Batch record search that answers beside the scanned page, the step boxed and its criticality marked
Any batch-record question answered in seconds, cited to the page and step.
See one case, screen by screen ↓Every lot of a biologic leaves an executed batch record behind: hundreds of scanned pages, filled in by hand, in the plant's own language. The executed purification record for one Veltrimab lot runs to 318 pages in Dutch; the master records for the same drug substance run 386 to 412 pages per plant, in English, Dutch and French, in more than one version. Every question about a lot — a deviation, a release, a change control, an inspector's request — is answered by paging through them.
And the record is only a third of the answer. Whether the step is a critical process parameter sits in the criticality analysis, the registered limit sits in CTD Module 3, and the validated hold sits in a hold time study. A hold read against the wrong limit, a superseded version cited, or a CPP read as routine can reach a deviation, a release decision or an inspection.
Estimated split of the 42 minutes it takes to find and prove one answer by hand, and of the ≈ 14 minutes to a checked answer with the solution.
Five agents find the page, read the handwriting, translate and check criticality; a quality reviewer verifies at the boxed step.
Veltrimab drug substance, Plant 3 — Leiden, the September purification campaign. Dr. Maya Chen, quality reviewer for all sites, needs to know whether lot VLT-L3-26-0388 held its neutralized pool inside the limit. Here is what happens, screen by screen, in the working solution.
Search covers 3 products and Plants 1, 2, 3 and 5 — 2,598 batch records, with criticality and filing. An earlier answer is still open: lot VLT-L3-26-0418 held at low pH for 74 minutes, inside the 60–90 minute range, verified by Lena Ortiz. On the right, page 58 of the Dutch record with step 7.4.3 boxed and marked CPP.
“Answers cite the record, page and step · the reviewer verifies at the page.”
Each step shows as it runs. First: found lot VLT-L3-26-0388 — executed batch record, Plant 3 — Leiden, 318 pages, Dutch. Then page 58, step 7.5.1, where “before filtration” means the neutralized pool hold, and its handwritten entries are read. Then the criticality analysis, the filing and the hold study.
The steps fold to “4 steps · 3.9 s” once the answer is in.
The lot held the neutralized pool for 5 h 17 min before depth filtration — 09:45 to 15:02 on Sep 2, at 21.8 °C. The batch record allows at most 8 hours at 15–25 °C. Page 58 opens beside the answer with step 7.5.1 boxed and the operator's and checker's initials circled “signed”.
The Batch Record Agent never rounds, converts or merges a range, and says “not found” instead of guessing.
Neutralized pool hold time is a critical process parameter in criticality analysis CA-VLT-DSP-004 v3.0, table 6-3: quality attributes aggregates (SEC) and bioburden, severity 8 of 10. Filed in Module 3.2.S.2.4 at ≤ 8 h at 15–25 °C; validated for 12 h at 15–25 °C in hold time study HTS-VLT-012. A bar shows the actual hold against the limit and the validated hold.
A difference from the filing always goes to a person.
“Houdtijd vóór dieptefiltratie: maximaal 8 uur bij 15–25 °C (CPP)” becomes “Hold before depth filtration: at most 8 hours at 15–25 °C (CPP)”. The page header, the step table and the labels on the handwritten entries switch to English; the values, units and step numbers stay as written, with decimal commas shown as points.
The original step is always shown beside the translation.
J. de Vries entered the hold at 15:04 and L. Ortiz made the second-person check at 15:20 — both initials are on page 58. Maya reads the boxed step on the scanned page herself and verifies the answer there. It is now on record and reusable by the team.
“Verified at page 58 by you.”
“Show the hold before filtration for every lot this campaign.” Six lots, 0388 to 0418, page 58 of each executed record, step 7.5.1. All held inside the 8-hour limit, from 2 h 55 min to 7 h 41 min. Lot 0402 came within 19 minutes: filtration started late while filter housing F-3502 was replaced, noted on the page with deviation DV-2026-1162.
Every entry carries the operator and second-person initials.
Every question asked across the plants, with its status, confidence and answer time. Omar Haddad's questions on Protein A resin cycles and the neutralized pool hold at Plant 5 need her check; Lena Ortiz's question on AEX load conductivity at Plant 3 is marked “Differs from filing”.
Plant 3 sets AEX load conductivity at 5.0–8.5 mS/cm in step 8.2.3; Module 3.2.S.2.4 registers 5.0–8.0 mS/cm, and Plant 1 runs 5.0–8.0. The Dutch step sits beside its English translation, with the plant-vs-plant table below. Dana Okafor in Regulatory CMC has already been notified.
“Raise this with Regulatory CMC before the next Plant 3 campaign.”
To verify a full answer, Maya chooses what her signature means — the cited steps match the source documents, or fit for use in a deviation investigation or a change control — and where it is used: a verified answer for the team, change control CC-2026-0812 or deviation DV-2026-1187. Her identity is confirmed by single sign-on.
“Your electronic signature, its meaning and the cited versions are kept in the audit trail.”
The same records, seen by the four people who answer for them — QA, MSAT, site Quality and Regulatory CMC.
Five agents find the page, read the handwriting, translate and check criticality; a quality reviewer verifies at the boxed step.
Finds the record, page and step for a question; reads executed records' handwritten entries and both initials, boxes the step on the scanned page and writes the cited answer, saying whether the parameter is critical, key or non-key.
