A reliability program where every alert reaches the board with its cause, a fix and a 90-day test — and you work it by asking the agents
Every breach reaches the board with evidence and a fix; every fix is checked at 90 days.
See one case, screen by screen ↓An aircraft reliability program, often called a maintenance reliability program, is how an airline watches whether its maintenance program works: it tracks PIREP rates, removals and delays by fleet and ATA chapter against alert levels, investigates breaches, and takes corrective action through a Reliability Review Board. Aircraft Reliability Program is reliability program software in which seven agents run the monthly analysis, find the cause behind each alert with cited evidence, draft the board's agenda and corrective actions, and check each fix at 90 days, while the board decides and the Director of Quality signs.
How to choose AI for airlines: a buyer's guide to agentic AI solutions →
Updated
An aircraft reliability program is how a Part 121 carrier shows its maintenance program works: the continuing analysis and surveillance system (CASS) of 14 CFR 121.373 and, for reliability-controlled intervals, the reliability program methods of AC 120-17B. Every month a reliability engineer pulls PIREPs, mechanic reports, unscheduled removals, shop teardown reports, technical delays, deferrals and flight hours into a workbook, computes hundreds of rates by fleet and ATA chapter against their alert levels, and builds the board pack in slides by hand.
The risk sits in what the spreadsheet cannot see. A pack fault coded to the wrong chapter hides a valve problem in another line. An alert says ATA 21-51 is high on the A321neo, not that the valves come from one serial lot. An alert that is really one tail sends the fleet on an inspection it does not need. And last quarter's corrective action is closed when the work is done, not when the rate comes down — so nobody knows whether the fix worked.
Estimated split of about nine working days from data close to a board-ready agenda, by spreadsheet and with the solution.
Seven agents run the month's analysis, find the cause behind each alert, check each other and draft the board's agenda; the board decides.
Corvane Airways flies 300 aircraft in ten fleets. September's data closed at 23:58 on Tuesday, October 13, and the Reliability Review Board meets on Tuesday, October 20. Here is how the reliability engineer works the month with the agents — what they did on their own, what she asked, what the board decided — screen by screen, in the working solution.
Lena opens a home screen with three numbers: 7 items on October's agenda, technical dispatch reliability of 99.12% for September with the A321neo at 98.71% against its 99.05% target, and 1 of last quarter's 2 actions that worked. Under them, four things that need her, in order, and three notes from the agents' overnight run, each with its citation and a next step.
Board secretary agent: “From September's 2,418 pilot reports and 1,312 removals I drafted October's agenda: 4 alert breaches, 2 check-backs, 1 interval change.”
At data close the data agent collected September from five systems: 2,418 PIREPs, 1,176 mechanic reports, 1,312 removals, 391 shop reports, 462 delays and 21 cancellations. It proposed 61 ATA re-codes from the report text, merged 27 duplicate write-ups and filled 3 days of A220 flight hours from aircraft OOOI messages. Then the analysis agent computed 312 lines and 1,148 part numbers, the check agent recomputed every one, the evidence and corrective action agents worked the breaches, and the board secretary agent had the agenda and the FAA report drafted by 05:00. Lena approved the 8 re-codes that touched agenda lines, with a reason code; 53 wait for her.
“People in the loop: re-codes need the reliability engineer; the agenda goes to members only when the secretary sends it; nothing reaches the FAA without the Director of Quality's signature.”
Item 1 reads as one plain sentence. Pilot reports on the ATA 21-51 pack flow control valve reached 0.84 per 1,000 flight hours, against an alert level of 0.62 — the 2025 mean of 0.38 plus two standard deviations. It is the second month above the alert level and the first above the 0.74 upper limit. A 21-month chart shows the rate against the mean, alert and upper lines, and the agents' notes sit beside it.
Evidence agent: “Lot FC-24 explains the alert: 26 of 30 reports and 8 of 11 removals; 6 of 7 teardowns found the same shaft-bearing wear.”
Under the chart the evidence is folded: 30 pilot reports, 11 unscheduled removals with a valve MTBUR (mean time between unscheduled removals) of 7,870 FH against a 9,000 FH floor, 7 shop findings and 16 delays and deferrals — 9 delays of 486 minutes, a cancellation and 6 one-pack deferrals. Lena opens teardown report TR-26-0909 for valve FC24-0603, removed from N441CV: butterfly shaft bearing wear, early stage, with intermittent binding below −20 °C.
The evidence agent never states a cause without a quoted source.
The corrective action agent drafted three options: replace the FC-24 valve on the 9 tails that still carry one, under warranty and vendor bulletin 3291478-21-012; replace on failure only; or replace all 88 valves. Lena asks the agents why 9. The evidence agent answers that lot FC-24 filled 17 valve positions on 14 tails, 8 have already been removed, and the other 79 valves run at 0.09 per 1,000 FH, below the 2025 mean.
