Zero inventory deviation, CCC protected

The problem

Inventory deviation is never one failure. It is a month of small ones.

All of that translates, month after month, into a deteriorated Cash Conversion Cycle and excess-and-obsolete inventory provisions discovered at the accounting close, when it's already too late.

• A reel booked in wrong at goods-in, an approved alternate consumed without being declared, a scrapped panel nobody backflushed, a changeover that left material on the cart.
• Each one looks harmless on its own, and each one passes whatever control exists today. Together they are the reason physical WIP and the ERP stop agreeing.
• It surfaces at the monthly accounting close as an inventory deviation and, sooner or later, as an excess-and-obsolete provision — discovered when it is already too late to act on it.
• In an EMS business that is not an accounting annoyance: components are the bulk of the cost, they are bought against a customer program, and every day they sit unaccounted for is a day of deteriorated Cash Conversion Cycle.

How it fits the IRIS system

Four coordinated rings — it is everything above, orchestrated.

4 coordinated rings:

No single ring solves it: each of them already exists today in some form and the deviation still appears at the close. The strength is in the cross-check, because what gives a deviation away is four independent sources ceasing to agree, work order by work order.

• the ERP declares the BOM and expected consumption per work order;
• the AOI/placement machine reports real consumption via API;
• Edge cross-checks what was physically produced at changeover and box-build, with JIDOKA AI flagging deviations;
• every work order closes with an evidence dossier. SMED AI speeds up changeover to protect line time.

See the full IRIS architecture →

Before and after

The four rings, and what each one closes

Ring	Which cases it draws on	What it closes
1 · Declared	Kitting traveler (1) + reel receiving (7)	What the ERP says the work order should consume
2 · Reported	AOI and MES (8) + shortage (3)	What the placement machine says was really consumed
3 · Physical	Feeder changeover (10) + box-build (9)	What Edge saw physically produced and set up
4 · Closed	Evidence pack (11) + verification (5)	Every work order closing with its dossier and its signature
Cross-check	JIDOKA AI across the four rings	The deviation flagged live, not at the close
Line time	SMED AI on changeover	Changeover shortened so the control costs no capacity

Inventory deviations discovered at the monthly accounting close → flagged in real time, work order by work order. Reactive E&O provisioning → proactive prevention.

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 6-12 months, and the return is financial before it is operational.
• Protection of the account's Cash Conversion Cycle: inventory deviations flagged in real time, work order by work order, instead of discovered at the monthly close.
• Excess-and-obsolete provisioning moves from reactive to prevented — and the provision you never have to take is the cleanest line in the whole business case.
• And SMED AI shortens changeover, so the added control is paid for in recovered line time rather than charged against the capacity you sell to customers.

Protects the account's CCC · payback 6-12 months. Estimate to be validated.

And the fair question from the production manager

“Do we have to deploy all twelve cases to get this?” — no, and framing it that way is the surest way never to start. The rings build in layers and each case pays on its own from the first month. With the kitting traveler and reel receiving you already hold ring 1, which is where most of the missing declaration sits. On reliability, every cross-check is an anchored comparison between declared and reported numbers, 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.