01 / CREW Ai
Flight and cabin crew on one roster
Captains, first officers, senior cabin crew and cabin crew share one monthly roster grid, with each day's flight, layover or day off. Synthetic demonstration data.
CREW Ai · Airline crew management system Build rosters the rules accept. Recover without starting over.
CREW Ai is an airline crew management system for building flight and cabin crew rosters within flight-time limitations, working out how many crew a season needs and comparing ranked recovery options when a disruption hits. A constraint solver builds each roster, and one flight-time engine checks every duty. Rosters go live only with every seat covered by qualified crew and a named administrator's sign-off, and recovery waits for a duty manager's approval.
Email me the brochure, CREW AiIn brief
What is CREW Ai?
CREW Ai is FlightAtom's airline crew management system for flight and cabin crew. A constraint solver builds rosters, a flight-time limitation engine for FAA Part 117, EASA Subpart FTL and the OTAR scheme checks every duty, and a roster is published only when seat coverage, qualification currency and a named approver's sign-off all pass. Duty managers approve disruption recovery options, and administrators can query the operation through an AI assistant or their own AI client.
- Category
- Airline crew management system
- Built for
- Crew planners, duty managers, workforce planners and flight and cabin crew
- Applied AI
- Read-only AI assistant, plain-English disruption intake and MCP for AI clients; a constraint solver, not a language model, builds rosters and recovery plans
- Availability
- Available · independent product
- Fits your operation
- As your operation grows or changes, we adapt CREW Ai to fit, and you approve each change before it goes live.
An example workflow · CREW Ai
From an optimized roster to the disruption decision.
A new period needs a roster, and the day of operation will change it. One flight-time engine judges the roster the solver builds, the version that is published and every recovery option.
- 01
Optimize the roster
A constraint solver assigns crew to every sector, replays each candidate through the flight-time engine against real duty history, and reports whether each stage was proved optimal.
- 02
Publish with sign-off
The roster goes live only when every seat is covered, every assignee holds current qualifications and a named administrator signs off. Otherwise the previous roster stays live.
- 03
Decide the disruption
A cancellation, retime or sick call opens a disruption. The recovery solver locks every unaffected duty and ranks options that pass the rules, each explained in plain English with a draft crew message, and the duty manager approves one.
The resulting record
A published roster and an applied recovery, each re-checked against the flight-time rules and kept with its approval in the revision ledger.
The human decision
A named administrator signs off publication, and the duty manager chooses the recovery option. The roster changes only when they approve.
How it works · CREW Ai
Every flight-time verdict comes from one engine.
One flight-time limitation engine sits at the centre of CREW Ai. The roster solver, the publish check, recovery, hand-entered standby and crew-demand simulation all take their legality verdict from it.
Flight-time limitation engineApplies FAA Part 117, EASA Subpart FTL and OTAR Appendix F, from duty periods and rest to standby and cumulative limits.
- Roster optimizerEach candidate roster goes to the engine; a duty it rejects becomes a cut and the solver searches on.
- Publication gateBefore a roster goes live, the engine checks the uncommitted candidate's duties; if they fail, nothing is committed.
- Disruption recoveryEvery recovery option must pass the engine, and the chosen option is checked again before it applies.
- Hand-entered standbyA standby entered by hand is replayed through the engine with the crew member's duty history and refused if it, or any later duty, breaks a limit.
- Crew-demand simulationRosters modelled to size the crew establishment are replayed through the engine, and minimum-crew search looks for the fewest legal lines.
Inside CREW Ai
See the work, from start to finish.
Explore the current product interface with demonstration data. Open any screenshot for a closer look.
02 / CREW Ai
Check duties against flight-time limits
A read-only check of the rostered duties from 12 to 17 October shows each duty's remaining margin on FDP, rest and the cumulative duty limits, plus flight-time limits for flight crew, and highlights the two tight rest margins. Synthetic demonstration data.
03 / CREW Ai
Ranked recovery options
A first officer's sick call for 22 and 23 October is logged on the Disruption Desk, and recovery proposes ranked options that change only the affected seat. Options 1 and 2 each name the off-duty crew member they would call and cannot be approved until the duty manager acknowledges the callout. Synthetic demonstration data.
04 / CREW Ai
Approve one option, with a crew message ready
The duty manager ticked the callout acknowledgement and approved option 2, so David Williams takes the first officer seat and the other options are superseded. Explain then gives a template summary and a suggested message for the crew member. Synthetic demonstration data.
05 / CREW Ai
Crew self-service in MyCrew
In the MyCrew portal a first officer sees flights and duty days for the next 30 days, qualification status and upcoming duties with local departure times, with links to bidding, pay, leave and safety reports. Synthetic demonstration data.
