Trace each result to the work it describes.

A traditional EM system files the count. Seal keeps the work it describes.

A traditional environmental monitoring system is a log of results. It files the plate, the count and the date it was read, while the batches, interventions and cleaning that the sample describes are recorded somewhere else.

A culture-based result describes the air, surface or person sampled days earlier. By the time the plate is read, the batches that ran during collection may have moved on. When samples, locations and production records sit in separate systems, linking a count to the work it describes is manual reconstruction, and it begins only once the result is known.

Seal keeps each sample with its location, collection window, media and incubation, beside the batch and intervention records, so the investigation starts from the collection time. Review rules are versioned. When the sampling plan or a rule changes, it becomes a new version with its own approval, and earlier and missed collections keep the plan they were taken under.

For sterile manufacture, EU GMP asks for exactly that reach. When an action limit is exceeded, the operating procedure should prescribe a root cause investigation and an assessment of the potential impact to product, including the batches produced between the monitoring and the reporting.¹ The same monitoring information is used for routine batch certification and for investigation.²

Figure 1 follows one sampling point. EM-⁠026-⁠03 was collected at P-⁠01 on 1 September, from 09:10 to 09:20 UTC, and read on 7 September: eight colonies in 1.00 m³ of air. It is the third successive increase at P-⁠01, so rule EM-⁠R-⁠04 v02 raises signal EM-⁠SIG-⁠026, although the count is below the rule’s action threshold of 10 CFU/m³. The collection window then finds the work that overlapped it: batch F-⁠041 and the INT-⁠041 transfer. F-⁠042 started later. The sections below follow that sample from its record to the programme change it may lead to.

You can start with one sampling point or trend, alongside the LIMS you already run: Seal connects to samples, tests and results in systems such as LabWare, LabVantage and SampleManager. The advantage grows as monitoring, the laboratory and manufacturing share the same records: fewer counts carried by hand between systems, and an investigation that starts from the work rather than from a search for it.

The collection time, not the read date, finds the work.

A sample record keeps its collection window, sampled volume, media and incubation, so the count can be read back against the work that was running. The read date says only when the answer arrived.

EM-⁠026-⁠03 was collected at P-⁠01 on 1 September, from 09:10 to 09:20 UTC. Its final read was recorded on 7 September at 08:45 UTC, at the end of incubation INC-⁠041. The read date does not identify which batch was running; the collection time does.

The record retains eight colonies, 1.00 m³ of air, media M26-⁠04 and incubation INC-⁠041. Collection, incubation and the final read are separate recorded events, each with its own time. Organism identification is pending, so no organism, common source or root cause has been assigned.

An unread plate, a zero count and an invalid result are recorded as different outcomes, because each leads to different follow-up. A missed or invalid sample must not become a zero count or disappear from schedule adherence; it keeps its reason, its assessment and any replacement or follow-up work.

Annex 1 asks for the programme to be set by risk assessments that establish the sampling locations, the frequency of monitoring, the methods and the incubation conditions, and for those assessments to be reviewed regularly.³ Each sample carries those choices with it: the location and method it was taken under, the media lot it used and the incubation it went through.

Find the work that overlapped the collection.

Seal compares the collection window with the recorded production and interventions, and keeps the result as an assessment: which work overlapped, why each candidate is included or excluded, and what evidence is still missing.

Batch F-⁠041 and an equipment transfer overlap the collection; F-⁠042 starts later. Timing narrows the candidate work, but it does not establish contamination or batch impact. Location, airflow, product exposure, interventions, cleaning, adjacent results and organism evidence still need assessment, and a later batch is not automatically excluded if a condition could have persisted.

Annex 1 asks for monitoring that captures all interventions and transient events,⁴ and for the potential impact of an organism found in grade A or B areas to be evaluated for each batch implicated.⁵ Keeping the candidate operations, their inclusion and exclusion reasons, the missing evidence and the resulting review together lets that evaluation start from the record instead of from schedules and logbooks.

The impact assessment stays open until the people responsible complete it. An overlap in time is evidence to assess, not a cause, and product disposition is a separate decision with its own authority.

A signal needs its rule and its source.

A signal in Seal keeps the rule version, the population and the source observations that raised it, so anyone can see why it appeared. A recurring increase stays separate from an identified organism or an established cause.

The three observations at P-⁠01 are 5, 6 and 8 CFU/m³, collected on 18 August, 25 August and 1 September. The example’s EM-⁠R-⁠04 v02 rule raises a review for three successively increasing results at the same point and method, even below its action threshold of 10 CFU/m³. These are fictional support-area criteria, not cleanroom-grade limits or regulatory defaults.

