SOP training impact assessment that retrains only the people a change touches, before it takes effect
Only the people a change touches retrain, all before it takes effect
See one case, screen by screen ↓It is the decision, made when an SOP revision is approved, of who must retrain on it, by which method and by when. SOP Training Impact Assessment has agents compare the approved revision with the one in effect, class every change, find who performs each changed section on the training matrix and propose a method per role; the training coordinator reviews and QA e-signs.
How to choose AI solutions for life sciences →
Updated
When a GMP procedure is revised, everyone on its training matrix has to be trained on the new revision before it takes effect. Working out who actually needs what means comparing the two revisions section by section, reading the change-control record for the reason, checking the matrix for which roles perform each changed section and deciding a method for each. With several revisions on the runway, each with its own effective date, most sites do the safe thing: the whole SOP goes to everyone to read and sign.
That costs hundreds of hours for people whose work did not change — a 26-page SOP re-read by the 784 people on its matrix is 653 hours. And it dilutes the change that matters: a new hands-on step at a pass-through hatch gets the same read-and-sign as a corrected figure number, people sign what they did not read, and assignments created late leave too little time before the effective date.
Estimated split of coordinator time on one SOP revision, by hand (about 2.5 h) and with the solution (about 21 min). Hours.
Five agents read the change, find who performs it, pick the method and write the check; the coordinator reviews, QA e-signs, and a sixth assigns the training.
Northwind Bio, Plant 3 in Leiden: SOP-MFG-0204, Entry of people and materials to classified areas, goes from Rev 11 to Rev 12 under CC-2026-0431. It is on the training matrix for 784 people and takes effect on Oct 21. Here is what happens next, screen by screen, in the working solution.
Rev 12 of SOP-MFG-0204 arrives from change control, approved Oct 7, owner Sam Patel, taking effect in 14 days. Priya's screen shows the 26-page outline on the left and, on the right, the four groups on its training matrix — QA on the floor, production supervisors, material transfer operators and 720 other cleanroom entrants. A blanket re-read is 50 minutes each: 653 h.
“784 people on the training matrix for SOP-MFG-0204 · a blanket re-read is 50 min each”
The Revision Reader aligns Rev 11 and Rev 12 by section: 26 pages, 5 changed passages. The Change Classifier finds 1 changed limit, 2 new steps and 2 editorial changes, each cited. The Training Matrix Mapper finds 4 roles, the Method Advisor proposes methods under rule set v4, and the Knowledge Check Writer drafts 3 questions. Each changed passage is now wired to the groups who perform it.
“Drafted by the agents Oct 7 · 10:15. Priya checks the changes and the method for each group, then sends it to QA. Nothing is assigned until Omar Haddad signs.”
The full review sets Rev 11 against Rev 12 for each change. In §8.2 the sporicidal contact time in the pass-through hatch goes from 5 to at least 10 minutes. The classifier calls it a changed limit at 0.97 confidence and ties it to the change-control reason: the disinfectant efficacy study DES-2026-03. It applies to material transfer operators, production supervisors and QA on the floor.
The Change Classifier never classes a change to a number, limit or time as editorial.
All three changed steps are the material transfer operators' own task at the hatch and the interlock, so the 42 operators show them to a qualified trainer: on-the-job demonstration, 30 minutes, rule R4. The 14 production supervisors and 8 QA on the floor oversee transfers but do not perform them, so they read the changed sections: 15 minutes, rule R5. The 720 others enter through the personnel airlock, §4 to §7, which did not change — they acknowledge Rev 12 in 2 minutes, rule R1.
“Blanket read-and-understand 653 h · Targeted, as proposed 51 h · Avoided 603 h”
One question per substantive change, each citing its section: how long an item stays in the hatch after the wipe, where the outer bag is removed, what goes on form MFG-0204-F3. The pass mark is 80 %. Priya can edit a question, add one, or preview the check as an employee sees it.
“‘5 minutes’ is the Rev 11 time — kept as a distractor so a pass shows the change was learned.”
Training impact record TIR-2026-0431 holds the revision, every change with its class and confidence, each role with its people, method, rule and who decided it. It updates as Priya works and downloads as a PDF. Until QA signs, it reads “Draft — not signed”.
Priya checks all five changes against the compare and accepts the agents' proposals as drafted. She sends the record to Omar Haddad, QA training approver: 1 changed limit, 2 new steps, 2 editorial, 784 people in 4 roles, 51 h against a blanket 653 h. Had she changed a class or a method, the reason would go on the record with it.
Note to QA: “Checked all 5 changes against the compare.”
Omar reviews the classes, the roles and methods, and the knowledge check, then signs with his user ID and password. The meaning of his signature is “Approved — training impact assessment”.
“Your signature is linked to this record, with your name, the date and time and its meaning. Nothing is assigned until you sign.”
With the signature verified, the Learning System Assigner creates 784 assignments on Rev 12, due Oct 20 — the day before it takes effect — and cancels 71 open assignments on Rev 11 so nobody trains on an old version. The 784 people and their supervisors are told what changed and how they train.
The assigner runs only after a valid QA e-signature, and never sets a due date after the effective date.
The same revision, seen by the four people who carry it — what it used to cost them, and what it looks like now.
Five agents read the change, find who performs it, pick the method and write the check; the coordinator reviews, QA e-signs, and a sixth assigns the training.
Aligns the revision in effect and the newly approved revision section by section, by number and heading rather than page, and lists every changed passage with its old and new text, section and page.
Classes each change as editorial, clarification, new step or changed limit, explains why in plain sentences, gives a confidence and cites the passage and the change-control reason.
Finds every role on the training matrix for the SOP, with head-count, work area and grade, and which changed passages touch each role through the sections it performs.
