Flight operations manual search that answers from the revision in force, with the page open beside it
Every answer from the revision in force that day, with the page open beside it.
See one case, screen by screen ↓Flight operations manual search is AI search for flight manuals: pilots and dispatchers ask a question in plain words and get the answer from the current manual, page cited, instead of guessing which manual holds it and searching by keyword. Flight Operations Manual Search adds seven agents that answer in two sentences from the revision in force on the day of the flight, open the page beside the answer, ask back when the answer depends on type, station or phase, and send what no manual answers to Flight Standards.
How to choose AI for airlines: a buyer's guide to agentic AI solutions →
Updated
An airline flies on its manuals, and flight operations manual search has to cover all of them: the Flight Operations Manual, an FCOM, QRH and MEL for every fleet, the Dispatch Manual, the OpSpecs and a briefing for every station. In the working solution that is 46 manuals and 61,380 pages, carried as electronic flight bag manuals on 4,330 EFBs, and they change every week. 14 CFR 121.137 asks every crewmember to keep their copy up to date and have it accessible while working. Yet finding an answer today still means knowing which manual holds the answer, opening the bookmark, running a keyword search that matches words rather than meaning, and checking the revision on the corner of the page.
The risk in aviation manual search sits between revisions and between manuals. A device that missed last night's sync shows the superseded page. A question no manual answers goes to the chief pilot's office or a Flight Standards inbox, and nobody counts how many crews asked the same thing. After a merger, two FOMs fly side by side and transitioning crews carry the old habit. And when a revision changes an answer, the crew notice goes to everyone — the pilot who asked that exact question last week is never told their answer changed.
Estimated split for one Flight Standards manager at a large network airline, by hand and with the solution.
Seven agents answer from the revision in force, ask back when they must, and catch what a revision changed; Flight Standards decides.
Thursday 15 October. Corvane Airways flight CVA 118, a 787-9 from Denver to London Heathrow, is in cruise at FL380, due in at 09:55Z. The LHR ATIS gives runway 27L, wind 230° at 24 knots, light rain, runway wet, and the crew plans an autoland. Here is what happens next in the working solution, from the flight deck to the Flight Standards office.
Erin types “Max crosswind for autoland, 787-9, wet runway?” The trip strip already shows CVA 118, N827CV, the Corvane certificate and her library in force. In 3.8 seconds the Manual agent answers in two sentences — 25 knots on a wet runway with braking action good, 15 knots with braking action medium or worse — and 787 FCOM page 2.11.4 opens beside it with the row highlighted and revision 42, effective 12 Oct 2026, on its corner. From the ATIS it works out the crosswind for 27L: 15 knots, within the limit.
“From the revision in force for your flight, every sentence on its page. The pilot in command decides.”
A note under the answer says revision 41 said 20 knots, with no separate limit for braking action medium. Compare revisions puts revision 41 (superseded 12 Oct) beside revision 42 (in force), the old 20 knots struck out and the new lines marked. When she follows up with “And the A321neo?”, the answer comes from the A321neo FCOM, revision 38, with a note that revision 39 takes effect Mon 19 Oct and does not change this limit.
The Manual agent never answers from a superseded or not-yet-effective revision.
“Noise abatement departure procedure?” depends on the station, so the agent asks which one, with the choices already filled in from her trip: LHR, her next departure on CVA 119 tomorrow at 12:30Z, or DEN, home base. She taps LHR and gets NADP 2 from the Corvane Airport Briefing LHR, revision 15, page 7, with the page open beside the answer and the procedure highlighted.
The Flight context agent never guesses a fleet or a station; when the trip does not say, the Manual agent asks.
Erin asks whether she may wear noise-canceling earbuds instead of the headset in cruise. No page supports an answer, so the agent gives none. It lists the nearest pages — FOM 2.4.6 on headsets below FL180 and 787 FCOM 5.20.2 on the headset jack — and one tap sends the question to Flight Standards, where it joins group N-77: 40 other crews have asked the same thing this month.
“I found no page that supports an answer, so I won't give one.”
Nadia's home shows the week: 18,642 questions, 94.1% answered with a page, 7 flags open and 214 answers a revision changed. Six items need her, each with its clock — a flag on a dated FAA list due in 2 h 40 min, an urgent safety question about lithium battery smoke in the crew rest, and the revision 42 note to 187 people due today. Beside it, the agents' morning brief cites the run behind every line.
