AI gap analysis for aviation operations manuals
Navlyt reads your manuals the way an auditor would — mapping each procedure to a requirement and flagging what's missing, outdated, or misaligned, with the evidence to back it up.
Aviation manuals are long, dense, and cross-referenced. Finding the gap between what a manual says and what a regulation requires is exactly the kind of careful reading that eats compliance teams' time — and it's easy to miss something.
Navlyt applies a grounded, auditable AI process to that reading. It only maps to requirements from a fixed corpus, quotes your manual verbatim, and abstains when it isn't confident — so you get evidence you can trust rather than a black box.
Recurrent testing procedure does not name the knowledge areas §135.293(a) requires
Finding NV-0142Your manual says
“Pilots complete recurrent ground training every 12 calendar months, documented on Form OPS-12 and retained in the pilot’s training file.”
General Operations Manual · §4.2.1 · Rev C, p. 87
The regulation requires
“No certificate holder may use a pilot… unless, since the beginning of the 12th calendar month before that service, that pilot has passed a written or oral test, given by the Administrator or an authorized check pilot, on that pilot’s knowledge in the following areas—”
14 CFR §135.293(a) · Initial and recurrent pilot testing requirements
Gap: the manual covers training but not the required knowledge test, and does not enumerate the subject areas §135.293(a) lists.
Grounded, not guessing
Navlyt maps only to a fixed list of regulatory requirements and never invents a regulation or citation. Low-confidence matches are flagged for review instead of asserted as gaps.
- Maps only to a defined requirements corpus
- Never fabricates a regulation or CFR cite
- Abstains ('review') when confidence is low
Every finding is cited
Each finding quotes your manual verbatim and names the requirement it relates to, so a reviewer can verify it in seconds — no hunting through the document.
- Verbatim source quotes from your manual
- Requirement reference on every finding
- Confidence score for prioritization
From gap to corrective action
Accept a gap and Navlyt turns it into a tracked corrective action with a plan and target date, then updates document coverage as you close it.
- One-click accept into a corrective action
- Coverage recomputes on resolution
- Re-analysis refreshes proposals without stacking duplicates
Human-in-the-loop
AI proposes; people decide. Findings are written as proposals and only enter your compliance record when a reviewer accepts them.
- Proposed findings, not automatic changes
- Accept, resolve, or dismiss each one
- Full audit trail of decisions
How citation verification actually works
When the analysis proposes a finding, the quoted manual text is checked character-for-character against the parsed source document. Only a verbatim match earns a displayed citation; anything less is withheld and the finding is downgraded to a review item for human eyes. This is a deliberate trade: we accept missing an occasional gap (a false negative a human reviewer can still catch) to guarantee we never show you a quote your manual doesn't contain. In compliance, the unsafe error is the overclaim.
What the analysis will not do
It will not mark your operation compliant — findings are proposals until a qualified person accepts them. It will not silently rewrite your manuals — analysis is read-only. And it will not claim frameworks it doesn't cover — Part 135 and Part 91 are live; EASA, TCCA, and SMS mapping are on the roadmap, labelled as such on every page including this one.
Frequently asked questions
How does Navlyt avoid AI hallucinations?
Navlyt constrains the model to a fixed requirements corpus, requires verbatim quotes from your manual, and marks low-confidence matches as 'review' rather than asserting them. Findings are proposals a human reviews, not automatic changes.
What does a gap finding include?
A gap finding includes the requirement it relates to, a verbatim quote from your manual, a confidence score, and a severity flag — everything a reviewer needs to verify it quickly.
What happens when I re-analyze a manual after editing it?
Re-analysis refreshes the proposed findings rather than stacking duplicates, and it preserves any findings you've already accepted or resolved so your decisions and audit trail are never lost.
Related reading
See the gap analysis on your manual
Upload an operations manual and get grounded, cited gap findings you can act on — with a human in the loop the whole way.