Capabilities
One set of rules, from first draft to sign-off.
Every roster decision takes its legality verdict from the same flight-time engine, so what planners build, publish and repair, and the duty records auditors receive, are judged against one set of rules.
Legal rosters without building pairings first
A constraint solver assigns flight and cabin crew to individual sectors, so duties take shape inside the solve instead of in a separate pairing step. Every candidate roster is replayed through the flight-time engine against each crew member's real duty history, and any illegal duty is cut before the search goes on. Seat coverage is settled before workload balance and roster stability, day-off bids are honoured as hard constraints, and each stage reports whether its result was proved optimal.
Recover a disruption without reshuffling the roster
A crew member's sick report, or a schedule-feed cancellation or retime of a crewed flight, opens a disruption on the Disruption Desk by itself. The recovery solver keeps every unaffected assignment locked, ranks options that pass the flight-time checks, including positioning and off-duty callouts, and shows which check ruled crew out of any seat it cannot fill. Nothing changes until a duty manager approves an option and confirms any off-duty callout, and CREW Ai re-validates the option before it applies.
Only a complete, legal, signed-off roster goes live
A roster is published only when one check passes in full: every seat filled, including a senior cabin crew member wherever cabin crew fly; every duty accepted by the flight-time engine; every assignee current on licence, medical, type and route qualifications; approved leave, planned training and manual duties respected; each crew member already at the station where the duty starts; confirmed crew records for every flight in the 84 days before the period; and source feeds fresh. A named administrator signs off, and if anything fails the previous roster stays live unchanged.
Know how close each duty is to its flight-time limits
Planners re-check the live roster or test a hypothetical duty at any time and see each duty's remaining margin on flight duty period, rest and cumulative limits, with the rule named for any breach. A single engine applies FAA Part 117, EASA Subpart FTL and Appendix F to OTAR Parts 121 and 135, covering flight duty periods, rest, standby, split duty, in-flight relief, days off and cumulative limits. The optimizer, publication, recovery, hand-entered standby, duty records and simulation all take their final verdict from it.
How many crew the season needs, legal and costed
Workforce planners size the crew establishment for the live schedule, a draft version or a copy retimed in the Itinerary Lab, and every modelled roster is replayed through the flight-time engine and costed at the operator's own salary, hotel, per diem and positioning rates. Minimum-crew search looks for the fewest legal roster lines, and season planning turns monthly lines into individual crew to give hiring numbers with and without reserve. Monte Carlo analysis shows how often a reserve level still leaves a flight uncovered, and the smallest extra standby per day that reaches a chosen service level.
Crew acknowledge, bid and report sick themselves
In the crew self-service portal and the CREW Mobile app for iPhone and iPad, crew see their roster and standby, acknowledge the roster and each change, bid for days off and trips, request leave, and view pay, qualifications and routes. Reporting unavailable opens a disruption on the Disruption Desk straight away. Planners get recovery approval, publish review, simulation and workforce tools in CREW Admin on iPhone and iPad and in CREW for Mac.
Fatigue reporting and screening
Crew file fatigue reports and commander's discretion records from the portal or the app, and discretion is kept as evidence that never relaxes a limit. For fatigue risk management, a published sleepiness model (McCauley et al.) screens rostered duties at group level from the rest the roster allows, and elevated results are flagged for qualified human review.
Verifiable duty records and audit trail
Schedule, roster, recovery, qualification and sign-off changes are written to an append-only ledger, chained with SHA-256 and protected from edits by the database. Administrators pull per-crew duty and flight-time records, with each duty's limit, margin, rest and rolling totals, and revision packs that can be verified offline, through the API as CSV or JSON.
Pay and credit to the cent
A versioned pay and credit engine calculates block-or-better credit, duty and trip rigs, reserve and positioning credit, monthly guarantees, overtime and per diem with decimal arithmetic, so a period reprices to the same cent every time. It reads the published roster and never changes it.
Applied AI
Ask in plain English. The solver and the rules decide.
Language models help where planners lose time: finding facts across the operation, writing up a disruption and explaining a plan. Rosters and recovery options come from the constraint solver and the flight-time engine, and nothing an AI client requests skips the checks or sign-off a planner would face.
An assistant that shows its working
Administrators ask about rosters, schedules, disruptions or legality, and the assistant answers by calling read-only CREW tools, listing each one it consulted. Ask who could cover a seat and it runs the checks recovery uses, then says why every other crew member is ruled out. Crew names and identifiers are replaced with aliases before anything reaches the model provider.
Plain English at the Disruption Desk
A duty manager can describe a disruption in plain English and get a structured entry to edit and confirm, then ask for any solver plan to be explained with a draft message to crew. Sick calls and simulation requests typed into chat are recognised by rules, not the model, and become proposal cards that do nothing until a person confirms them.
