Skip to main content

Development runs. Tech transfers. Batch reviews. Neil does the legwork. Your team decides.

It reads your runs and records, then drafts the comparison, the gap assessment or the review, so you start from a draft, not a blank page.

Book a demoWatch Neil answer an email
Neil › Jobs
4 Neils awake

QualityWorking

Prepare batch B-2407 for review

Reading the batch recordChecking QC against the specTracing DEV-038 to its batches

Open the job
50+ companies, from GxP labs to GMP manufacturing.

Every Neil has its own computer.

A Neil for each kind of work, with its own instructions, access and computer. Each works on many jobs at once.

See how CDMOs keep client work separate
Neil › Personas › Development
awake, 2 jobs

Development

Compare development and manufacturing runs, and list what a transfer still needs before the first run.

Access
Development, Manufacturing. Proposes changes.Never more than the person asking it.
Computer
Its own. Keeps the script behind each comparison, so you can rerun it.
Skills
Run comparisonTransfer gap assessmentScale-upApproved by your team, shared by every Neil.
Signs
Nothing. People sign.

Quality

Check each batch record against its master record and the open deviations, ready for QA to review.

Access
Quality, Manufacturing. Proposes changes.Never more than the person asking it.
Computer
Its own. Keeps the checks it runs on every batch record.
Skills
Batch review briefCoA checkCAPA recurrenceApproved by your team, shared by every Neil.
Signs
Nothing. People sign.

Manufacturing

Configure approved procedures as processes in Seal, step by step, with their limits and checks.

Access
Manufacturing, Processes. Proposes changes.Never more than the person asking it.
Computer
Its own. Keeps its notes on each procedure, step by step, for review.
Skills
Procedure to processStep limitsApproved by your team, shared by every Neil.
Signs
Nothing. People sign.

Validation

Trace each requirement to the test that covers it, and list what nothing covers yet.

Access
Validation, Equipment. Proposes changes.Never more than the person asking it.
Computer
Its own. Keeps the script behind each trace, so you can rerun it.
Skills
Traceability matrixTest scriptsApproved by your team, shared by every Neil.
Signs
Nothing. People sign.

Your own

The work you hand it, and how your team likes it done.

Access
What you grant it, as you would a colleague.Never more than the person asking it.
Computer
Its own, ready for its first job.
Jobs
What you hand it, many at once.
Skills
The methods your team approves.
Signs
Nothing. People sign.

It shows its work.

Every finding links to the record, version and field behind it. What the records don’t show, Neil says.

Review AI and data controls

It proposes. Your team releases.

The way engineers ship code: Neil proposes the change, its tests run, QA signs, and the next run follows the rule.

See how continuous validation works
Neil › Jobs › Post-thaw viability
waiting for QA

  1. MSAT

    Why is post-thaw viability falling?

  2. Neil

    Longer holds coincide with lower viability. All eight runs stayed within the approved limit.

    From the batch records, their timestamps and QA-PROC-014 v04. The records don’t establish a cause.

  3. Neil

    Proposed changeset #14 to QA-PROC-014, for QA to review.

    Step 03: Post-thaw hold

    • Hold timeLimit 45 30 min, thaw to count
    • Over the limitOpens a deviation for QA

    12 of 12 tests passed on v05

  4. QA

    Awaiting signature

  5. RUN-218

    Follows v04 until you release v05.

It checks the work as it’s done.

Every result is read against its limits, its controls and the runs before it as soon as it’s saved. A problem reaches the person who can fix it before anyone signs.

See how CDMOs share checked work with clients
Neil › Checks › Today
6 to look at

  • 31 earlier recordsNothing to raise

ELN-2291Saved 16:40Flagged 16:41

Spike recovery below its limit, marked Pass

  1. 1Reads the result

    ELN-2291, results

    Spike recovery
    46%
    Result
    Pass

  2. 2Opens the method the record cites

    HCP-ELISA v4, section 7.2

    Spike recovery must be within 50 to 150% for a result to be reported.

  3. 3Opens the plate data

    hcp-0412.xlsx, spiked wells, ng/mL

    WellAddedFoundRecovery
    C320.09.246%
    C420.08.844%
    C520.09.648%

  4. 4Flags it at 16:41

    Spike recovery is 46%, below the method’s 50% limit, but the record says Pass. With recovery this low, the lot’s result could be about half the true value.

    To the analyst, before signing.

ELN-2307Saved 15:11Flagged 15:12

Reported from a plate whose control failed

  1. 1Reads the result

    ELN-2307, results

    Sample potency
    98%
    Reported to
    Stability summary SS-01

  2. 2Opens the method’s control limit

    POT-CB v2, section 6.1

    The reference standard’s EC50 must fall within 1.5 to 3.0 ng/mL, or the plate is invalid.

