From the shared folder to the ERP, no re-typing

The problem

The batch is released on a spreadsheet the ERP never sees.

The quality lead fills in a spreadsheet for every batch with pH, density, active-ingredient concentration and bioassay results. That sheet lives in a shared folder the ERP never sees. The official ERP receives that data weeks late, hand-transcribed by someone who spends time every month copying and pasting — and sometimes gets it wrong.

• The quality lead fills in a spreadsheet for every batch: pH, density, active-ingredient concentration, CFU count, viability and the result of the efficacy bioassay.
• That sheet lives in a shared folder. It is the real source of truth for whether a biopesticide or a biostimulant can be released, and the one quality trusts most, because it is where the analyst actually works.
• The official ERP receives the same data weeks late, transcribed by hand by someone who spends hours every month copying and pasting — and sometimes gets a decimal wrong. On a CFU count, a misplaced exponent is the difference between a compliant batch and a useless one.
• And because the ERP charges per active user, several lab staff hold a license only to type data in, which is the most expensive way to keep the same record in two places at once, and neither of them is fully reliable.

How it fits the IRIS system

Connect in listen-inward mode — the sheet is read on save, not retyped.

iLEAN Connect in 'listen inward' mode reads the file when it's saved, parses each cell with a grounded LLM, validates it on screen with the lead, and posts it to the ERP via API.

The lab keeps working in the spreadsheet it trusts. iLEAN reads the file the moment it is saved, checks each cell against the expected range, shows it to the quality lead for validation and posts it to the ERP. Only one person needs a license to do that, and nobody has to learn a new screen.

See the full IRIS architecture →

Before and after

The lab sheet in a shared folder versus the lab sheet posted to the ERP

Aspect	Today	With iLEAN Connect
Delay between sheet and ERP	Weeks	Minutes after saving
Bioassay and CFU results in the ERP	Typed by hand, sometimes wrong	Parsed cell by cell and validated
Out-of-range value	Noticed when someone looks	Flagged before posting
Users logging in just to type	Several licenses	Only the validating lead
Monthly copy-and-paste	Hours of a qualified analyst	None
The lab's working tool	The spreadsheet	The same spreadsheet

Today the ERP charges per active user: if several lab staff log in just to enter data, that's several licenses. With Connect, only the validating lead needs to log in — a typical 50 to 80% reduction in quality-module licenses. Estimated payback 4 to 9 months.

Impact estimate

Impact estimate — to be validated against your actual ERP 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.

• Estimated payback 4-9 months.
• A typical 50-80% reduction in quality/lab-module ERP licenses, because only the lead who validates needs to log in.
• Release data reaching the ERP in minutes instead of weeks, which is what lets planning ship a batch as soon as the bioassay clears, instead of holding it until the ERP catches up.
• And the transcription errors that surface at the worst moment — an inspection, a field complaint, a registration renewal — stop being produced, because nobody transcribes.

Estimated payback 4-9 months · typical 50-80% reduction in quality/lab-module ERP licenses.

And the fair question from the production manager

“And if it reads a CFU value in the wrong column?” — parsing a spreadsheet whose structure the lab uses every day is an anchored task where the best models drop below 1.5% error [1]. On top of that, every value is checked against its physical and specification range, and the quality lead validates the screen before anything is posted. A value outside its range is never posted silently: it waits for a decision.

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