700 SOPs to assess.
The 47 that mattered, in an afternoon.

SAP ECC's end-of-life is a SOP problem disguised as an ERP problem.

Every Pharma org running SAP ECC has the same calendar entry: Q3 2027 — migrate or perish. The IT side is loud about it. The Quality side, usually, is quiet — until you ask which SOPs need to change. Then it turns out: nobody knows.

A typical mid-to-large Pharma site has somewhere between 400 and 1,200 GxP-relevant SOPs. SAP touches dozens of them — Material Master, batch release, financial closing, change control, user access, audit trail review. But not all of them. Some SOPs reference SAP transaction codes that are changing. Some reference data flows that aren't. Some are completely unrelated. Until you read each one, you don't know which.

The instinct is to start with the obvious ones — IT-Change SOPs, User Access SOPs — and hope nothing slips. The auditor will eventually disagree. Hope is not a regulatory strategy.

Step one: rank all 700 by SAP-impact confidence.

traqx's SOP Impact Assessment works on the entire Tenant SOP corpus. The QA team uploaded the 700 SOPs to the Tenant Knowledge Base in one afternoon — direct from SharePoint export, no reformatting required. By the time the upload finished, traqx had indexed every paragraph against the company's internal standards and the relevant industry regulations (Annex 11, 21 CFR Part 11, GAMP 5).

Then the Project Owner described the change in two sentences: "SAP ECC migration to S/4HANA. New transaction codes. New data model. Same business processes." Click. traqx returned a ranked list:

• VERY LIKELY (12 SOPs) · 91–94% confidence · e.g. SOP-IT-014 User Access Management, SOP-FIN-007 Financial Closing — every one of these references SAP transaction codes that change in S/4HANA
• LIKELY (28 SOPs) · 70–88% confidence · Master Data Maintenance, Batch Release Procedure — data model changes, but business logic stable
• POSSIBLY (47 SOPs) · 50–69% confidence · indirect process touchpoints, audit-relevant edge cases
• UNLIKELY / NO IMPACT (613 SOPs) · <50% · the rest, with rationale per SOP

The QA team reviewed the top 87 SOPs (Very Likely + Likely + Possibly), confirmed or rejected each with a one-line rationale, and ended the day with a signed, audit-defensible scope-of-impact: 47 SOPs to update, 653 to leave alone, every classification in the audit trail.

AI proposes the diff. The SOP Owner signs the change.

Each of the 47 SOPs opens in Edit-Mode: a side-by-side view with the current SOP on the left, traqx's AI-proposed changes on the right. The proposed changes reference the SAP S/4HANA migration context — new transaction code patterns, updated authorisation concept, the unified data model — and cross-reference the customer's own change-control SOP for the editing pattern.

The SOP Owner accepts the changes that read correctly, modifies the ones that don't, and rejects the ones that miss the local context. Each acceptance is logged. Each rejection becomes training data — traqx gets sharper for the next 46 SOPs. By the end of week one, all 47 SOPs had a draft revision in the company's QMS, ready for the formal approval workflow. No re-typing. No copy-paste from a Word doc.

Before traqx · After traqx

Three architectural decisions made the 700 → 47 number defensible.

• RAG over the Tenant SOP corpus. traqx doesn't generate from generic LLM training data — it retrieves from your SOPs. The "Very Likely" SOPs aren't a guess; they're the ones whose paragraphs literally reference SAP transaction codes.
• Confidence is reasoned, not asserted. Every classification ships with a rationale ("SOP-IT-014 §3.2 references transaction codes ME21N, ME22N, ME23N — all replaced in S/4HANA"). The QA team confirms the rationale, not the score.
• House-style preservation. Edit-Mode operates inside the customer's SOP template. The output goes back into the QMS in the format the QMS expects. traqx is not the document of record — your QMS is, your QMS approval workflow is, your distribution is. traqx is the authoring layer that makes the work tractable.