The heat-control spreadsheet, synced to the ERP

The problem

The real heat control is a file in a shared folder, and the ERP is a late copy of it.

The shared warehouse folder holds heat control data: supplier, grade, thickness, heat number, certificate. But the ERP demands manual line-by-line transcription, costing half a day a month and weeks of drift from reality.

• The shared warehouse folder holds the heat control the business actually runs on: supplier, grade, thickness, heat number and the certificate that goes with each one.
• The ERP wants exactly the same thing typed in line by line, by hand, by somebody holding a named license that gets paid for every year whether it is used for five minutes or five hours.
• Half a day a month goes into that transcription, and between one transcription and the next the ERP runs weeks behind reality.
• Which means the copy people quote from, and the copy a customer's paperwork is built on, is not the copy that is true.

How it fits the IRIS system

Connect in listen-inward mode — it reads the file when it is saved, not the person.

Connect listen-inward mode: reads the file on save, parses it with an anchored LLM, and after a single human validation, syncs it to the ERP via API.

Nobody is asked to standardize the sheet, and that is deliberate: asking the warehouse to change the file it has used for years is the fastest way to end up with two files, one of which is a lie. iLEAN reads the one that exists — columns, merged cells, notes in the margin and all — and the only new step in anyone's day is a single sign-off.

See the full IRIS architecture →

Before and after

Today's spreadsheet versus the connected spreadsheet

Aspect	Today	With iLEAN Connect
Where the truth lives	In the shared folder only	In both, and they agree
Getting it into the ERP	Line by line, by hand	Automatic after one sign-off
Time spent	Half a day a month	Minutes of review
Gap with reality	Weeks	Zero latency
The sheet's format	Has to be respected	Read as it is, even as it changes
ERP licenses on the module	4 people typing	1 person validating

Monthly manual transcription and weeks of drift → automatic sync after a single sign-off.

Impact estimate

Impact estimate — to be validated against your own 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 named user. If four people hold a quality and stock license purely to type heat data, that is a 50-80% reduction on the module, recurring every year.
• Half a day a month of somebody's time recovered, month after month.
• From weeks of drift between the sheet and the ERP to a sync after a single sign-off.
• And quoting stops being done from a copy that is out of date, which is where the expensive mistakes start.

Direct ERP license savings: from 4 licensed users down to 1 validator, -50 to -80% on the quality/stock module. Estimate to be validated.

And the fair question from the production manager

“That sheet has been passed around and edited for years — will it cope?” — that is the expected case, not the exception. Parsing does not depend on fixed cell positions: it interprets columns, units, merged cells and notes in the margin without anyone standardizing anything first. It is an anchored task, where the best models drop below 1.5% error [1], and the sync only happens after one person signs off on what was read.

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