creates the foundation for industrial intelligence.

🗄️ Smart Tool Box | Managing Tool Flow

Every tool movement is recorded and traceable.

⏱️ Tool Lifespan Management | Managing Tool Usage

Actual machining workload and remaining tool lifespan are continuously monitored.

📡 SCB Smart-CNC-Box | Understanding Machine Conditions

Machine status and wear trends are transformed into readable data streams.

🧠 Tool Management Operating System | Building the Data Foundation

With more than 200,000 material records and 20,000+ workpiece records managed within the system, every cutting operation generates valuable industrial data.

The real purpose of this ecosystem is simple:

Turn every action on the shop floor into data first.

04 | An Agent Is Not Installed. It Grows.

At WAIC 2026, hundreds of Agents were introduced.

But manufacturing Agents will not simply walk from a conference stage into a factory.

They must grow from the industrial data foundation.

After visiting hundreds of exhibitors, industry observers pointed out a clear trend:

AI is moving beyond demonstrations. The real competition is shifting toward real-world scenarios and data infrastructure.

The direction is becoming clear:

An Agent’s ability is not measured by the volume of a product launch.

It is measured by the depth of its industrial data foundation.

The Knowhy Process Agent follows exactly this path.

Built on top of the cutting management ecosystem, it learns to:

• Understand engineering drawings

• Interpret machining tolerances

• Match tools and processes

• Check inventory availability

• Identify abnormal conditions

Turning questions like:

“Which tool should we use for this new part?”

from an experienced machinist’s personal knowledge into a systematic, data-driven answer.

The soil has been prepared.

Now, let intelligence grow.

WAIC 2026 is over.

But one question is worth bringing back to every company promoting AI Agents:

Can this Agent survive on my shop floor?

Knowhy Insight

Before every cutting operation can become intelligent,every cutting action must first become data.