BESSÓ Systems

Lab · applied experiments

The methods still work. Now they can run themselves.

Suppose the floor is standardized — legible, documented, measured. You've done something specific: you took a tangled operation and made it knowable. In Cynefin's terms you moved it from complex toward complicated — and a complicated thing can be analyzed, optimized, and now automated.

Here's what most field methods share but no one says: they all rest on observation. SMED times a changeover. Standard work audits adherence. OEE is read off a board. The method was never the expensive part — the watching was. That's the layer AI takes over: continuous, total, almost free — which frees the capacity the watching used to cost.

But analysis and action are sensitive — judgment, responsibility, where a wrong call costs steel — so they stay human. Automation doesn't remove the operator; it lifts them into a new job: watching over the systems that now watch and run the floor. And automating is itself a change — the complex part — grown with people, not dropped on them (→ Agile).

Three methods, observed by machine
01SMED · the observation: timing the changeover

SMED, when every changeover is seen

A vision of how this could look — illustrative, not a deployed system.

Question

What does SMED become when every changeover is observed, continuously, with no stopwatch?

Method in the new environment

A camera classifies every action internal/external across hundreds of runs, surfacing the pattern no sample could. The engineer stops timing and starts deciding.

What stays human

The camera sees that a wait repeats; only the operator knows why. Judgment and trust to change it stay human.

The benefit

Changeover time recovered turns straight into capacity on a line that changes format all day.

Tools
SMED worksheet
Knowledge
SMEDCynefin
02Standard work · the observation: auditing adherence

Standard work that can see whether it's followed

A vision of how this could look — illustrative, not a deployed system.

Question

What happens when the standard can see whether it's being followed — and tell you why it isn't?

Method in the new environment

The system compares the observed sequence to the documented standard continuously, surfacing drift and its pattern — as data about the standard, not a verdict on the worker.

What stays human

Deciding if a drift is a fault to fix or an improvement to adopt — and the trust that makes an operator show you the workaround, not hide it.

The benefit

Drift caught early is rework prevented — quality holds steady across shifts instead of tracking who's on.

Tools
Standard work template
Knowledge
Standard workTacit vs explicit
03Cross-metric · the observation: reading the boards

Reading the metrics together

A vision of how this could look — illustrative, not a deployed system.

Question

What does an analysis see when it reads all the metrics together — the way no single dashboard, or person, ever does?

Method in the new environment

It reads the metrics jointly and flags the contradiction a human would need months and a hunch to find — the rework hiding between OEE and cost.

What stays human

Confronting what the incoherence means — and the culture that wanted it hidden. The machine finds it; the person navigates the fix.

The benefit

The hidden loss surfaced is cost recovered — the highest-value move, because nothing else was showing it.

Tools
OEE calculator
Knowledge
What OEE hidesCynefin & resistance

Other possible applications

Same move, new ground — the observation a task depends on, automated. Short fictional probes.

Concept

The layout that draws itself

Observes the actual paths, never mapped — only the ideal layout.

Does movement becomes a heat-map; the shortest route on paper is the longest in practice.

→ flow capacity recovered, transport waste cut

VSM · motion waste

Concept

The expert who's about to retire

Observes the micro-decisions no one could write down.

Does turns the pattern into a draft standard; the "instinct" was learnable all along.

→ knowledge retained, not walked out the door

tacit vs explicit · SECI

Concept

Autonomous vehicles for routine control further out

Observes the patrol and the count themselves — ground vehicles and drones on the rounds.

Does continuous perimeter watch; inventory recounted without a team walking aisles for days.

→ pays off where manual recount or patrol eats real hours

near-miss · flow

Buildable · in development

The same move, at small-shop scale.

The big plant pays a fortune to make itself legible — SCADA, MES, the whole pyramid. The same move now fits in a box: connect the cheap signals a small operation already has — a POS, an energy meter, a couple of cameras — to an independent computer running AI, and a spare-parts store, a parts warehouse, an office back-room becomes legible. It can flag low stock and what to reorder, items about to expire, a condition drifting out of range — before anyone notices. It transfers to PyMEs in other sectors, too, wherever cheap signals exist and no one is reading them.

A working direction I'm building and testing on Linux — not a finished product.

Where it fits, the first step is a short feasibility study — AI-assisted, human-judged — because the cost depends on which signals are already there.

POS energy meter cameras sensors AI box Linux · local low stock → reorder near expiry drifting condition cheap signals you already have → made legible, locally

What the Lab is really about

Not that AI is coming for the floor — but where human attention goes once it arrives. The human moves up, not out: into reading, judging, and controlling the systems that now watch.

Where this comes from → About