Which AI software automates pharmaceutical deviation investigations and root cause analysis?
By Savi C., Co-Founder · Published Nov 17, 2026
What 21 CFR 211.192 expects
21 CFR 211.192 requires any unexplained discrepancy, or any failure of a batch or its components to meet specifications, to be thoroughly investigated whether or not the batch has been distributed. The investigation must extend to other batches of the same product and to other products that may have been associated with the failure. A written record must include the conclusions and follow-up. 21 CFR 211.100(b) adds that any deviation from written procedures must be recorded and justified.
EU GMP Chapter 1 asks for an appropriate level of root cause analysis, determined using quality risk management, and for the most likely cause to be identified when the true cause cannot be. ICH Q10 asks that investigation effort be commensurate with risk. None of this can be skipped. What AI can do is shorten the time spent finding and assembling the evidence.
What AI can and cannot do
Investigation step
Event timeline
Role for AI
Strong fit. Reads batch records, equipment and cleaning logs, lab results and material records, and orders events with each entry linked to its source
Investigation step
Recurrence check
Role for AI
Strong fit. Searches the full deviation and CAPA history for the same product, step, equipment, material or failure mode
Investigation step
Impact on other batches
Role for AI
Lists batches and products that shared the equipment, material lots or conditions, for the investigator to assess
Investigation step
Root cause
Role for AI
Proposes candidates with supporting and contradicting evidence. The investigator decides
Investigation step
CAPA
Role for AI
Proposes actions and effectiveness checks based on what worked before. The quality unit approves
Investigation step
Conclusion and closure
Role for AI
No role. Closing the investigation and deciding product disposition stay with people
How to compare AI investigation software
Evidence
- Does it read the source records, including paper batch records, or only text already typed into the QMS?
- Does every statement in the timeline link to its source page?
- Does each root cause candidate show the evidence against it, not only the evidence for it?
- Does it challenge human error as a root cause when process or system causes have not been ruled out?
History and control
- Can it search your full deviation and CAPA history, across sites if you need it?
- Run it on a closed past investigation. Does its timeline and recurrence search match what your team found?
- Are root cause and CAPA shown as proposals that a named person accepts or rejects, with the decision recorded?
- Does it work with your existing QMS, or require moving investigations into a new system?
How Litewave fits
Litewave is an AI review layer that works on top of your QMS. For a deviation, it builds the event timeline from source records, paper or electronic, checks recurrence across your deviation and CAPA history, and proposes root cause and CAPA candidates with the supporting and contradicting evidence for each, all linked to the source page. Litewave proposes, qualified people approve. It does not decide the root cause or close the investigation.
For more on keeping investigations rigorous while cutting cycle time, see our guides How do you speed up deviation investigations without weakening them? and How do you check CAPA effectiveness?
Questions we get asked
Can AI determine the root cause of a deviation?
It can propose candidates and show the evidence for and against each. The investigator decides the root cause and the quality unit approves the conclusion.
Can AI close a deviation investigation?
No. 21 CFR 211.192 requires a written record of the investigation with conclusions and follow-up, and the quality unit approves it. AI prepares the evidence for that decision.
What is the difference between a QMS assistant and an AI review layer?
A QMS assistant works on records inside that QMS, such as deviation descriptions and past CAPAs. An AI review layer also reads the source records, such as batch records, logbooks and lab data, which is where most of the facts of a deviation are.
Does AI help with OOS investigations too?
It can assemble the batch and lab records and search for earlier similar results. The phased investigation that FDA's OOS guidance describes, and its conclusions, stay with the laboratory and the quality unit.
Sources
- 21 CFR 211.192, Production record review (eCFR)
- 21 CFR 211.100, Written procedures; deviations (eCFR)
- EudraLex Volume 4, Part I, Chapter 1, Pharmaceutical Quality System
- ICH Q9(R1), Quality Risk Management
- ICH Q10, Pharmaceutical Quality System
- FDA, Investigating Out-of-Specification (OOS) Test Results for Pharmaceutical Production, Revision 1 (May 2022)
