Skip to main content

The GxP execution system.

Seal runs your whole GxP operation on one record, from eBR and QMS to LIMS.Neil, your AI agent, does the paperwork across all your systems. You sign.

Illustrative CAR-T release. Eight runs of version 4 of a procedure in Seal, post-thaw viability drifting down toward its 90% specification and reaching 91.4%. Neil drafts a reduction in the post-thaw hold from 45 to 30 minutes. The process owner reviews and approves version 5, whose runs hold around 96%.

50+ companies, from GxP labs to GMP manufacturing.

Every run better than the last.

Usually the deviation, the procedure, its tests, the training and the batch record live in separate systems. In Seal they are one.

  1. Found, Mon 19:40: Neil

    Lot 26-0917 › Record review, shown below

    A 21‑minute stop the batch record left out.

    Neil drafts the deviation for QA. The lot stays held.

  2. Agreed, Tue 09:07: QMS

    DEV-041 › Review, shown below

    Neil finds 4 earlier jams. QA agrees an hourly check.

    A look through the glass, so no new media fill.

  3. Approved, Tue 11:20: VLMS

    SOP-FIL-007 › Changeset #15, shown below

    All 12 test cases pass. QA approves version 13.

    3 new: Neil drafts, Validation approves, then they run.

  4. Effective, Tue 21:34: DMS and LMS

    SOP-FIL-007 › Version 13, shown below

    Night shift trained. Version 13 in force.

    The day shift's seven train before their next shift.

  5. On the line, Wed 02:10: eBR and MES

    Lot 26-0918 › Fill, shown below

    The next lot's record calls the hourly check.

    The fix: a new feed track, then ten lots without a jam.

30 hours from the finding to the check on the line, every step of change control signed. The procedure's last change, raised before Seal, took 15 weeks.

Blubber Biologics is a fictional plant. What customers measured is below.

GxP execution system

eBR, MES, QMS, DMS, LMS, LIMS, ELN and CMMS on one record. Use one, or all of them.

Explore the platform
Lot 26-0917's record at Mon 9 Nov 18:04, all eight steps signed, as eight sheets one over another: 1,912 entries in the eBR and, one square each, the counts and weighings, changes, procedures, trainings, samples, studies and calibrations they stand on in the MES, QMS, DMS, LMS, LIMS, ELN and CMMS. In focus, in red, the fill weight check at 01:07 on Fri 6 Nov, means of 10.19 to 10.22 g from 5 vials on each of four pumps, signed by A. Okafor; in ink, joined to it by a thread through the sheets, the four entries it stands on: the weighings from balance B-04, SOP-FIL-007 v12 with limits 10.00 to 10.40 g, A. Okafor's training on v12, and B-04's calibration on 3 Nov
Line 2 › Lot 26-0917
Mon 9 Nov, 18:04
  • This entry
  • What it stands on
  • One square, one entry
Wed 4 NovFri 6 NovSat 7 NovMon 9 NovDispense1Formulate2Filter3Fill4Cap5Inspect6Label7Pack8QA reviewQA
eBRentered and signedMESread from equipmentQMSdeviations and changesDMSprocedures in forceLMStrainingLIMSsamples and resultsELNstudies citedCMMScalibration, cleaning
  • This entry
  • What it stands on
  • One square, one entry
Wed 4 NovFri 6 NovSat 7 NovMon 9 NovDispense1Formulate2Filter3Fill4Cap5Inspect6Label7Pack8QA reviewQA
eBRentered and signedMESread from equipmentQMSdeviations and changesDMSprocedures in forceLMStrainingLIMSsamples and resultsELNstudies citedCMMScalibration, cleaning
Fri 6 Nov, 01:07

Fill weight 10.19 to 10.22 g

8 of 8 steps signed, 1,912 entriesNeil starts the record review at 18:04

Neil

Neil reads your records where you keep them, from SAP to AVEVA PI, and drafts the paperwork. People decide.