Capt. Ji-ho Park, who joined from Halden Air, flagged an answer on single-engine taxi: “The Halden FOM I trained on says 5 minutes. Which one do I fly?” The Conflict agent checked every manual: the Corvane FOM 4.6.2 says 3 minutes, the Halden Air FOM 2.9.1 says 5, the 737 MAX FCOM gives no cool-down time, and FOM 1.3.2 says the Corvane FOM applies on every Corvane flight. The Flag agent drafted a reply to confirm the answer, at confidence 0.93.
The Conflict agent never decides which manual is right. The Flag agent never closes a flag or replies to a pilot.
Nadia picks the reason code “Answer correct · the manuals differ by certificate” and reads the drafted reply before it goes. It tells Captain Park the Halden time applies only to E175 flights under the Halden certificate until 1 Mar 2027, and that harmonization item H-37 is open. The reply goes to his EFB and to the two duplicate flags, signed by her and logged with its reason.
“Signed by Nadia Brooks, Flight Standards Manager. The reply goes to the pilot's EFB; the decision is logged with its reason code.”
787 FCOM revision 42 went into force on Mon 12 Oct and changed 63 pages. The Revision agent replayed 12,940 real questions on the 787 against the new pages and found 214 answers that are now different, asked by 187 people — Erin among them, told 20 knots on 1 Oct. Each row shows the question, the answer then and the answer now; 1,212 of 1,560 crew have read and acknowledged the general crew notice.
The note CN-26-118 is drafted and waits for her. She approves and sends it with the reason code “A revision changed their answer”. Each of the 187 people gets their own question, the old answer and the new answer, each with its page — not one more notice to everyone.
The Revision agent never sends a note; Flight Standards approves.
At the Denver dispatch desk, Marcus asks whether CVA 412, an A321neo with MEL 30-11-01 open for the left wing anti-ice valve, can go into ORD tonight. In 3.2 seconds: no — with that item open the flight may not operate in known or forecast icing, and the ORD TAF has freezing drizzle from 18:00Z to 21:00Z, overlapping the arrival window. The MEL page opens beside the answer, and one click adds a cited line to the release remarks.
“Ask about the flight you are releasing. The release is yours.”
The same library, seen by the five people who rely on it — what their week looked like, and what it looks like now.
Seven agents answer from the revision in force, ask back when they must, and catch what a revision changed; Flight Standards decides.
Answers in two sentences from the revision in force on the flight's date, with every sentence on its page; asks back on type, phase or station when the answer depends on it. Searches the manuals by meaning, not keywords, and reads the ATIS and the forecast to work out values such as the crosswind component.
Reads the trip — flight, tail, fleet, station, phase of flight and certificate — from crew scheduling and the release, and checks which revisions the device holds.
Checks every sentence against its passage and checks that the page is in the revision in force, before anyone sees the answer.
When a revision is published, compares it page by page with the one in force, replays 90 days of real questions against the new pages and drafts the “your answer changed” note for the people who asked.
Finds where manuals disagree on the same subject: the FOM, the FCOM and station pages, and the airline's and the acquired airline's manuals.
Groups the questions no manual answers by meaning, finds the nearest pages, and marks safety-related groups urgent with a 4-hour clock.
Turns a pilot's flag into a case: the question, the answer, the page, a check across manuals, duplicates, and a proposed reply with a reason code.
Replies and notes; pilot in command and dispatcher decide. The agents propose; a named person decides.
Flight Standards can question any agent about its work. Every answer cites the revision index, the run or the page behind it.
Why revision 42 and not 41?
The flight is on Thu 15 Oct. Revision 42 has been in force since Mon 12 Oct, so it is the only 787 FCOM revision I may answer from. I said what revision 41 said, because the answer changed.
When do you refuse to answer?
When no page supports a sentence, or my confidence is below 0.85. Then I say “the manuals do not answer this”, list the nearest pages and file the question. This week: 596 times.
How did you find the 214 answers?
I replayed 12,940 real questions on the 787 from go-live on 7 Sep to 11 Oct against the 63 changed pages and kept each answer that came out different. 214 did, asked by 187 people. The note waits for Flight Standards.
Which differences between the two FOMs matter most?
