Real deployments, examined in full.
Each use case presents an operation as it ran, and the same period with spiderAI™ in place.
SYSTEM TYPEDiverse machines, one processHeterogeneous functional
A month below plan, and nobody could say what to fix first.
This is a real sinter line at a steelworks. Over thirty days it delivered 9.7 days less output than planned. Fourteen of its fifty machines flagged a problem, three had gone silent for weeks and were counted as healthy, and on day twenty the line was stopped so a crew could open machines and look. Every fact needed to rank those fourteen already existed on site. Nothing put them in order. Step through the month to see how the gap built. Change the nouns and this is a fab, a desalination train or a data-centre hall.
100%90%50%NOTHING PRODUCED
0%
Nothing has failed. The line is already below the rate it is asked to hold, and nothing tells anyone where to look.
A concern has been raised, and the only way to learn what it means is to stop the machine and take it apart.
Seven machines share one electrical supply and show the same read. Nothing relates their readings to each other, so each is inspected as though it were alone.
These have sent no signature for 23, 39 and 66 days. Silence is not health, but nothing silent ever raises a concern.
The same fourteen, in no order. This is what is left after the line has been stopped, the machines opened and the morning gone.
EVERY MACHINE HERE IS FITTED TO SEND A CURRENT SIGNATURE, CONTINUOUSLY, WHETHER IT IS RUNNING OR NOT.
THIS IS A SINTER LINE AT A STEELWORKS. CHANGE THE NOUNS AND IT IS A FAB, A DESALINATION TRAIN, A DATA-CENTRE HALL.
The line is asked to hold one hundred per cent of its planned production rate. Across the thirty days shown it almost never held that rate: it runs just short of plan for the first week, falls through the second, sits near half rate in the middle of the month, and on day twenty production stops entirely so a crew can open machines and look for the fault. It climbs back unevenly and ends the month below plan. Fourteen of fifty machines have raised a concern, seven of those share one electrical supply, and three machines are sending nothing at all. At the end of the month the same fourteen concerns remain, in no order.
NOTHING HERE WAS UNKNOWN. IT WAS UNORDERED.
A month spent below plan, and nothing on the board that says what to fix first.
The 9.7 lost days are not a mystery. Fourteen machines raised a concern and stayed on the board unranked, three sent nothing for weeks and were carried as healthy, and the only way left to find the fault was to stop the line and open machines. Every fact needed to rank that morning already existed on site, held by different people, in different systems, on different clocks.
The same month. Run at plan, no stop on day twenty, the fourteen ranked.
Here is what spiderAI™ decided on that line, day by day: the production call, the jobs for the shift, every machine placed with a next step, the fourteen concerns ranked by what each costs the line, the top job opened with its decision rule, the silent machines owned, and what is still open from last month. Ask again after the field reports back and the ranking updates.
What to do with production today, before any machine is named.
100%NOTHING PRODUCED
0%
- WITHOUT spiderAI™ · 9.7 DAYS LOST
- DAY 20 · LINE STOPPED, MACHINES OPENED, NO ORDER TO WORK IN
- WITH spiderAI™
- DAY 20 · LINE RUNS · THE ANSWER WAS WHICH MACHINE TO MEASURE FIRST, WHAT TO MEASURE, AND WHY THE OTHER THIRTEEN CAN WAIT
- DAY TWENTY · NO STOP · FOURTEEN CONCERNS, SEVEN OF THEM ON ONE SUPPLY, AND THREE MONTHS OF REPAIR HISTORY, WEIGHED TOGETHER, SAID NOTHING JUSTIFIED STOPPING THE LINE
- FOURTEEN CONCERNS · RANKED ONE TO FOURTEEN · BY WHAT EACH COSTS THE LINE · FOUR WORKED THIS SHIFT, TEN RUN ON WITH A JOB CARD
- THREE SILENT MACHINES · NAMED ON DAY ONE, EACH WITH AN OWNER AND AN ANSWER DUE BEFORE SHIFT END