Meet Neil
Neil's first night on Line 2: it reviews the record of lot 26-0917 overnight, across Seal, the line's readings Seal Edge put on the lot's record and Line 2's shift channel in Teams, and drafts the record review for QA: 45 of Blubber's 50 checks met, each opening to the entry it cites, and seven things to decide. QA confirms all 50 and decides the seven. The first of them is drawn on one clock, the fill on Fri 6 Nov from 01:30 to 03:30: the batch record has its fill weight checks signed every half hour, at 02:07 and 02:37 among them, and step 4.3 says “No interventions”; F-02's feed track current is 0 A from 02:08 to 02:29; and at 02:11 the night shift wrote “Stopper bowl jammed again” in Line 2's channel
Batch records › Lot 26-0917 › Record review

Fri 6 Nov, the fill

The batch recordStep 4.3: “No interventions”

F-02's feed track, by Edge0 A from 02:08 to 02:29

Line 2's channel, Teams02:11 “Stopper bowl jammed again”

Neil, first night on Line 2WorkingDone 00:18

  1. Read the batch record, 1,912 entries
  2. TeamsRead Line 2's shift channel and handovers: 653 messages
  3. EdgeLine 2 stopped 21 min; the record says no interventions
  4. Teams“Stopper bowl jammed again”
  5. DEV-041 drafted on Blubber's form, for QA
  6. Endotoxin run invalid, spike recovery 31%
  7. Bulk held 29 h 40 min; the study covers 26 h
  8. Checked Blubber's review checklist (RC-014 v6): 45 of 50 met
  9. Cited each met check to its entry
  10. Brief ready for QA: seven to decide

Lot 26-0917: record review

Blubber's review checklist, 50 checks45 met

QA confirms all 50. Each met check opens to its entry

For QA to decide2 outside the checklist, 5 not met
Nothing is accepted until QA signsFrom Edge's readings of F-02, step 4.3 of the record, Line 2's shift channelFrom the record, Wed 18:40 to Fri 00:20, and study EXP-332 (6 to 26 h)From Room 1.214's readings during formulationFrom step 1.6 of the record and its audit trailFrom step 3.2 of the record and its audit trailFrom LIMS: the fill's 14 plates and glove plates, incubating since Fri 6 NovFrom LIMS run R-1005-02: spike recovery 31%, limit 50 to 200%

GxP › QMS

Deviations, CAPA and change control on the record they are about. Neil takes the minutes, for QA to accept.

Explore QMS
Deviation DEV-041 in review, Tuesday 10 November 09:00 to 09:30, with Neil taking the minutes; what is said is an excerpt, 09:04 to 09:09. Asked whether the stopper bowl has jammed before, Neil answers from Line 2's handovers, stand-ups and shift channel that it has jammed four times since August, none raised as a deviation, each lot's record saying no interventions. Over the review the twelve weeks are drawn, a square for each thing Neil read of Line 2 on its day: 1,240 messages, 61 meetings and 9 entries in F-02's maintenance log, over the 13 batch records of the lots filled in them; the four are marked among what was said, and none is in a batch record. It minutes what the people in the review decide: the lot held, the other lots to assess and who is told today, including Rookery Bio, the client; two causes, each with its action, the worn feed track with CAPA-019 and a step that never said to record a cleared jam with version 13 of the procedure; the hourly look at the bowl until the new track; and the training, each with its owner, for S. Amato to accept
Deviations › DEV-041 › Review
Neil is taking the minutes
  1. Detected
  2. Drafted
  3. QA triage
  4. Investigation
  5. CAPA
  6. Closed
Line 2 since August, as Neil read it
  • 1,240 messages, 61 meetings, 9 maintenance entries
  • 13 batch records
Asked by S. Amato09:04

Has this happened before?

4 timesAll on Line 2, none raised as a deviation

Nothing compared the records with the machine or the messages until Neil joined the line on Monday.

Minutes

Tue 10 Nov, 09:00 to 09:30
Filed on DEV-041 when S. Amato accepts itStep 4.4 of the lot's record. INT-07 is covered by media fill MF-26-02S. Amato at 09:05. Told: E. Marchetti (QP) and Rookery Bio, the clientT. Nakamura at 09:06: logged worn 8 Oct, current up 0.7 A in EdgeK. Brandt at 09:06, agreed at 09:07: a look, not an interventionS. Amato at 09:08. Media fill MF-26-02 covers INT-07, so no new oneRaised by Neil at 09:08; K. Brandt owns the training

GxP › Continuous validation

Validated, not frozen. Every change is tested against your processes, then your QA approves it.

