Chemical customer specs and CoA with AI — the spec stays live, the certificate is automatic, the complaint never arrives.

The problem

The CSR lives in Excel and the CoA is composed by hand every night.

A chemical manufacturer with a broad portfolio really has N products × M customers × P spec revisions — and almost all of it lives in an Excel folder with murky permissions. Customer A orders product X with a tighter range on five parameters and one extra parameter; customer B orders the same product X but with a different accepted analytical method; customer C emailed a revision of its CSR four months ago that the sales rep read but the lab never received. When the batch comes through, the quality technician composes the CoA by copying from the LIMS into a Word template and checking whether the customer's file is up to date — and very often it is not.

The problem is not the quality team. It is information living in islands: the spec in Excel, the result in the LIMS, the customer and the product in the ERP, the latest revision in an email inbox. When a CoA slips through with a value in the wrong unit or an old range, the complaint does not just cost money — it costs the commercial relationship with the customer. And it was almost always avoidable.

How it fits the IRIS system

iLEAN does not replace your LIMS — it seals the crack between LIMS, ERP and the customer's CSR.

The data for a perfect CoA is already in your plant: results in the LIMS, the order in the ERP, the customer's spec in its CSR. iLEAN acts as the putty between the three, without asking you to change any of them. The spec stops being an Excel folder and becomes versioned, structured data.

Writer composes the CoA. Tracer captures every CSR revision from the customer. The agent cross-references LIMS + ERP + CSR at batch close. The person signs — never the other way around.

The three iLEAN pieces applied to chemical spec and CoA management:

• Writer — composes the per-batch Certificate of Analysis in the format agreed with each customer (parameters + value + method + seal), with a signed declaration of conformity. It supports per-customer variants without duplicating logic: one template, N customizations driven by the CSR.
• Tracer — captures every CSR revision that arrives from the customer (an email from the customer's quality manager, a supplier portal, an annex to an order) and proposes the spec update as a new version, with a diff against the previous one. The quality manager reviews, signs, and the new version becomes active for the next batches.
• Agent — at batch close it cross-references LIMS results + ERP order + the customer's active CSR. If everything checks out, it generates the CoA and leaves it ready for signed dispatch. If any parameter is out of spec, it holds the shipment and alerts the quality manager for a decision. The line does not ship a nonconforming batch without a human signature.

See the full IRIS architecture →

Before and after

Manual CSR + CoA management vs. with iLEAN

Aspect	Excel + LIMS + ERP in silos	With iLEAN Writer + Tracer + Agent
Each customer's CSR	Excel folder with murky permissions	Structured data with version and signature
CSR revision from the customer	Email to the sales rep, who may or may not tell the lab	Captured by Connect and proposed as a new version
CoA generation	Copy-paste from LIMS into Word	Automatic, at batch close
Out-of-spec detection	The quality technician's eye	Automatic hold + alert
Traceability of which CoA used which version	A folder of PDFs	Signed, searchable history
Complaints over outdated specs	Recurring, hurts the commercial relationship	Reduced — the customer's CSR is always live

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your portfolio. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Chemical manufacturer with 20-80 active customers, each with its own CSR; LIMS and ERP in silos; CoA generated by hand every shift.
• Writer + Tracer pilot focused on the 5-10 customers that generate the most complaints or move the most volume. First value expected within a few weeks: a structured CSR library + automatic CoA generation for those batches.
• Indicative payback between 4 and 9 months, depending on the average cost of today's complaints and the time spent composing CoAs by hand.
• Expected reduction in complaints over outdated specs and CoA errors of over 30% (conservative estimate).

And the quality manager's reasonable doubt

“What if the AI composes a bad CoA and we send it to the customer?” — no CoA is sent without a human signature. The agent composes, the quality manager reviews and signs. For anchored tasks like reading a LIMS result, cross-referencing it with the CSR and formatting it into the customer's template, the best models brought error below 1.5% ^[2]. The AI does not invent the data — it reads it from the source and recontextualizes it into the expected format. And even so, the three iLEAN safety rings guarantee that a nonconforming batch is held; nothing goes out unless the person signs.

^[1] Typical quality standard in automotive ≈ 25 PPM. Source: Symestic.

^[2] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.