Connect your own AI client
A built-in Model Context Protocol (MCP) server lets administrators connect an AI client such as Claude over OAuth to read rosters and schedules, run simulations and check legality. Publishing or applying a recovery needs a write scope an administrator explicitly grants, and runs through the same checks and sign-off. Crew appear by the operator's crew code rather than by name.
Evaluate CREW Ai
Make the next step measurable.
Agree a representative sample and a baseline from your current process. These are proposed evaluation measures, not published performance results.
Bring a representative example
A representative roster, current constraints, and a disruption scenario.
Inspect the complete workflow
Compare proposals with the existing roster and run operational validation.
Compare with your baseline
Planning time and solver runtime; uncovered duties; constraint violations; roster cost.
Questions & answers
Before you book a demo.
What does CREW Ai cover as a crew management system?
CREW Ai covers schedules and draft timetables, rostering and bidding, flight-time limitation checks, qualification currency, leave and training planning, roster publication with sign-off, disruption recovery, fatigue reporting, crew-demand and season planning with crew cost, pay and credit, and duty records for audit. Planners work in the operations console and the CREW Admin apps, and crew use the self-service portal and the CREW Mobile app for their roster, bids, leave, pay and reports.
How does CREW Ai make sure a roster is legal?
A constraint solver builds the roster and replays every candidate through a flight-time limitation engine for FAA Part 117, EASA Subpart FTL and OTAR Appendix F, using each crew member's real duty history, and cuts any illegal duty before searching on. Before publication, one all-or-nothing check confirms seat coverage, qualification currency, leave and training, crew location, prior crew history and feed freshness, and a named administrator signs off. If any check fails, the previous roster stays live.
What happens in CREW Ai when a crew member calls in sick?
The crew member reports unavailability in the portal or app, which opens a disruption on the Disruption Desk automatically. The recovery solver keeps every unaffected assignment locked and ranks options that pass the flight-time checks, including positioning and off-duty callouts where allowed. A duty manager reviews the options and approves one, and CREW Ai re-validates it before applying it to the roster.
Can CREW Ai check a roster built in another system?
Yes. CREW Ai can import the crew roster held in the connected flight-operations system as a mirror period and check every duty against the flight-time rules, so the Disruption Desk, recovery and compliance reports work on periods CREW Ai did not publish. Schedules arrive from the flight-operations feed, by CSV upload or as seasonal IATA SSIM files, and qualification validities sync from the training-records system.
Can CREW Ai tell us how many crew we need for a season?
Yes. Crew-demand simulation sizes the establishment for a live or draft schedule with a choice of eight construction methods, and every modelled roster is replayed through the flight-time engine. Minimum-crew search looks for the fewest legal roster lines, season planning chains monthly lines into individual crew to give hiring numbers with and without reserve, and workforce planning shows how many leave slots each day can grant without breaking coverage.
Does CREW Ai plan crew leave and training?
Yes. Crew request leave in the portal or app, and administrators approve it and plan recurrent, type-rating, captain-upgrade, line and new-hire training. CREW Ai refuses a training event that clashes with the trainee's or instructor's approved leave, other training or hand-entered roster entries. The roster solver treats approved leave and planned training as fixed days off, and workforce planning projects crew supply against demand month by month.
Does CREW Ai calculate crew pay and credit?
Yes. A versioned pay and credit engine calculates block-or-better credit, duty and trip rigs, reserve and positioning credit, monthly guarantees, overtime and per diem from the published roster. It uses decimal arithmetic with a single rounding step, so a period reprices to the same cent every time.
What records can CREW Ai export for a regulator?
Administrators export per-crew duty and flight-time records through CREW Ai's API as CSV or JSON, with each duty's flight duty period against its limit, the rest before it, rolling totals and any violation. Schedule, roster, recovery, qualification and sign-off changes sit in an append-only, SHA-256 hash-chained ledger that the database protects from edits, and revision packs can be verified offline.
Can AI assistants such as Claude connect to CREW Ai?
Yes. A built-in Model Context Protocol (MCP) server lets administrators connect an AI client such as Claude over OAuth to read rosters and schedules, run simulations and check legality. Publishing a roster or applying a recovery needs a write scope an administrator explicitly grants, and passes the same checks and sign-off as a planner. CREW Ai's own in-app assistant can only read and simulate, so it cannot change the roster.
What happens to CREW Ai when our operation grows or changes?
As your fleet, routes, rules or teams change, we adapt CREW Ai to match, and you approve each change before it goes live.
See CREW Ai in action
Request a demo Start with your workflow.
A focused walkthrough, with the people and decisions that matter to your team.