How each change is tested
Changesets › SOP-FIL-007 › Changeset #15
in review
Aseptic filling
v12 → v13

Step 4.3: Stopper feed

  • Bowl checkEvery 2 hours hour during filling
  • If it jamsStop the line. Clear it as intervention INT-07. Record it before restarting.

Proposed by Neil at 09:31 from DEV-041's review. The check is a look through the glass; clearing a jam is still INT-07, so no new media fill.

v12Next check in 45 min

v13Overdue: the record holds its next entry

12 of 12 test cases passed on v13 at 09:54

3 new: Neil drafted, R. Haldane approved

  1. Clearing a jam needs a recorded intervention
  2. Its start, end and rejected vials are required
  3. A stop over 10 min opens a deviation for QA

9 standing: written by Blubber at go-live

  1. The check falls due after 60 min
  2. An overdue check holds the record's next entry
  3. Steps 4.1 and 4.2 are unchanged
  4. Steps from 4.4 on are unchanged
  5. Fill weight limits are unchanged
  6. Lot 26-0917's record stays on v12
  7. The next fill starts on v13
  8. Untrained operators can't sign 4.3
  9. The audit trail records both versions
Validation

R. Haldane10 Nov, 09:52 GMT

3 new test cases approved
Production

K. Brandt10 Nov, 10:40 GMT

Reviewed
QA

Awaiting signature

GxP › DMS and LMS

A new version is approved, trained and effective before the next shift. The line can't run on the old one.

Explore document control
Version 13 of the aseptic filling procedure in document control. QA approved it at 11:20. The evening is drawn on one clock: three of the night shift's four operators signed for it at 20:48, 20:59 and 21:12, and version 12 is still in force. Before gowning, the last of them reads its one change with Neil at the tablet, answers three questions and signs at 21:34. By the procedure's own rule version 13 is then effective on Line 2 and version 12 is withdrawn, 26 minutes before the shift starts at 22:00. The line's other seven operators are due before their next shift.
Documents › SOP-FIL-007 › Version 13
Effective 21:34
  1. Proposed
  2. Tested
  3. Approved
  4. Trained
  5. Effective
SOP-FIL-007 Aseptic fillingTue 10 Nov, 21:34

Version 13 is effective

  1. A. Okafor20:48
  2. J. Lindqvist20:59
  3. M. Rossi21:12
  4. R. Mensah21:34
  • Version 12In forceWithdrawn 21:34
  • Version 133 of 4 signedEffective 21:34
4.3 Stopper feedv12 to v13

Check the stopper bowl every 2 hours hour during filling. If it jams, stop the line. Clear it as intervention INT-07. Record it before restarting.

Step 4.3 can't be signed by anyone not trained on the version in force.

Audit trail

Version 13
Line 2's other seven operators are due before their next shiftDrafted by Neil at 09:31. Then 12 of 12 test cases passed on v13, 09:54Signed for Production, with password and authenticator codeMeaning: approved. DEV-041's review asked for it before tonight's fillEveryone on Line 2 who signs step 4.3. K. Brandt owns the trainingRead the change with Neil, answered 3 of 3 and signed before gowningRead the change with Neil, answered 3 of 3 and signed before gowningRead the change with Neil, answered 3 of 3 and signed before gowningRead the change with Neil, answered 3 of 3 and signed before gowningThe rule QA approved with v13: effective when the shift running it is trained

GxP › eBR and MES

The batch record directs the run. It calls each check, reads the instrument, and Neil flags a trend before it reaches a limit.