  3. 3Opens the plate reader’s export

    pot-0917.csv, curve fits

    CurveEC50, ng/mLR²
    Reference3.90.991
    Sample3.30.994

  4. 4Flags it at 15:12

    The reference standard failed its limit, so this plate can’t give a valid potency. Its 98% is already in the stability summary.

    To the analyst and the stability lead, before the summary is signed.

ELN-2351Saved 14:30Flagged 14:30

A third run of a sample that failed twice

  1. 1Sees a run of a sample tested before

    ELN-2351, started 14:30

    Sample
    STB-09-05
    Test
    Purity by SEC, 9 months

  2. 2Finds the earlier runs

    Records for STB-09-05

    RecordPurityComment
    ELN-234093.1%Invalid, retest
    ELN-234493.4%Invalid, retest

  3. 3Opens the stability protocol

    STP-12, table 2

    Purity by SEC: at least 95.0% at every time point.

  4. 4Looks for an investigation

    Deviations, STB-09-05

    Open
    None

  5. 5Flags it at 14:30

    Two results below the 95% limit were set aside with no cause recorded, and no investigation is open. This point needs one, not a third run.

    To the analyst and quality, as the run started.

ELN-2368Saved 10:58Flagged 10:59

A signed Fail, unlocked and now a Pass

  1. 1Reads the record’s history

    ELN-2368, audit trail

    WhenWhat
    Yesterday 11:20Version 3 signed: Fail
    Today 10:58Replicate 4 removed, no reason
    Today 10:58Version 5: Pass, not signed

  2. 2Recalculates with every replicate

    ELN-2368, replicates, % of label

    123456RSD
    101.696.898.6118.998.7102.77.9%
    101.696.898.698.7102.72.4%

  3. 3Opens the qualification protocol

    QP-031, section 5.3

    Intermediate precision: RSD no more than 5.0%. A result is excluded only with a documented cause.

  4. 4Flags it at 10:59

    A signed Fail became a Pass when replicate 4 was removed with no reason given. With every replicate counted, precision is 7.9% against a 5.0% limit.

    To quality and the first reviewer, before it’s signed again.

ELN-2433Saved 10:19Flagged 10:20

One expired standard behind three studies

  1. 1Reads the reagents used

    ELN-2433, reagents

    Reference standard
    RS-77
    Kit lot
    K-5512

  2. 2Finds every run on RS-77

    Records using RS-77

    RecordRunControl
    ELN-241014 Oct0.82
    ELN-241521 Oct0.79
    ELN-2433Today0.71

  3. 3Opens the certificate and the log

    RS-77 certificate, storage

    Use within 60 days of opening. The reagent log has it opened on 12 August.

  4. 4Flags it at 10:20

    Three studies used RS-77 after its use-by date of 11 October, and its control has fallen in every run. The two that passed may be affected too.

    To all three teams.

ELN-2452Saved 09:47Flagged 09:48

Raw data from a different study

  1. 1Reads the record

    ELN-2452, run

    Run
    Today
    Samples
    P-2452-01 to 06

  2. 2Opens the attached file

    titre-0814.cdf, instrument header

    Acquired
    14 Aug, 09:12
    Samples
    P-2219-01 to 06

  3. 3Finds where the file belongs

    ELN-2219, raw data

    File
    titre-0814.cdf
    Peak areas
    The same as today’s

  4. 4Flags it at 09:48

    The raw data attached here comes from a study run on 14 August. Today’s results have no raw data behind them.

    To the analyst, as the file was attached.

Ready for the inspector’s questions.

The questions inspectors ask about AI in GMP work, answered by how Neil works.

Who made this decision?
A person. Neil drafts and proposes. It signs, approves and releases nothing.
What is this based on?
The records. Each finding links to the record, version and field behind it.
What didn’t it find?
It says so. Neil names what the records don’t show. It doesn’t fill the gap.
Could the AI change the process?
Only by proposing. The change’s tests run, then your team releases it.
What could it see?
What you grant it. Never more than the person asking.
Does our data train the model?
No. Under Seal’s agreements, it isn’t used to train the underlying models.
How Neil fits Annex 22

Neil works in your Word, Excel and PowerPoint files.

Open Neil in the document you’re working on, with nothing to upload or paste. It changes the file only once you accept edits. You check the change and save.

Install Neil for Office

Give Neil a real job.

Start with the work that takes your team longest. Each job shows what Neil reads and the draft it returns.

Book a demoAll 52 jobs