Single-engine taxi (3 against 5 minutes) is asked most: 612 questions this week. I do not decide which manual is right; FOM 1.3.2 says the Corvane FOM applies on every Corvane flight.
13 screens from the working solution, on its sample data. Pick one to see it large.
Thresholds, rules and service levels are settings. Each change is saved as a new version and logged, and the safety rules are locked by Flight Standards.
| Setting | Default | Choose from |
|---|---|---|
| Answer confidence floor | 0.85 | Below it: “the manuals do not answer this”, and a person is offered |
| Sentences per answer | 2 | A third only for a computed value |
| Revision recall window | 90 days of questions | Replayed against each new revision |
| Golden test pass mark | 97% with the right page | Run weekly and after every revision |
| Device behind, flag after | 24 hours | Pilot and EFB administrator notified |
| Revision by effective date and flight date | On | Never a superseded or not-yet-effective revision · locked by Flight Standards |
| Safety-related no-answer is urgent | On | 4-hour clock, copied to Safety · locked by Flight Standards |
| Reply to a flag | 3 business days | Warn at 75% · Flight Standards Manager, then the Director |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate; the 2 min 50 s baseline is an estimate from the sample airline's 412-question timing study. Regulations named as context: 14 CFR 121.133, 121.135 and 121.137 (manuals kept current and accessible; electronic manual revision dates per Amdt. 121-388, 30 May 2023), 14 CFR 121.533 (operational control), FAA AC 120-76E (12 Jun 2024, EFB authorization). Corvane Airways, Halden Air, and the people, flights and tails named on this page are fictional — characters and sample data in the working solution. The limits on its manual pages are fictional content written for the demo, not quoted from any manufacturer's manual.
Flight operations manual search is AI search for flight manuals: pilots and dispatchers ask a question in plain words and get the answer from the current manual, page cited, instead of guessing which manual holds it and searching by keyword. Flight Operations Manual Search adds seven agents that answer in two sentences from the revision in force on the day of the flight, open the page beside the answer, ask back when the answer depends on type, station or phase, and send what no manual answers to Flight Standards.
It chooses the revision by its effective date and the flight's date. On 15 Oct it answers from 787 FCOM revision 42, in force since 12 Oct, and says what revision 41 said when the answer changed; an upcoming revision is noted but not used until it takes effect. An EFB that has not synced still gets the server's revision in force, with a “behind” notice.
Every sentence must sit on a page. The Citation checker tests each one against its passage and removes any it cannot support, and below a 0.85 confidence floor the answer is “the manuals do not answer this”, with the nearest pages. The question is then filed for Flight Standards and grouped with others like it.
The Revision agent replays the last 90 days of real questions against the new pages and lists every answer that is now different and who asked it — 214 answers and 187 people for 787 FCOM revision 42 in the working solution. It drafts a note to each of them with their question, the old answer and the new one, and Flight Standards approves it before it is sent.
Yes. A dispatcher asks about the flight being released, and the Flight context agent has already read the release: tail, MEL items, weather and alternates. The answer cites the MEL, the Dispatch Manual or the OpSpecs from the revision in force, and a cited line can be added to the release remarks. The release stays the dispatcher's decision.
14 CFR 121.137 asks each crewmember to keep their manual up to date and accessible while working. The solution answers only from the revision in force, shows the revision and effective date on every page, lists every EFB behind on the revisions in force, and keeps a chained record of what each person was told and from which page. It works alongside your manual system and your EFB program under FAA AC 120-76E; it does not replace them.
Yes. The agents never change a manual, never send a crew note and never reply to a pilot. A Flight Standards Manager decides every flag, no-answer group and changed-answer note with a reason code, and the pilot in command and the dispatcher decide what to do with any answer.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your manuals with their revision history, roles and audiences, and a set of golden questions from Flight Standards — and it goes live once every quality gate has passed. In the working solution the golden test is 412 questions, each with the right page.
Recovery plans in minutes, each one rehearsed and costed, signed before anything moves.
1-min demoFlight Dispatch ReleaseEvery release checked before signature, with a source behind every number.
1-min demoNOTAM Briefing238 notices become the 6 that change the flight, each with its reason and original text.
1-min demoWe'll run Flight Operations Manual Search on your own manuals and a sample of the questions your crews really ask.