Explore batch records
The batch record of lot 26-0918 on the filling line's tablet: it calls each check as it falls due, takes each pump's fill weight, the mean of 5 vials, from balance B-04, calls the hourly stopper bowl check, which Neil saw done through the operator's glasses and the operator confirmed, no jam, carries Neil's draft of what the operator said about the feed track, with the frame kept, for them to sign, and Neil's warning to D. Osei, the night-shift supervisor, that pump 2's fill weight has risen at three checks running while still inside its limits of 10.00 to 10.40 g, as pumps 1, 3 and 4 hold at 10.19 to 10.22 g. At that rate it would have passed 10.40 g by 03:10. D. Osei trimmed the pump at 02:14 and the next check read 10.22 g: pump 2 never reached its limit.
Line 2 › Lot 26-0918 › Fill
Live 02:38
  1. Dispense
  2. Formulate
  3. Filter
  4. Fill
  5. Cap
  6. Inspect
  7. Label
  8. Pack
  9. QA review
Step 4.2Every 30 minHourly bowl check, new in v13

Fill weight check

10.22gPump 2, mean of 5 vials

Trimmed at 02:14. Pump 2 never reached its limit.

Batch record

SOP-FIL-007 v13
Ready to sign, none of it typedMean of 5 vials taken under INT-02, read from balance B-04 at 01:07:09; calibrated 3 NovCalled by the record at 01:10; seen by Neil through A. Okafor's glasses at 01:10:22; frame keptMean of 5 vials taken under INT-02, read from balance B-04 at 01:37:41; calibrated 3 NovMean of 5 vials taken under INT-02, read from balance B-04 at 02:07:14; calibrated 3 NovRaised by Neil from pump 2's last three means; D. Osei answered at 02:09Called by the record at 02:10; seen by Neil through A. Okafor's glasses at 02:10:31; frame keptSaid by A. Okafor at 02:12:05; drafted by Neil with the frame, linked to CAPA-019From filler F-02's controller, under D. Osei's loginMean of 5 vials taken under INT-02, read from balance B-04 at 02:37:06; calibrated 3 Nov

Neil › Voice and vision

Hands in the glove ports can't type. Neil sees what they do through smart glasses and drafts the entry for the operator to sign.

Explore voice and vision
Through operator A. Okafor's smart glasses at the filling isolator, Neil sees them take 20 vials through the glove ports for the fill weight check, times it from 02:06:02 to 02:07:04 and drafts the intervention's entry, INT-02, for them to sign, as they signed the last one at the tablet at 01:41; at the hourly stopper bowl check, a look through the glass, Neil sees no jam and A. Okafor confirms it; their note that the feed track hesitated is drafted on step 4.3, linked to CAPA-019; all three wait for their signature at the tablet
Line 2 › Lot 26-0918 › Fill
Live 02:13

Batch record

SOP-FIL-007 v13
Ready for A. Okafor to sign at the tabletSeen by Neil, 01:36:35 to 01:37:33; signed by A. Okafor at the tablet at 01:41Seen by Neil through A. Okafor's glasses, 02:06:02 to 02:07:04, vials counted; frame keptCalled by the record at 02:10; seen by Neil through A. Okafor's glasses at 02:10:31; frame keptSaid by A. Okafor at 02:12:05; drafted by Neil with the frame, linked to CAPA-019

Edge

Seal Edge reads your equipment, instruments and historian from inside your network. Each reading lands on its lot and its step.