“Answers cite their source; nothing is approved or sent until a person says so.”
Victor wants serial numbers, not an alert. The full analysis shows how the rate was computed — 12 reports on 14,286 flight hours — and the check agent's own recompute of 0.840. By serial lot, 26 of the 30 reports fall on FC-24, which filled only 17 of the 88 positions. The 9 tails in the proposal are listed, with the three that had two September reports each marked.
The check agent recomputed all 312 lines from the raw tech log without seeing the first answer, and agreed on 311. On the last it challenged the draft: the 737-800 nose-wheel steering rate of 0.54 against an alert level of 0.45 comes from 5 of 9 reports on one tail, N517CV, “shimmy on taxi”. Without that tail the fleet is at 0.24. Version 1 of the agenda went to members with the challenge open; on Thursday morning Grace Liang confirmed N517CV is a repeat defect, and the corrective action agent replaced the fleet inspection with a tail-level action, keeping the first draft as option B.
The check agent never edits a draft itself; a person confirms its challenge before the proposal changes.
Every action carries a success test written when the board approved it. The 777-300ER bleed valve fix met its test: 0.38 in August and 0.32 in September, below the 0.58 alert level. The A320 forward cargo door fix did not: 0.47 against 0.30. The seals were serviceable; the sensor bracket was loose on 8 of 11 doors. The agent escalated it to Marcus Bell, the Director of Quality, and drafted a second-level action.
The check-back agent never closes an action; a failed one goes to the Director of Quality.
Members pre-voted. Victor: “Agree with A. Arden confirmed exchange stock for 9 units.” Grace Liang, Director of Line Maintenance, asked for the three tails with two September reports first. In the meeting, with the chair and 4 voting members present for a quorum, Lena records the decision: approved with a change — N431CV, N433CV and N438CV within 14 days, the other 6 within 60 — a 90-day check-back at the January 19, 2027 board, owner Victor Osei, reason code D2, and her e-signature.
The board secretary agent never records a decision a person did not make.
Seven decisions in one meeting. The board secretary agent writes the minutes — attendance, quorum, each decision with its reason code and action — and creates CA-2610-01 for Victor Osei, due December 14, 2026, with the 9 tails and proposed DEN overnights in Elena Kowalski's planning queue; N431CV, N433CV and N438CV are marked within 14 days. The monthly reliability report for the FAA principal maintenance inspector moves to version 3 with the board's decisions, for Marcus Bell to sign.
Nothing reaches the FAA without the Director of Quality's signature.
The same month of the reliability program, seen by four of the people who carry it — what it looked like by spreadsheet, and what it looks like now.
Seven agents run the month's analysis, find the cause behind each alert, check each other and draft the board's agenda; the board decides.
Pulls the month's PIREPs, mechanic reports, removals, teardown reports, delays and flight hours from the M&E, ops and utilization systems; re-codes miscoded reports from their text; merges duplicates; fills gaps from aircraft messages.
Computes rates per 1,000 flight hours by fleet and ATA chapter-section, component MTBUR and technical dispatch reliability, and compares each with its alert level by the program's rules.
For each breach, reads the pilot reports, removal records, teardown PDFs, delays and deferrals, finds the pattern — a serial lot, a tail, a station, a flight phase — and cites every passage.
Drafts options with cost, owner, window and a measurable success test, from vendor bulletins, past actions, warranty terms and the planning calendar.
Recomputes every number on its own from the raw records and challenges weak items — an alert that is one tail, a sample too small to call a trend.
Measures each approved action against its success test at its check-back and proposes close, extend or escalate.
Writes the agenda, the minutes and the actions from the recorded decisions, the monthly reliability report for the FAA and interval-change letters.
The engineer approves re-codes; members pre-vote and decide; the chair signs the minutes and the FAA report. The agents propose; a named person decides.
Beside every agenda item and every screen, ask about the month, a line or a proposal. Every answer cites its source, and nothing is approved or sent until a person says so.
Why 9 tails and not all 44?
Because the problem is one serial lot. Lot FC-24 was fitted in 17 valve positions on 14 tails; 8 of those valves have already been removed, so 9 remain, one each on N427CV, N429CV, N431CV, N433CV, N436CV, N438CV, N440CV, N441CV and N443CV. The other 79 valves run at 0.09 per 1,000 FH, below the 2025 mean.
What did the teardowns find?
6 of the 7 returned valves had butterfly shaft bearing wear with binding below −20 °C; one was no fault found. That matches the pattern in the reports: most faults in climb or cruise, when the valve is cold.
What does it cost us if we wait?