Proposes the training method for each role from the strongest change that applies, under the site's approved rules, with the reason, the minutes per person and the hours it costs.
Drafts one multiple-choice question per substantive change from the changed text, each citing its section, with the previous value as one of the distractors.
After QA signs, creates one assignment per person with the method, revision and due date from the signed record, cancels open assignments on the superseded revision and notifies each person and their supervisor.
Review and e-sign. The agents propose; a named person decides.
Ask about any revision, who must train, or describe a rule to add. Every answer points to the passage behind it.
Why do only 42 people retrain on the transfer SOP?
SOP-MFG-0204 Rev 12 is on the training matrix for 784 people, but its three changes to what people do are all in §8, material transfer through the pass-through hatches: the 10-minute contact time, the outer bag removed at the interlock and the new transfer record. Only the 42 material transfer operators perform §8, so they show the steps to a trainer. Supervisors and QA on the floor (22) read the changed sections. The other 720 enter through the personnel airlock, which did not change, so they acknowledge Rev 12 (rule R1). That is 51 h instead of 653 h.
What changed in the gowning SOP?
SOP-MFG-0217 Rev 07 has 9 changes; 4 matter for training: re-gown after 3 hours, not 4; any growth on a Grade A glove print is an excursion; a new step to sanitise gloves after any touch outside Grade A, the corrective action for DEV-2026-0331; and supervisors check gowning qualification before Grade B entry. The other 5 are 2 clarifications and 3 editorial changes.
Why do editorial changes not need retraining?
Because nothing anyone does changes. Rule R1, approved by Omar Haddad on Sep 2, sends editorial-only changes to an acknowledgement: the person confirms they have the new revision — 2 minutes instead of re-reading the SOP. The class, its reason and the QA signature sit in the impact record — that is the answer an inspector gets.
Who is at risk of missing an effective date?
SOP-QC-0203, HPLC system suitability, takes effect on Oct 12, in 5 days, with 7 people still to train — mostly QC analysts on night shift. Reminders went out at 06:00. If they are not done by Oct 11, the learning system takes them off HPLC work until they pass. Everything else is on track.
13 screens from the working solution, on its sample data. Pick one to see it large.
The strongest change that applies to a role decides its method. The rules, the training window and the checks are settings, versioned and on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| R1 · Only editorial changes, or none, in the sections the role performs | Acknowledge | On-the-job demonstration · Knowledge check · Read & understand · Acknowledge |
| R3 · A new step or a changed limit in the role's own task | Knowledge check | On-the-job demonstration · Knowledge check · Read & understand · Acknowledge |
| R4 · A new hands-on step in a cleanroom task | On-the-job demonstration | On-the-job demonstration · Knowledge check · Read & understand · Acknowledge |
| R6 · The same change is already tested through another open revision | Read & understand | On-the-job demonstration · Knowledge check · Read & understand · Acknowledge |
| Assignments are due | 1 day before the effective date | 1 · 3 · 5 days before the effective date |
| Reminders to the person and their supervisor | 7, 3 and 1 days before due | 7, 3 and 1 days before due · 3 and 1 days before due · Daily in the last week |
| Block work for people not trained by the effective date | On | On · Off |
| Knowledge check pass mark | 80 % | Step up or down; below it, the person re-reads the changed sections and retakes |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate · a blanket re-read is the whole SOP, 50 minutes a person. Context: 21 CFR 211.25(a) and EU GMP Part I Chapter 2 (training in the operations each person performs, before the work); 21 CFR Part 11 and EU GMP Annex 11 (electronic signatures). People, companies, sites and documents named on this page are fictional — characters and sample data in the working solution.
It is the decision, made when an SOP revision is approved, of who must retrain on it, by which method and by when. SOP Training Impact Assessment has agents compare the approved revision with the one in effect, class every change, find who performs each changed section on the training matrix and propose a method per role; the training coordinator reviews and QA e-signs.
The Training Matrix Mapper lists every role on the matrix for the SOP and the sections each role performs. A role whose sections did not change, or changed only editorially, acknowledges the new revision. Roles that must know a change but do not perform it read the changed sections. Roles whose own task changed get a knowledge check or, for a new hands-on step in a cleanroom, an on-the-job demonstration — the strongest change that applies decides.
The Change Classifier classes each change as editorial, clarification, new step or changed limit, with a reason, a confidence and citations to the passage and the change-control record. A change to a number, limit or time is never classed as editorial, and anything below 0.85 confidence is flagged for the coordinator.
No. Read & understand stays one of the four methods, alongside on-the-job demonstration, knowledge check and acknowledgement. What changes is that it goes to the roles that need it — on the sample revision, 22 supervisors and QA on the floor read the changed sections rather than 784 people re-reading the whole SOP.
Yes. The method rules, the training window, reminders, block-on-effective, cancelling assignments on the superseded revision, the knowledge-check pass mark and the approvers per area are settings. A rule change is a proposal until the QA training approver approves the new rule set.
Yes. The agents propose; the training coordinator can override any class or method with a written reason, and the QA training approver e-signs or returns the record. Nothing is assigned in the learning system until that signature is in.
It reads approved revisions from document management, the reason for change from change control, the matrix and completions from the learning system and roles from the site roster. After QA signs, it creates the assignments and cancels open ones on the superseded revision.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your training procedure, method rules, training matrix and a few recent SOP revisions — and it goes live once every quality gate has passed. We will walk you through it on your own revisions first.
An inspector's training question answered with cited records in minutes
1-min demoHR Policy Self-ServicePolicy answers for each employee's own country in seconds, clause quoted.
1-min demoEU Pay Transparency ComplianceEvery pay-information request answered within the legal deadline.
1-min demoWe'll run SOP Training Impact Assessment on a few of your own recent SOP revisions and your training matrix.