The batch starts at the bank.

1A traditional MES runs the recipe. Seal carries the lineage.

A traditional MES guides operators through recipe steps and files the batch record. A biologics batch also depends on living lineage, many-to-one and one-to-many transformations, product-contact genealogy, time-sensitive intermediates and laboratory decisions that gate the next step, and in most facilities that evidence sits in LIMS, spreadsheets and paper beside the MES.

A biologics batch does not begin when the production order opens. It begins with the cell substrate: the bank vial selected, the expansions that lead to the production culture and the materials and assemblies introduced along the way. Harvest, purification, viral-safety steps, holds and testing then change the physical shape of the batch several times before release.

Seal records each of those transformations as it happens, with its materials, equipment, process evidence, samples and acceptance decision, on the same platform as the laboratory and quality work. The result is one manufacturing record rather than a batch record beside a separately maintained genealogy report, and QA reviews completed stages while manufacturing continues instead of rebuilding the lineage by hand.

The process can then be refined across campaigns. Neil can propose a process change from the run evidence; an approved change applies to the next lot, and every earlier lot keeps the process version it ran.

2Neil prepares the work. Your team decides.

Neil is Seal’s AI agent. It works across the same records your team does, within the permissions of the person asking, and what it produces is work in Seal: a lineage trace, an assembled investigation, a release package or a proposed process change.

Give Neil a supplier notice, a resin concern or a freezer excursion, and it follows the affected population forward through the lineage: each assembly, culture, harvest, pool and drug-substance lot, with the contact interval and the record behind each link. It can draft the impact assessment and the evidence requests that remain. Everything it produces is preparation. Nothing takes effect until a qualified person has reviewed it, and each lot’s disposition stays with the quality unit.

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.

3Give every manufacturing system one authority.

A biologics facility relies on ERP, distributed control systems, equipment controllers, historians, LIMS, chromatography data systems and quality tools. The design works only when each object and decision has one owner.

ERP can own demand, purchasing and financial inventory. Control systems own process control and high-frequency acquisition; historians retain dense time series; CDS and instruments own analytical acquisition and processing. Seal can own the GMP order, the approved process, bank and material genealogy, accountable execution, samples, equipment eligibility, quality events and disposition. Historian and control-system connections read values; they do not change setpoints, acknowledge alarms or operate equipment. See all integrations.

Seal adds the context a manufacturing decision needs. A harvest decision carries culture state, critical values, samples, limits, reviewer and destination pool. A chromatography result carries column and resin identity, cycle, fractions, pool and acceptance.

Interfaces define acknowledgement, retry, outage and reconciliation. Missing automation or laboratory evidence stays visibly pending rather than becoming a blank field filled from memory.

4The lineage starts at the bank vial.

A master or working cell bank is not ordinary inventory. Its identity, passage history, characterisation, storage, use authorisation and remaining population determine whether a manufacturing lineage may begin.¹ Seal manages bank structures and individual vials as controlled records. Withdrawal identifies the exact vial, user, time, location, storage exposure and process order, and the vial cannot be silently substituted once execution has begun.

Each expansion, from thaw through shake flasks and seed bioreactors to the production vessel, is both a process step and a lineage event. It records source culture, target vessel, medium and supplement lots, inoculation quantity, passage, timing, samples and acceptance. Cell count, viability and contamination checks gate the next stage. If a stage misses its window or fails a criterion, the approved response is recorded against that culture rather than in a comment on the eventual production batch.

The same chain works in both directions. A freezer excursion or characterisation concern traces forward from the affected vial population to every seed, culture, harvest, pool and drug-substance lot. A released-lot investigation traverses it backward without a separate genealogy spreadsheet. The cell bank and seed lot blueprint covers the bank lifecycle in detail.

5Prepared solutions and single-use assemblies are part of the batch.

Media, buffers and single-use assemblies bring their own lots, preparation, integrity and product-contact history into the batch. Seal keeps them as linked records, so a raw-material or component finding traces into every culture or pool it reached.

Media, feeds and buffers have their own formulas, source lots, preparation equipment, adjustments, filtration, samples, hold limits and destination. Seal can execute preparation as a linked record, so the prepared lot retains its consumed materials, actual quantities, preparation values and release. At the bioreactor or purification step, scanning verifies the correct lot, status, expiry and destination, and a raw-material finding traces into every culture or pool that received it.

Bags, tubing sets, filters, sensors and connectors carry identity, lot, expiry, sterilisation and integrity requirements. Seal defines the approved assembly pattern and instantiates the physical assembly for the run: scanning checks each component and position, and connection, integrity testing, use and disposal stay linked to the equipment train and exposure window. A wrong component, expired lot or failed integrity test blocks the normal path. An authorised replacement keeps both the removed and installed configuration and the product-contact interval each experienced, so a supplier issue traces to every culture, harvest or pool the component lot contacted.

