Use cases

Each Scribe workflow is a different kind of job with its own reference documents. Enable the ones you need for a Project (see Getting started), then attach the documents listed below when you create a routing rule.

Workflow	Status	Reference documents it needs
Summary Generation	Live	Output template
Certificate of Analysis Review	Live	Material Specification, Reconciliation Report Template
Batch Record Review	Live	Master Batch Record, Prescreening Checklist Template
PQR compilation	Coming soon	—

All live workflows accept PDF or Word (.docx) source documents, produce output in the same format as your template, and cite the source for every entry. Scanned or handwritten documents are not supported, as there is no OCR step.

Summary Generation

Generate a summary of incoming CDMO data files.

The general-purpose workflow. Drop a source document, and Scribe fills in a summary template you designed, with each entry cited to the page or section it came from.

Attach

• Output Template — the Word or PDF document to fill in. It can be as simple or structured as you like: summary headings, tables, a one-page brief.

You get a copy of your template with the blanks filled in as tracked changes (Word) or colored additions (PDF), each with a citation back to the source. If you save a new version of the source file, the existing output is updated in place rather than duplicated, and fields you've already edited by hand keep your changes.

Good for technical data packages, supplier updates, meeting packs, or any recurring document you'd otherwise summarize by hand.

Certificate of Analysis Review

Review and summarize incoming Certificates of Analysis against spec.

Checks a supplier's CoA, parameter by parameter, against your own specification.

Attach

• Material Specification (role: Comparison baseline) — your internal specification for the material or product: the tests, methods, and acceptance criteria.
• Reconciliation Report Template (role: Output template) — the report layout that gets filled in with the results.

You get a completed reconciliation report with a row for each parameter, marked:

• In range — the reported result meets your specification.
• Out of range — the reported result misses your acceptance criteria.
• Method mismatch — the CoA used a different test method or unit than your spec.
• Missing from CoA — your spec requires a test the CoA doesn't report.

Each row shows the spec criteria beside the CoA's reported result, with citations to both documents, and the report carries an overall disposition. The same results appear on the run page under Reconciliation results, with flagged items counted up front.

Your specification is the only authority for pass or fail. A different specification printed on the supplier's CoA is not treated as correct, since catching that mismatch is the point.

Notes

• Every CoA is treated as new work: a new lot is never a revision of an earlier output.
• Scribe makes the call to reconcile and flag, but disposition (accept, reject, quarantine) stays with QA.

Batch Record Review

Summarize and flag exceptions in executed batch production records.

A completeness and data-integrity prescreen of an executed batch record, run before it reaches QA disposition.

Attach

• Master Batch Record (role: Comparison baseline) — the site's blank, unexecuted MBR. It defines the required steps, sign-off points, lot fields, in-process specs, and yield range the executed record is checked against.
• Prescreening Checklist Template (role: Output template) — the report layout for the checklist results.

You get a checklist with a status for each item (pass or flag, with a note) and an overall summary, grouped by ALCOA+ dimension:

• Attributable, Complete, Accurate are checked item by item against the MBR: initials, timestamps, and signatures present; lot numbers recorded; deviations cited; in-process results and yield within range; identification matches.
• Legible, Contemporaneous, Original, Consistent are checked across the whole record by scanning for what the record says about itself, such as a correction note, a mention of illegibility, a back-filled entry, or a mismatched identifier. A clean dimension shows an explicit pass rather than silence.
• Enduring and Available can't be judged from a single document, because they depend on how the record is stored afterward. The template carries a fixed explanation instead of a false pass.

On the run page, Prescreening checklist lists every item grouped by dimension and counts the flagged ones.

Good for catching blank sign-offs, uncited deviations, and out-of-range yields before they cost a review cycle.

PQR compilation

Coming soon

Compiling a full year of batch data, deviations, complaints, and stability results into an Annual Product Review template is planned but not available yet. Join the waitlist to hear when it ships.

Choosing a model

Every routing rule picks its own model. Stronger models are noticeably more reliable on long documents, such as a 100-page batch record, and on large tables. If a run flags a lot of false problems or misses obvious ones, create a new rule on a stronger model and compare. See Routing rules for how to choose one.