Translates Dutch and French batch-record steps into English beside the original, keeping units, ranges, equipment tags and step numbers exact.
Lines up the same process step across two plants or two versions — matched by meaning, not step number — and marks a row as different only when the values differ after unit and decimal normalization.
Finds each parameter's classification in the effective criticality analysis and its registered range in Module 3.2.S.2.2 and 3.2.S.2.4, and flags any batch-record range outside the filing, saying by how much.
Reads new and revised records in place in the quality document system, labels them by plant, product, lot, version, language and step, maps step numbers to pages and retires superseded versions.
Verifies at the boxed step and signs. The agents propose; a named person decides.
Ask in plain words — or say what to change. Every answer cites the record, page and step.
What was the hold time before filtration on lot 0388?
Lot VLT-L3-26-0388 held the neutralized pool for 5 h 17 min before depth filtration, inside the 8-hour limit. It is a critical process parameter in criticality analysis v3.0. Page 58 opens beside the answer.
Which answers still need my check?
Three answers are waiting for you. AEX load conductivity at Plant 3 sits 0.5 mS/cm above the filed range; Protein A resin cycles and the neutralized pool hold at Plant 5 need a signed check before use.
Compare Plant 3 and Plant 5 for viral inactivation
Lining up step 7.4.3 at Plant 3 with step 7.3.4 at Plant 5. They start the clock the same way, but Plant 5 holds 50–80 minutes at 16–24 °C. The table is open on Compare.
What changed between Plant 3 batch record v5.1 and v5.2?
Three ultrafiltration steps changed under change control CC-2026-0741: the TMP target narrowed to 0.8–1.2 bar, diafiltration rose from 7 to 8 diavolumes, and a new 4-hour limit applies to the over-concentrated pool.
13 screens from the working solution, on its sample data. Pick one to see it large.
How answers are written, checked and signed are settings — some, like citations, cannot be turned off.
| Setting | Default | Choose from |
|---|---|---|
| Answer language | English + original language | English + original language · Original language only · English only |
| Cite document, page and step | Always on | Required for every claim — cannot be turned off |
| Only cite effective versions | On | On · Off — superseded versions are used only when a question names them |
| Say “not found” instead of guessing | On | On · Off |
| Flag when plants differ on a critical parameter | On | On · Off |
| Notify Regulatory CMC on a filing difference | On — Dana Okafor | On · Off |
| Answers used in a deviation or change control need a signed check by | Dr. Maya Chen | Dr. Maya Chen · Omar Haddad · Lena Ortiz |
| Answers that differ from the filing need | Dr. Maya Chen + Dana Okafor | Dr. Maya Chen + Dana Okafor · Dr. Maya Chen |
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 records · “proven” = measured in production at a global medtech manufacturer, against its own expected answers · “target” = the design goal, measured in the live solution (time to a checked answer) · “estimated” = our estimate · “typical” = published figures, vendor or consultant reported (OxMaint — 6–8 hours of manual QA review per complex paper record; SAP Life Sciences symposium 2025 — ~19% of batch release cycle time; MasterControl survey cited by IntuitionLabs — ~7% with MES fully deployed across all sites). People, products, plants, lots and documents named on this page are fictional — characters and sample data in the working solution.
Search across master and executed batch records that answers a question about a lot or a step with the source itself. Batch Record Search returns the answer beside the scanned page of the executed record, with the step boxed, the entry and initials read, and whether the parameter is a critical process parameter — for a quality reviewer to verify at the page and sign.
Yes. The Batch Record Agent reads the handwritten entries and both initials on the scanned page. On lot VLT-L3-26-0388 it read the end of neutralization at 09:45, the start of filtration at 15:02 and the initials of J. de Vries and L. Ortiz on page 58 of a 318-page record.
Yes. The Language Agent puts Dutch and French steps into English beside the original and never changes a number, unit or step number. The answer language is a setting: English with the original, the original only, or English only.
The Criticality & Filing Check Agent looks the parameter up in the effective criticality analysis — critical, key or non-key, with its quality attribute and severity — and checks the registered range in Module 3.2.S.2.2 and 3.2.S.2.4 and the validated hold. A range outside the filing is flagged and goes to Quality and Regulatory CMC.
Measured in production at a global medtech manufacturer: 0.95 accuracy against the client's own expected answers, up from 0.88 and 0.75 on earlier pipelines, in under 6 seconds per answer with its citations. The dashboard keeps scoring it against an expert question set written by MSAT and Quality.
No. Records are read where they live in your quality document system, read only — nothing is copied or migrated. The Document Intake Agent labels new versions, maps steps to pages and retires superseded versions once the owner confirms the effective date.
Yes. The agents answer and check; they never edit a record, release a lot or close a deviation. A quality reviewer verifies each answer at the cited page and signs with its meaning, and answers that differ from the filing need Quality and Regulatory CMC.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your master and executed batch records, criticality analyses and filings — and it goes live once every quality gate has passed. We will walk you through it on your own records first.
Protocols drafted in minutes, every value traced to its source.
1-min demoManufacturing Batch AnalyticsBatch analysis that took three months now takes hours, with the query behind every answer.
1-min demoDeviation & CAPA ManagementEvery deviation closed inside the 30-day target, with a CAPA aimed at the cause.
1-min demoWe'll run Batch Record Search on a sample of your own master and executed batch records.