6Upstream execution carries the process state.

Control systems capture temperature, pH, dissolved oxygen, agitation, gas flow, feed and weight at high frequency. The GMP record still has to show which recipe phase was active, which data supported a decision and what accountable action followed.

Seal orchestrates the approved operation while control systems and historians remain authoritative for automation data. Critical events, phase summaries, excursions, additions and source references enter the batch record; manual connections, additions, observations and transfers receive contemporaneous attribution and prerequisite checks. Reviewers navigate from the accountable step to the relevant trend rather than to a few hand-selected numbers.

Process samples are created from the active culture, pool or hold with their method, criteria and next decision attached. The approved result returns to the waiting step without being emailed or retyped, and can permit harvest, release a pool to the next operation or open an investigation.

If a historian or controller interface is unavailable, the batch shows the missing evidence and follows the approved continuity path. Recovery reconciles message identity and time range; an interface running again does not mean all batch data has been recovered.

7Pools and columns keep their own evidence.

Harvest, clarification, capture, polishing, concentration and final filtration change the shape of the batch. Each transformation, and each column cycle behind it, keeps its own evidence and stays linked to the pools it produced.

Seal records source and destination containers, actual quantities, yields, location, storage condition and time limits. Pooling is a controlled many-to-one transformation and splitting a one-to-many transformation; neither becomes a generic inventory receipt.

Purification depends on more than the skid program. Column packing, resin lot and reuse history, sanitisation, pressure, cycle count, prior products and lifetime limits determine eligibility. Seal links each column, resin or membrane instance to the cycle and to the fractions or pool it produced, and enforces campaign and lifetime limits from actual exposure. A later resin concern identifies every cycle, load, pool and batch in the affected history.

8Viral safety is recorded for the actual pool.

The records of viral inactivation and removal steps show that the approved safety strategy was executed for the pool that received them, not only that a study supports it.

An inactivation step records start condition, setpoint range, mixing, exposure interval, end condition and review; a virus filter records identity, use conditions, throughput, differential pressure and integrity. Clearance studies support the control strategy.² The executed batch shows that the defined conditions were achieved for the actual pool. A failed or interrupted condition identifies the exact pool and downstream material requiring assessment. The viral safety blueprint covers the study side.

9Hold clocks and cold state travel with the container.

A pool held within temperature for too long is not acceptable because the current reading looks correct. Seal starts each hold clock from its defined event and carries the deadline with the container.³

Harvest holds, pool holds, filter-use limits, ambient exposure, freeze-thaw allowances and shipping windows can apply at once; Seal identifies the next limiting condition, and pause or reset behaviour is part of the approved rule rather than an operator assumption.

Storage and transport observations map to the affected interval. When an excursion occurs, impact assessment begins with the exact containers, exposure period, process stage and downstream lots. Scheduling sees material approaching a limit before it becomes an exception, so a delayed test or equipment outage shows which downstream slots remain viable.

10Change, monitoring and release read the same lineage.

Biologics processes are sensitive to changes in banks, media, single-use components, equipment, scale, methods, sites and suppliers. Seal links controlled configuration to its source documents and validation evidence, so a proposed change shows what depends on it.

A proposed material change identifies the recipes, stages, specifications, active batches and training that depend on it. Execution history shows which lots ran under which approved state, which supports comparability assessment and tech transfer.

Process validation continues through commercial production.⁴ Because parameters, attributes, yields, holds, deviations and results keep their process version, bank, material and equipment context, they can be trended by the right population, and a changed method or process version stays visible rather than mixed into one chart. A signal can open an investigation, CAPA or change with the source lots attached.

Release brings together bank and seed genealogy, materials, upstream execution, pools, viral-safety evidence, testing, equipment state, holds, deviations and changes. QA reviews completed stages while manufacturing continues, and exceptions appear against the transformation where they occurred. Disposition remains the quality unit’s decision.⁵ Later evidence, such as a supplier issue or stability trend, attaches to the original genealogy without rewriting the historical disposition.

11Start with one bank-to-release lineage.

Prove one representative lineage rather than configuring upstream, downstream, laboratory and quality as separate modules. Select one bank, seed train, production culture, harvest, purification path, final pool and release panel, with representative single-use assemblies, automation data, samples and holds.

Exercise the failure cases: a wrong bank vial, failed seed acceptance, missing historian interval, single-use integrity failure, pool quantity discrepancy, failed viral-safety condition, exceeded hold and a post-release supplier finding. The operating model is ready when each scenario blocks or permits the correct transition and identifies every affected downstream object, and when forward trace from the bank and backward trace from a released container both hold.

You can begin without replacing the systems you rely on. Where an existing MES, LIMS, eQMS or ERP stays in place, connect the records the lineage needs and agree which system owns them.