In Q3 this line caused 9 delays, 486 minutes and one cancellation — about $48,600 in delay minutes at roughly $100 a minute, before the cancellation — and six one-pack deferrals that cap the flight at FL315. The exchange itself is free under warranty.
Draft the change Grace asked for
Done: option A now starts with N431CV, N433CV and N438CV within 14 days (first overnights Oct 22, 23 and 26), then the other 6 within 60 days. The success test is unchanged. It stays a draft until the board decides.
13 screens from the working solution, on its sample data. Pick one to see it large.
The alert rules, the board and the agents' modes are settings. A change to an alert rule changes every line, so it is a proposal the chair signs, kept with its reason.
| Setting | Default | Choose from |
|---|---|---|
| Alert level | Previous calendar year's mean plus 2 standard deviations | 1.5 · 2 · 2.5 · 3 standard deviations |
| Upper limit (one month above it is an alert) | 3 standard deviations | 2.5 · 3 standard deviations |
| Consecutive months above the alert level | 2 | 1 · 2 · 3 |
| Minimum sample, below which a line is flagged small | 5 events | A number of events you set |
| Fleets with under 12 months of data | Provisional level from the previous operator's records | Provisional level from the previous operator's records · No alert level until 12 months |
| Component MTBUR floor | Previous year's MTBUR minus 2σ, or the vendor's guaranteed MTBUR if higher | Set per part number |
| Check-back, chosen when the board decides | 90 days after approval | 60 · 90 · 120 days |
| Agent mode | Propose only for the corrective action, check, check-back and board secretary agents | Acts, then reports · Propose only — set per agent |
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 · “typical” = published figures (Airlines for America, U.S. passenger carrier delay costs — $100.76 per aircraft block minute in 2024). Regulations named as context: 14 CFR 121.373 (continuing analysis and surveillance) and FAA AC 120-17B (reliability program methods). People, the airline Corvane Airways, the vendor Arden Flow Systems, lot FC-24, bulletins, tails and records named on this page are fictional — characters and sample data in the working solution.
An aircraft reliability program, often called a maintenance reliability program, is how an airline watches whether its maintenance program works: it tracks PIREP rates, removals and delays by fleet and ATA chapter against alert levels, investigates breaches, and takes corrective action through a Reliability Review Board. Aircraft Reliability Program is reliability program software in which seven agents run the monthly analysis, find the cause behind each alert with cited evidence, draft the board's agenda and corrective actions, and check each fix at 90 days, while the board decides and the Director of Quality signs.
In the working solution the alert level for each line is the previous calendar year's mean monthly rate plus two standard deviations, and the upper limit is the mean plus three. A line is in alert after two consecutive months above the alert level, or one month above the upper limit. The method, the sigma, the months and the minimum sample are settings, and a change is a proposal the chair signs.
The evidence agent reads the pilot report text, removal records, vendor teardown reports, delays and deferrals behind the breach and breaks them down by serial lot, tail, station and flight phase. In the demo it traced 26 of 30 A321neo pack-valve reports to one serial lot, with every passage cited. It never states a cause without a quoted source.
Every action gets a measurable success test when the board approves it, such as a rate below the alert level for two months. The check-back agent measures it at the check-back board, 90 days by default, and proposes close, extend or escalate. An action that did not work goes to the Director of Quality with a second-level action drafted.
Yes, on evidence. The board secretary agent packages an interval change with the task findings — in the demo 214 accomplishments over 24 months with none beyond limits — and the chapter rate against its alert level, and the check agent adds its risk notes. The board recommends, the Director of Quality approves, and under the airline's program manual the FAA principal maintenance inspector is notified before it takes effect.
It drafts it. The report covers technical dispatch reliability by fleet, lines in alert and the board's actions, check-backs, interval changes, data quality and program changes, with every number linked to its record and a new version after the board's decisions. The Director of Quality signs and sends it; the agents never do.
Yes. The agents propose; the reliability engineer approves re-codes, board members pre-vote and decide, the secretary records each decision with a reason code and an e-signature, and the chair signs. The agents never change a source record, set an alert level, close an action, book a slot, order a part or send anything to the FAA.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your Reliability Program Manual, alert levels and a past month of PIREPs, removals and teardown reports — and it goes live once every quality gate has passed. We will walk you through it on your own fleet's data first.
No tail flies past an AD, and every line of the status report is proven.
1-min demoService Bulletin EvaluationEvery bulletin decided in days on the fleet’s own evidence, every tail and dollar cited.
1-min demoDeferred Defect ManagementNo aircraft flies past its MEL limit, and every fix lands on a planned overnight.
1-min demoWe'll run Aircraft Reliability Program on a past month of your own PIREPs, removals and teardown reports.