Explore Edge
F-02's feed track current on lot 26-0917's record, read every second by Seal Edge from Line 2's filler since 8 Sep. On Fri 6 Nov it climbs to 2.7 A at 02:08 and drops to 0 A until the line restarts at 02:29. Each reading carries its lot, 26-0917, its step, 4.3 Stopper feed, its equipment, F-02, and its procedure, SOP-FIL-007 v12. Widened to Line 2's last five lots, each its own span, the mean steps up from 1.9 A to 2.6 A, and lot 26-0914 stops for 40 s on Tue 20 Oct. No limit was ever set on this reading. Neil finds the rise of 0.7 A at 19:40 on Mon 9 Nov, the first night it reads Line 2's records, and deviation DEV-041 cites the readings themselves. CAPA-019, opened at its review on Tue 10 Nov, sets an alert at 2.2 A, live from Sat 14 Nov, when the new track is fitted and reads about 1.8 A; until then an hourly bowl check stands in for it. Had the alert existed, it would have fired at 00:20 on Tue 27 Oct, on lot 26-0915, two lots before the stop
Equipment › F-02 › Feed track current
Alert at 2.2 A from 14 Nov
Lot 26-⁠0917, Fri 6 NovLots 26-⁠0913 to 26-⁠0917Tue 10 Nov, 09:30
2.70AFri 6 Nov, 02:08:00
Read each second since 8 Sep by Seal Edge over OPC UA, read-only, on each lot's recordRead by EDGE-L2 over OPC UA from F-02 at 02:08:00 on Fri 6 Nov. Quality good.Lot 26-⁠0913, Wed 14 Oct: 34,800 feed track readings in step 4.3, its span from the batch recordLot 26-⁠0914, Tue 20 Oct: 34,800 feed track readings in step 4.3, stopped 40 s at 03:41Lot 26-⁠0915, Tue 27 Oct: 34,800 feed track readings in step 4.3, its span from the batch recordLot 26-⁠0916, Mon 2 Nov: 34,800 feed track readings in step 4.3, its span from the batch recordLot 26-⁠0917, Fri 6 Nov: 34,800 feed track readings in step 4.3, stopped 21 min at 02:08From lot 26-⁠0917's batch record, running on F-⁠02 at 02:08:00The step lot 26-⁠0917's batch record was on at 02:08:00Line 2's filler, read by EDGE-⁠L2 in the line's control cabinetThe version the lot was filled on; v13 became effective at 21:34 on Tue 10 NovFound by Neil at 19:40 on Mon 9 Nov, the first night it read Line 2's recordsDEV-⁠041 cites these 2,401 readings, 02:00:00 to 02:40:00, not a picture of themLive Sat 14 Nov, with the new track at about 1.8 A. Until then, an hourly bowl checkHad this alert existed, it would have fired at 00:20 on Tue 27 Oct, on lot 26-⁠0915

Measured by the teams that run on Seal.

Read the customer stories

ERP

Costing on the record production already wrote: what a stop cost, to the dollar, and stock free to sell the minute the lot is released.

Explore the ERP
Lot 26-0917 in Seal ERP on Monday 23 November: when E. Marchetti, the Qualified Person, certifies and releases its 41,339 vials, 17 days after the fill (of 41,860 filled, 61 were rejected at the stop, 318 at inspection and 142 sampled and retained), its cost closes from the batch record, the scanner and the LIMS at $39.08 a vial, $1,615,400 for the lot against a standard of $1,611,552; the $3,848 it is over is the 21 minute line stop, $1,470, and the 61 rejected vials, $2,378, of deviation DEV-041; an order takes 40,000 vials and Neil drafts the invoice for finance, who approve it before it posts to the ledger in NetSuite
ERP › Lots › 26-0917
Released 10:12
  1. Planned
  2. Issued
  3. Filled
  4. Released
  5. Allocated
  6. Invoiced
mAb-24 drug productMon 23 Nov, 10:14

Cost a vial

$39.08$1,615,400 over 41,339 vials
Over its standard of $1,611,552$3,848
  • 21 min line stop$1,470
  • 61 vials rejected$2,378

Both are DEV-041. Neil puts five jams since August at about $18,940.

Held by QA
0
Free to promise
1,339
Promised
40,000

41,860 filled, 61 rejected at the stop, 318 at inspection, 142 sampled and retained, 41,339 held41,860 filled, 61 rejected at the stop, 318 at inspection, 142 sampled and retained, 41,339 released

Ledger

Lot 26-0917
Finance approves, then it posts to your ledger in NetSuiteDS-26-0331 as the batch record weighed it at dispensing, at its cost10R vials, 20 mm stoppers (ST-5512) and caps, each lot scanned onto Line 2The batch record's own start and end, 00:20 to 10:00, at $4,200 an hourThe lot's 19 tests in the LIMS at each method's cost: 19 of 19 passed, sterility lastCertified and released by E. Marchetti, QP: DEV-041 closed, EM and glove plates within limits, 19 of 19 results including sterilityThe disposition on DEV-041, at standard cost; DEV-041 closed Fri 20 Nov, after the sterility resultThe order of Rookery Bio, the client: the first released lot that meets its shelf life40,000 at $62.00, drafted by Neil for finance; issued on delivery

Portal

Your clients and partners see a deviation on their lot, and the change it brings, the same day your QA does. They see only what you share.

