The lab spreadsheet reaches the ERP on its own

The problem

Two parallel worlds: the real spreadsheet and the official ERP.

Two parallel worlds coexist. The real spreadsheet, where the live batch data are and where the analyst actually works. And the official ERP, where the same data arrive retyped, incomplete and weeks out of date, because somebody spends half a day a month copying them and introduces typos along the way. When the auditor asks for the cross-check between lab result and batch release, that cross-check has to be done by hand. And every lab colleague who needs ERP access purely to key in results consumes a quality-module license.

• The spreadsheet holds the live batch data and is where the analyst actually works: pH, viscosity, density, oxidizing agent content, stability and shade deviation against standard.
• The ERP receives the same data transcribed, incomplete and weeks late, because somebody spends half a day a month copying it and introduces typos along the way.
• When the auditor asks for the cross-reference between laboratory result and batch release, that cross-reference has to be built by hand.
• And every laboratory person who logs into the ERP purely to type results consumes a quality module license.

How it fits the IRIS system

Connect in inward listening mode — it reads the file, not the person.

Connect "listen inward". iLEAN observes the shared folder — the file, not the person — and reads the delta the moment the sheet is saved. A grounded LLM parses the structure, which drifts over the years and between analysts and is exactly why a rigid parser has never worked, normalises it against the batch data model, shows it on screen for the quality supervisor's sign-off and posts it to the ERP through the API.

The sheet changes over the years and between analysts, which is why parsing does not depend on fixed positions. Asking the laboratory to standardize its spreadsheet is the fast route to the system going unused: here we read the sheet that exists, not the one that would be convenient.

See the full IRIS architecture →

Before and after

Today's laboratory versus the connected laboratory

Aspect	Today	With iLEAN Connect
Gap between real data and ERP	Weeks	Second zero
Manual transcription	Half a day a month	Zero
Lab ↔ release cross-reference	Manual, for the auditor	Automatic and auditable batch by batch
Copying typos	Unavoidable	Eliminated
Sheet format	Has to be respected	Tolerated as is, even as it changes
Quality module licenses	Five	One, the supervisor who validates

Impact estimate

Estimated impact — to validate against your actual invoice.

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.

• 💰 CFO angle: the ERP bills per active user. If five laboratory people hold a quality module license purely to enter results, with Connect only the validating supervisor logs in: from five down to one, up to 80 % less recurring every year.
• Half a day a month of qualified staff recovered.
• From weeks of gap between real data and ERP data, down to second zero.
• Lab ↔ batch release cross-reference automatic and auditable, instead of rebuilt by hand for every audit.

CFO angle — direct ERP license saving: the ERP bills per active user. If five people in the lab hold a quality-module license purely to key in results, that is five licences. With Connect only the supervisor who signs off goes in: one. Typical reduction of 50-80% in quality/laboratory module licences, recurring every year. Estimated payback of 4 to 9 months on the retyping saving, plus the recurring license saving, plus removing the risk of a typo in a release datum. *Estimate to validate* against the actual ERP invoice.

And the fair question from the production manager

«What if the analyst changes the sheet layout?» — that always happens, which is why parsing does not depend on fixed positions: the model interprets the structure tolerating merged cells, margin notes and mixed units. It is an anchored task, where the best models drop below 1.5% error [1], and the quality supervisor validates before anything crosses into the ERP.

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