GxP operations.
Evolve with AI.

Each traditional system was built for one part of the work. A traditional eQMS is a system of record: it files deviations and CAPAs as forms beside the work, and it cannot put a check in front of the next operator. A LIMS beside the batch reports a result without the step it governs, so a finished test becomes the start of another search. A page-based ELN keeps conditions and results in free text and disconnected files. With a traditional MES, carrying an investigation’s conclusion into the next run is another manual project.

In Seal the finding, its evidence and the response sit on the same record as the work. A corrective action can be released as a change to the workflow itself, carrying its procedure version, verification and approval, so the next batch runs the improved process instead of a revised document waiting to be applied.

You can start with one workflow alongside the systems you already have. The advantage grows as laboratory, manufacturing and quality use the same connected foundation: less reconstruction between teams, and a clearer path from a finding to an improvement that can be verified.

In this example Neil confirms the overrun against the instruction that was used: the recorded hold was 52 minutes, and MBR-014 v04 permits 45. Why the transfer was delayed, and whether the extra hold affected the product, is not yet established; the lab results alone do not close that question. Neil drafts the request to Manufacturing, asking what happened between mix completion at 10:12 UTC and transfer completion at 11:04 UTC, with the equipment, transfer and handoff records that explain any delay.

Everything in the middle column is preparation. Nothing Neil produces takes effect until a qualified person has reviewed it, and required people and signatures still control the decisions. Give Neil several jobs to work on in parallel; your team reviews the results and controls process and batch release.

In line with the EU’s draft GMP Annex 22 on AI, Neil is not used in GMP execution. It helps set up configuration, which your team verifies and releases under change control. Customer data is not used to train AI models, and the requester’s permissions govern what Neil can read. See how Neil fits Annex 22 and Neil at work.

Asked what it would change, Neil proposes keeping the 45-minute limit, calculating the hold from both timestamps and routing an overrun to an exception assessment. A missing timestamp is flagged as an incomplete check rather than a pass. This improves detection and escalation. It does not prove the cause of this delay or resolve the batch investigation.

On the step, the written instruction becomes behaviour. At or below 45 minutes the duration check is satisfied, and other checks and approvals still apply. Above 45 minutes the step requires an exception assessment before anyone decides how to proceed. A missing value cannot be treated as a completed check, and a transfer recorded before the end of mixing is sent back so the source values can be corrected.

Forms, calculations, dependencies and review gates define how the work runs. Permissions, signatures, audit history and process versions are shared controls across lab, manufacturing and quality, and the new fields reference the batch, its readings and the investigation as records rather than copied identifiers.

The first proposed version, v05-a, let the check complete without a start time. That failed run is kept as ANOM-006, and the corrected version, v05-b, adds a new result; it does not replace the failed run. In the example, 4 selected checks passed and 1 access-path check is not run. B-041 keeps v04, and the hold limit remains 45 minutes.

The change is not released yet. The process owner still verifies the exception route, including restricted-role and bypass checks. The quality reviewer assesses ANOM-006, its correction and its impact, because a later passing result does not close the anomaly by itself. The operations lead assesses training and cutover, and approval is recorded against the intended version.

Neil prepares the specifications and verification for a proposed configuration, so the evidence is drafted with the change rather than after it. GAMP 5 Second Edition supports risk-based lifecycle assurance, including iterative specification and verification. FDA’s CSA guidance addresses medical-device production and quality management system software. The assurance effort follows intended use and risk.

Your team assesses the configuration and supporting evidence, performs any additional user acceptance testing and approves release. Continuous validation does not mean automatic approval. See how continuous validation works.

Each connection is scoped before it goes live: what Seal may read, whether anything is written back, which system stays the source of truth and how the mapping is verified. Some connections are set up per project on the vendor’s API. See all integrations.

Neil works through the same connections. A question about a batch can draw on its ERP, LIMS and MES records, with each fact traced to its source.

Blueprints are starting processes for LIMS, ELN, MES, QMS, document control, training and more. See all blueprints.