Explore the portal
The portal of Rookery Bio, the client, on its order SO-1182, Tuesday 24 November: the 40,000 vials of mAb-24 in shipment SH-0771 leave Blubber Biologics at 07:30, and the shipper's logger shows them between 2 and 8 °C the whole way to delivery at 13:10. Beside them is what the quality agreement between Blubber and Rookery names, shared by Blubber's QA: the deviation's summary on 10 November, the day it was agreed, and change notice CN-112, both sent by S. Amato and CN-112 acknowledged by D. Hale at 15:10 that day; the earlier lots assessed on 13 November, with no product impact; the deviation closed on 20 November after the sterility result; and the certificate of analysis and the batch summary at the lot's release on 23 November, certified and released by E. Marchetti, Qualified Person. The invoice for $2,480,000 is issued on delivery, and Rookery's action is to confirm receipt
Portal › Rookery Bio, the client › SO-1182
Delivered 13:10
  1. Ordered
  2. In production
  3. Released
  4. Shipped
  5. Delivered
  6. Invoiced
40,000 vials, mAb-24Tue 24 Nov, 13:12

Shipment SH-0771

4.8°CDelivered 13:10, in range for 5 h 40 min

Shared with you

By Blubber Biologics
Your lots, their deviation summaries and change notices, as your quality agreement names; shared by Blubber's QAThe deviation and the change it brings, shared with you the day DEV-⁠041 was agreedWhat happened, its two causes, the 61 vials and four earlier lots to assess; shared by S. Amato, QASOP-⁠FIL-⁠007 v13, sent 11:32 under your quality agreement with the right to object; acknowledged by D. Hale, effective 21:34Lots 26-⁠0905, 0908, 0912 and 0914: plates within limits, sterility passed, the open vials rejected each timeClosed by S. Amato, QA, with 19 of 19 results in; bulk hold and endotoxin retest accepted 13 NovCertified and released by E. Marchetti, QP, 23 Nov 10:12; from the lot's 19 of 19 resultsDrafted by Neil from the lot's 1,912 entries; shared by E. MarchettiScanned out of Blubber's cold store against SO-⁠1182The shipper's logger, read on receipt; 5 h 40 min between 3.9 and 5.4 °CEvery reading, on the shipment's record40,000 vials at $62.00; drafted by Neil, approved by Blubber's finance

Deployment

Line 2 was off paper in 38 days; Blubber's last system took 14 months. A Seal strategist stays on your site until you accept the result.

The record of Blubber Biologics' Line 2 going live on Seal, on one calendar from 24 August to 9 November 2026. The scope, Line 2's batch record, was agreed on 31 August; a working process and its validation pack were ready for Blubber's QA to review 48 hours later, on 2 September; R. Iyer, Seal's deployment strategist, was on site from 7 September until the result was accepted on 2 November, 8 weeks; the first lot on Seal, 26-0909, was filled on 22 September beside the paper record; and paper was retired from lot 26-0912 on 8 October, 38 days after the scope, where Blubber's last system took 14 months. Fill to QA release, measured on 30 October: the last four lots on paper, 26-0905 to 26-0908, took 28, 27, 26 and 27 days, and the first four on Seal, 26-0909 to 26-0912, 17, 17, 15 and 18; 14 days of each is the sterility test. J. Whitcombe, Blubber's VP Operations, accepted the result on 2 November, and Neil joined Line 2 on 9 November.
Deployments › Blubber Biologics › Line 2
Result accepted 2 Nov
Neil joins Line 2 Neil reads Line 2's records from Mon 9 Nov, a week after the result

Ready for your quality and IT review.

SOC 2 Type 2 audited. Part 11 and Annex 11 e‑signatures and audit trail. Your data trains no AI. Neil decides nothing in GMP work.