Annex 1 asks for alert levels set so that adverse trends are detected, and for monitoring procedures to define the approach to trending.⁶ The trends it names map to what a signal keeps:

Retain the rule version, population, period, exclusions, source observations and outcome, so the signal can be reproduced later. The example compares three successively increasing results at one point and method; it is not a statistical contamination-source model. Species-level similarity stays separate from confirmed strain relationships or other source evidence, and the investigator’s assessment keeps the method, confidence and identification version it relied on.

When an alert level is exceeded, Annex 1 asks for assessment and follow-up, including consideration of an investigation or corrective action.¹ A configured workflow can prepare a linked investigation with the source samples, the rule and the context; EU GMP asks for the results of monitoring to be taken into account in batch release and in the investigation of deviations.⁷ A generated record is not a completed investigation, an established root cause or a batch decision.

Neil prepares the investigation. Your team decides.

Neil is Seal’s AI agent. Ask it about a signal and it follows the location, sample and production records the requester is permitted to see, separates observations from possible explanations and prepares the investigation as work in Seal, not text to paste somewhere else.

To trace the signal into the operation, ask Neil: “Investigate P-⁠01. Compare the collection window with production, interventions and adjacent observations. Show what evidence is missing.” Neil can prepare a candidate population, a chronology and linked requests for identification and operation evidence. Ask it to explain the inclusion and exclusion reasons and the observations that would distinguish competing explanations.

To configure the sampling workflow, bring the procedure: “Turn this procedure into linked collection, incubation and result templates. Include the failure paths and verification cases.” Neil can author the required media, device, volume and time fields, source references, review states and rule configuration, with tests for a missed collection, an invalid count, an absent identification and a changed rule version. People run the verification and approve publication.

To prepare a programme change: “Use this investigation to propose a targeted monitoring change and define how we will assess whether it helped.” Neil can stage the revised plan configuration, requirement links, training-impact work and an effectiveness review with a defined population and source records. The proposed location or frequency still needs scientific assessment and authorisation.

These are illustrative requests and the work they prepare, not a recorded Neil session. People assess the evidence, verify configuration and approve decisions. Identification, impact assessment, programme changes and product disposition stay with the microbiologists and quality staff responsible for them.

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 more about Neil.

Change the programme without rewriting its history.

A sampling point has a physical position, a method and a reason to exist. Seal versions its plan, frequency, volume, operational conditions and response rules, so a change takes effect from a known point and the collections before it keep the plan they were taken under.

Configure locations and frequencies from the approved plan, with the risk rationale and the effective version. Preserve the actual collection history and assess missed work. Keep missed or invalid collections when replacement work is created, so schedule adherence shows what happened rather than what was planned.

Annex 1 asks for the risk assessments behind the programme to be reviewed regularly to confirm its effectiveness,³ and for alert levels to be reviewed periodically against ongoing trend data.⁶ A programme change carries its reason, its proposed configuration and evidence that the programme remains suitable, and it is approved before it takes effect. An investigation can add targeted locations, times and methods without changing the approved routine plan.

The programme is reviewed as a whole: coverage, adherence, results, organisms, signals, investigations, changes and risks, with the decisions, actions and effectiveness that follow. Review monitoring alongside facility, cleaning and process controls and the site’s contamination control strategy, with authorised people deciding programme changes and product disposition.

Connect the instruments and the laboratory.

Monitoring results come from samplers, counters, incubators and the laboratory, each with its own timestamps and formats. Each source is connected under an agreed mapping, and the laboratory’s records stay in the laboratory’s workflow.

Instrument connections need source-specific mapping and verification for timestamps, units, gaps, duplicates and delayed files. Assess the interface for each source, whether an export, an API or another supported route, and verify original data, identities, units, timestamps, completeness and recovery from interruption. A vendor name does not mean every device and format is already integrated.

Culture-based findings include incubation and reading delays; they are not immediate sensor alarms. Where rapid microbiological methods are adopted, Annex 1 asks for validation that shows they are equivalent or superior to the established methods.⁸ Chamber events in a stability study need original sensor evidence and historical sample occupancy, and are assessed there rather than treated as a microbiological result.

The programme and the laboratory share sample identifiers, so the relationship survives between them. Where the laboratory runs in a LIMS outside Seal, the interface needs agreed ownership, mappings and verification. In the US, aseptic processing areas need a system for monitoring environmental conditions;⁹ keep its records connected to the work they describe.

Start with one sampling point.

Bring one sampling point or trend you need to investigate, with its procedure and recent results.

Neil prepares the collection, incubation and result templates and the review rule from the procedure, and your microbiologists inspect them against how the work should run. Start with sample material or arrange an NDA before sharing confidential records.

Test the complete cycle before you rely on it: a missed collection, an invalid count, an absent identification, a changed rule version and a signal that reaches the operations it overlapped. Verify both the allowed and the disallowed paths for the intended roles.

Seal can run beside the LIMS and monitoring systems you already use. Agree which system owns each record, which records Seal reads and how status crosses the boundary. To scope the first sampling point with us, book a demo.