The AOI that never drifts from the work order

The problem

Two modern systems, side by side, with a human in between.

Yield and reject data are read on a local panel and someone transcribes them into the daily report. If the work order changes and no one updates the inspection program, the AOI compares against the wrong program.

• The AOI is capable and the MES is capable, and both have documented interfaces. They simply were never stitched together, because that project never reached the top of the list.
• So yield and reject figures are read off a local panel and somebody transcribes them into the daily report, hours after the boards went through and long after the moment when reacting would have been cheap.
• The expensive failure is the other direction: if the work order changes and nobody loads the matching inspection program, the AOI compares the new board against the previous program.
• In a plant that changes customer program several times a shift, that is not a rare event — it is a permanent exposure, and what escapes is a defect nobody inspected for. The boards pass, the yield looks healthy, and the problem surfaces at the customer's incoming inspection.

How it fits the IRIS system

Connect stitches the two systems through their APIs — neither one is replaced.

Connect stitches two modern systems together via API. When the MES marks a work order start, it automatically publishes the correct inspection program to the AOI and receives yield and rejects back in real time, with no manual entry in between.

When the MES marks a work order start, the correct inspection program publishes to the AOI automatically and yield comes back in real time. The human step that could go wrong simply stops existing, which is a different thing from training people to be more careful about it.

See the full IRIS architecture →

Before and after

The isolated AOI versus the stitched AOI

Aspect	Today	With iLEAN Connect
Loading the inspection program	By hand, at changeover	Published automatically on work order start
Program misalignment	An occasional escape risk	Zero error of human origin
Yield and rejects	Read on a panel, typed later	Returned to the MES in real time
The daily report	Transcribed hours later	Built from live data
Several program changes per shift	More changes, more exposure	More changes, the same exposure: none
AOI or MES software	—	Neither is replaced

Occasional inspection-program misalignments (defect-escape risk) → zero human-origin error in program switching.

Impact estimate

Impact estimate — to be validated with your numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 3-6 months, counting only the transcription time and the changeover minutes recovered.
• Zero inspection-program misalignment: the defect escape that costs most is eliminated by construction, not by reminding people.
• Yield per work order and per customer program available live, instead of a figure typed into a report and then argued over in the next morning's meeting.
• And with high program mix, the number of changeovers stops raising the risk of inspecting against the wrong reference, so growth in customers stops meaning growth in exposure.

Payback 3-6 months · zero program misalignment. Estimate to be validated.

And the fair question from the production manager

“Wouldn't a proper integration project be cleaner?” — it would, on the day it reaches the top of the list, and that is precisely the problem: this is the classic modern island still isolated because there was never a window. This stitching replaces neither system, needs no line shutdown and does not compete in that queue. Mapping work order to inspection program is an anchored task where the best models drop below 1.5% error [1].

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.