How Seal fits GxP
The validation record of Neil's record review on Line 2. Intended use: it reads a lot's batch record against QA's 50-check review list and drafts findings for QA, who confirm all 50 checks and decide each finding. Evaluated from Monday 2 to Friday 6 November on eight batch records QA had already reviewed, lots 26-0909 to 26-0916, against criteria S. Amato of QA set before it: no finding of QA's missed, and at least 95% agreement. Neil matched QA on 398 of 400 checks: 387 that both found met and all 11 of QA's findings, the checks QA had found not met. The two differences were entries it flagged that QA had accepted with a comment. Approved by R. Haldane for Validation and S. Amato for QA on Friday 6 November, and in use on Line 2 from Monday 9 November. It reads the lot's record, the filler's readings and Line 2's handovers, stand-ups and shift channel, not private chats or mail, and a change of model or configuration is evaluated again on the same records under change control.
Neil › Jobs › Record review, Line 2
In use on Line 2
  1. Intended use
  2. Criteria
  3. Evaluated
  4. Approved
  5. In use
EvaluationMon 2 to Fri 6 Nov

Against QA's own reviews

Lots 26-0909 to 26-0916, each already reviewed by QA, read from the batch record alone

  • 387 met
  • 11 QA findings
  • 2 differences

Criteria, set by S. Amato before the evaluation

Setting Neil's result beside QA's, record by record2 differences: Neil flagged, QA had accepted

Validation record

Mon 9 Nov
In use on Line 2 from Mon 9 Nov. Lot 26‑0917 was the first record it readEvery check QA had found not met, Neil flagged too8 records, 50 checks each from Blubber's review checklist, RC‑014 v6A fill weight at step 4.2. From QA's review of lot 26-0910, set beside Neil'sFrom QA's review of lot 26-0913, set beside Neil'sEach job Neil does has its own intended use, evaluation and approvalEach met check opens to the entry it cites, so QA confirms it from the record, not from NeilOnly what Blubber connected. The readings, from Edge, and the channel are for findings outside the checklistIt runs on the model and configuration evaluated. A change to either is assessed like any other changeSigned with password and authenticator code. Meaning: approvedSigned with password and authenticator code. Meaning: approved

Questions about Seal.

What is Seal?

Seal makes every GxP run better than the last. The Seal platform runs lab, manufacturing and quality work, with QMS, batch records, MES, ELN and LIMS on one record. Neil, the AI agent, reads the systems you connect, drafts the paperwork and proposes improvements for your team to approve, with Seal or without. Seal ERP adds materials, costing, orders and invoicing on the same record. Seal Edge reads your equipment, instruments and historian on site and puts each reading on its lot's record, and Seal Portal opens the record to your clients, as far as you share it.

Why Seal over a standalone LIMS, QMS or MES?

Lab, manufacturing and quality run on one platform, so a result, its batch and its investigation are linked rather than reconciled across systems, and Neil can work across all of them. Start with one workflow and add the next when you are ready.

How are changes validated?

Each change carries its specifications, test evidence and required approvals, so a validated process can keep improving instead of freezing. Your team approves each change before it takes effect, and earlier runs keep the version they used.

Can Neil change a live process?

Neil proposes changes. Each is tested, and your team approves it before it goes live. Neil works within your permissions; you can disable AI across the organisation.

How is Neil itself validated?

For each job you give Neil, your team evaluates it on representative cases against acceptance criteria your experts set, and again after any change of model or configuration. In line with the EU's draft GMP Annex 22 on AI, Neil doesn't decide anything in GMP work: a qualified person reviews what it finds or drafts, and what runs is configuration your team has tested and approved under change control.

Do we have to replace our existing systems?

No. Neil reads the systems you keep, from SAP to AVEVA PI, and drafts its work for your team. Start Seal with one process and move more work into it when you’re ready.

Where is our data hosted?

Seal agrees the hosting region with you at onboarding, and confirms the AI provider, where it processes data, retention and model terms for your organisation during onboarding and supplier review. Under Seal’s agreements, your data is not used to train AI models.

Who uses Seal?

Paradromics, Bioventix, Acclinate, Skye Biologics and MiAlgae. Read their customer stories for measured results.

See your process in Seal.

Working 48 hours after we agree its scope: tested, with evidence your QA approves.

Book a demo