5S without traffic boundaries is theatre — visual management for AI

5S, dashboards, machine data, MES — you have all of it, and yet automation is not delivering the promised productivity. The reason rarely lies in the AI model. It lies on the floor. If the shop floor has no clear zones, traffic rules and markings, the system is analysing a process that is not stable. That is why visual management for automation is not an embellishment of the implementation — it is its precondition.

What Hannover Messe 2026 really showed

At Hannover Messe 2026, the analyst firm IDC put forward a thesis that overturns the usual thinking about the advantage of AI in manufacturing. The key sentence reads: "the data gold mine is not just in the machine data" — the source of value does not lie solely in machine data. The most valuable context is created where the worker and the process meet: in the operator's decisions, in the way people move around the shop floor, in where and how material is set down.

This is an important correction. For years, investment went into the digital layer — sensors, telemetry, integrations. Yet the quality of the conclusions AI draws depends on the depth of the operational context, not on the quality of the model alone. A rich stream of data from a machine describes the machine. It does not describe the chaos around it.

In practice, this means shifting the centre of gravity. Before you add another model or dashboard, it is worth asking whether the process around the machine is orderly enough for a repeatable pattern to be read from it at all. If not, every additional digital layer measures variability that would be easier to remove than to analyse.

Why 5S without traffic boundaries is just decoration

5S organises the workstation: sort, set in order, shine, standardise, sustain. The problem is that conventional 5S usually ends with tool outlines and a label on the rack, and does not reach the layer that determines the repeatability of the entire flow — traffic boundaries. When there are no marked-out routes, marked set-down zones and unambiguous rules on who gives way to whom, "order" exists only in the audit photo.

For a person, this lack is inconvenient. For an autonomous system and for analytics, it is simply noise. Visual management for automation closes what workstation-level 5S does not cover: it translates the rules of the process into signals that are visible from several metres away and the same on every shift.

Four places where AI loses context

If you want to check where your process is leaking before you add another digital layer, look at four points. This is where the context needed for a reliable analysis is most often lost.

1. Movement of people and trucks

Without marked-out routes, the real flow exists only in the operators' heads. Two people will make the same journey in two different ways, and the system has no way of distinguishing the norm from a deviation. Marked-out traffic routes turn "this is how we walk" into a rule that can be measured.

2. How the operator receives decisions

When a set-down location or a non-conforming material zone is not marked, variability increases. One operator sets a pallet down here, another two metres further on. Analytics records this variability as "process randomness", when in fact it is a lack of a visual standard.

3. Quality of process data

Without a clear standard, the cause of an event disappears. The system will see that the material stood for 40 minutes, but it will not find out that it stood in the wrong zone, because there was no zone. Data without spatial context describes the effect, not the source.

4. Repeatability of operations

There is no analytics without a repeatable process. If the same step is performed differently every time, AI learns from blurred material and returns blurred conclusions. Repeatability begins with the space enforcing a single way of doing things.

Visual management for automation: the order that works

IDC's observation leads to a simple operational conclusion: first a visible standard, then reliable data, and only at the end smart AI. Reversing this order is the costliest implementation mistake — analytics is added to a process that cannot yet be measured reliably.

A visible standard is a layer the company controls entirely on its own and can put in place without waiting for integrations. High-contrast floor marking tapes mark out routes and zones, and floor markers unambiguously mark set-down points and decision points. Then there is addressing: legible floor and rack labels ensure that a location is the same for both people and the system. Only on this foundation does data from the shop floor start to mean anything.

Where to start on the shop floor

You do not need to overhaul the entire layout to get started. Begin with these four steps — each one closes one of the places where context was previously lost:

• Mark out routes and traffic aisles. The movement of people and trucks should follow one visible pattern, not several variants depending on the shift.
• Mark set-down zones and decision points. A place for conforming material, a non-conformity zone, a buffer — each with its own separate, consistent marking.
• Standardise the colour code and addressing. If every department marks things its own way, the process stops being scalable and the data stops being comparable.
• Check legibility in poorer light and at an angle. A standard that cannot be seen from a distance is not a standard — it is a declaration.

A simple readiness test: will a new person understand the traffic rules on your shop floor within a few minutes, without explanation? If so, you have a layer that analytics and automation can rely on. If not, no model will make up for it — it will first make the chaos visible, then accelerate it.

Frequently asked questions

Is visual management necessary when robots have ever better perception?

The more a system "sees", the more it depends on the quality of the signals it reads. An ambiguous marking is a weak input signal regardless of the class of sensor. Better perception raises the demands placed on the environment rather than lowering them.

Where should we start if we have already implemented 5S?

Extend 5S from the workstation to the flow. Check whether, in addition to order at the workstations, you have marked-out routes, marked set-down zones and uniform addressing. This is usually the missing link between a "tidy shop floor" and a "shop floor ready for data and automation".

Is this an investment for AI or for people's day-to-day work?

For both. The same layer that gives the system reliable data shortens the time people spend searching and deciding, and reduces errors. Readiness for automation and readiness for people are, in practice, the same standard.

Sources

• IDC, Hannover Messe 2026: Industrial AI in Manufacturing — https://www.idc.com/resource-center/blog/hannover-messe-2026-industrial-ai-manufacturing/