Deviation Investigations
Investigations that start with everything you already know
Litewave builds the timeline from the source records, runs the recurrence assessment against your full deviation and CAPA history, and proposes root cause and CAPA candidates with the evidence for and against each. A first-of-a-kind event gets the same treatment, and where there is no precedent Litewave says so. Your investigator decides.
Investigations stall for two reasons, and neither is judgement
The clock on a deviation is a disposition clock. Every day an investigation stays open is a day of inventory that cannot ship.
The first reason is assembly. Pulling the batch record, equipment logs, calibration history, environmental monitoring, training records, change controls and prior lots by hand across five systems, then writing a chronology from them, is days of work before anyone has thought about cause.
The second is memory. Every investigation has to answer whether this event has happened before, and answering it well means reaching across years of records for the same vessel, the same product, the same process step, then recovering what the CAPA was and whether it held. The site holds that knowledge. It sits in an eQMS, in archived reports, and in the heads of one or two long-serving investigators, and it leaves when they do. When retrieval is that hard, the recurrence assessment leans on recall under a closure deadline rather than on the full history the site actually owns.
How it works
- The event record from your eQMS
- Executed batch record and in-process data
- Equipment logs, calibration and maintenance history
- Cleaning and changeover records
- Environmental and utility monitoring
- Analytical data and OOS records
- The site historical deviation and CAPA ledger
- Assembles a fact-checked chronology from the source documents, every entry cited
- Runs the recurrence assessment across the site history by equipment, product, process step, symptom, material lot and shift
- Drafts hypotheses with supporting and contradicting evidence
- Flags where the record is silent, which is frequently the real finding
- A pre-populated investigation record with scope and impact assessment
- A documented recurrence assessment, with the outcome and CAPA effectiveness of every similar prior event
- Candidate root causes with the evidence behind them
- A proposed CAPA and a draft effectiveness check
What it does
Assembling the facts
- Builds a minute-level chronology from batch records, equipment logs and system data, each entry cited to its source
- Runs scope and impact assessment across other batches on the same equipment, material lot or time window
- Identifies other lots requiring extension of the investigation
- Flags gaps where the record does not say what happened
Assessing recurrence
- Runs the recurrence assessment on every investigation, not only when a match is expected, and records what was searched and what it returned
- Retrieves prior similar deviations with their root causes, CAPAs and effectiveness outcomes
- Separates true recurrence from coincidence by trending frequency across equipment, product, process step, shift and operator group
- Surfaces CAPAs that were written to a symptom and did not hold, so an effectiveness failure is visible before the event returns
- States explicitly where no comparable prior event exists, so a first-of-a-kind conclusion is evidenced rather than assumed
Reaching a cause
- Drafts hypotheses in your methodology of choice, whether 5 Whys, fishbone or fault tree
- Presents evidence for and against each hypothesis, and states plainly where evidence is absent
- Proposes CAPA candidates linked to the root cause rather than the symptom
- Drafts an effectiveness check with a measurable criterion
Standing up to inspection
- Audits the closed investigation for whether the root cause is supported by the record
- Checks that the CAPA addresses the stated cause and that scope and impact are justified
- Tracks investigation aging and surfaces the ones that will breach your closure target before they do
Where the evidence comes from
At the world’s largest sulfamethoxazole producer, reviewers previously lacked immediate access to historical resolution data, forcing quality teams to investigate recurring anomalies entirely from scratch. The deployment now leverages historical deviation intelligence to compress future investigation cycles. That is the published finding, and we have deliberately not attached a cycle-time percentage to it: the 85% figure on our case study is a batch disposition number, not a deviation number.
Questions we get asked
Does Litewave decide the root cause?
No. It proposes candidates with the evidence behind each, including the evidence against. An investigator accepts, edits or rejects, and their reasoning is what goes into the record.
Do we have to move off our eQMS?
No. Litewave reads from and writes back to the system you already run. The investigation record stays in your system of record.
Our historical deviations are PDFs and Word files. Does that break it?
That is the normal starting state. Structuring the historical ledger is part of onboarding, and it is what makes similar-event retrieval useful from day one rather than after a year of new data.
Does it only help if a similar deviation has happened before?
No. Timeline assembly, scope and impact assessment, hypothesis drafting and gap analysis all run on the source records for the event in front of you, so a first-of-a-kind deviation gets the same support as a familiar one. The history search runs on every investigation regardless, because recurrence has to be formally assessed every time. When nothing comparable is found, Litewave records that the search ran and returned nothing, which is part of the investigation record rather than a silence in it.
What happens when there is no clear cause?
It says so. Litewave marks hypotheses as unsupported and lists what evidence would be needed to test them. An investigation that concludes "cause not determined, here is the containment and here is the monitoring" is defensible. One that asserts a cause the record does not support is not.
Related
Built for validated environments
Your data stays where your auditors expect it
Deploy in your environment
Your cloud, your data center, on the plant floor, or fully air-gapped. The same agents run identically in all four, because the models are open-source and served inside your perimeter, so inference never leaves it.
Your data never trains our models
Documents are never sent to third-party AI providers, and they never train a model that any other customer touches. Where a model does adapt to your site, the handwriting engine learning your operators, that loop is closed inside your deployment and what it learns stays yours.
21 CFR Part 11
Electronic records, electronic signatures and complete audit trails on every action, built to ALCOA+ data integrity principles.
EU GMP Annex 11, and building for what is next
Meets the currently effective Annex 11 expectations for computerised systems. We are building against the draft revised Annex 11 and the draft Annex 22 on AI while both remain in consultation, so nothing here is claimed as compliance with a standard that is not yet in force.
Validated like the computerised system it is
A risk-based validation package written to GAMP 5 second edition and Annex 15 expectations: specifications and traceability matrix, IQ, OQ and PQ protocols with executed evidence, and a supplier assessment pack, so your quality unit can qualify Litewave under its own CSV or CSA procedure.
AI governance you can inspect
Context of use, model versions, performance monitoring and change control are documented against the EMA and FDA guiding principles of good AI practice in drug development, so the model sits inside your quality system rather than beside it.
Evidence on every flag
Every extraction, check and recommendation links back to its source page. Nothing asserted is unverifiable.
Human in the loop by design
Agents propose, qualified people approve. Litewave never dispositions a batch on its own.
See it run on one of your own records
Bring something real. We will show you what Litewave finds in it and what it cites as evidence.