FLIGHT Ai is flight operations analytics software for the operations teams of airlines and aircraft operators: chase missing return documents, catch weight and fuel mismatches between the load sheet, NavLog, fuel slip, logbook and the recorded flight, and tune each aircraft's fuel planning bias from measured burn. It syncs flights, crew and files from ForeFlight Dispatch, and document AI reads crew uploads, including phone photos and handwritten logbook times. Cross-checks are rule-based, with each value labelled by its source document.
Email me the brochure, FLIGHT Ai
FLIGHT Ai is FlightAtom's flight operations analytics software for airlines and aircraft operators that plan with ForeFlight Dispatch. It synchronizes flights, crew and attached documents, uses AI to classify and read return documents such as load sheets, performance cards, fuel slips and logbooks, and cross-checks weights and fuel between those documents and the recorded flight. It measures actual fuel burn from the signed NavLog, recommends a fuel planning bias for each aircraft, analyses schedule performance, and uses trained models to predict fuel burn and flag unusual flights with the reasons behind each flag.
Document AI reads the return documents and its readings feed the rule-based checks directly; trained models predict from the operation's own flight history. People act on what is flagged.
Explore the current product interface with demonstration data. Open any screenshot for a closer look.
Return flights to the KJFK base that are missing required documents, with each required type marked present or missing, the uploaded photos and their document type, the captain and first officer, and a Send email reminder. Synthetic demonstration data.
The Compare data view lines up weights, centre of gravity, fuel and passengers from the final load sheet, performance data, NavLog, fuel slip, logbook and the recorded flight, and flags a 380 kg zero fuel weight difference. Synthetic demonstration data.
Uploads that the document AI left as General wait in a review queue, each with a preview and a one-click tag for every required document type, and a person's tag overrides the AI. Synthetic demonstration data.
Planned and measured burn on the same 180 flights over 30 days, the flights that reached the gate below alternate plus reserve fuel with their crews, and daily fuel per nautical mile and fleet bias trends. Synthetic demonstration data.
Each aircraft gets a rule-based recommendation to plan more fuel, less fuel or keep planning as it is, using a 1% band against its measured burn, and the same check over the past year broken down by cruise speed profile. Synthetic demonstration data.
Status of the fuel-burn, delay-risk and anomaly models trained on the flight history: the fuel model's backtest error, with a badge showing it beats the flight plan's own error on the same flights, and the delay model's AUC and base rate. Below, the predicted burn for upcoming flights sits beside the flight-plan figure. Synthetic demonstration data.
The delay model ranks the next three days' flights by their probability of departing more than 15 minutes late, beside plain per-station taxi-time percentiles, and below them the anomaly model lists recent flights with the values that stood out against typical ones. Synthetic demonstration data.
FLIGHT Ai keeps itself up to date. On a short, configurable cycle it syncs flights, crew and files from ForeFlight Dispatch, skips flights that have not changed, parses new NavLogs, and sorts and reads new documents, so the checks and analytics are ready when someone opens them.
Each flight returning to base is checked against the required return-document set: load sheet, NOTOC, loading instructions, takeoff and landing performance, NavLog, logbook, security search and fuel slip. One click emails the captain and first officer a list of what is missing and records the reminder. A monthly trend shows whether compliance is improving or worsening, and a pilot view shows who has gaps.
A takeoff weight on the performance card that disagrees with the load sheet, or a fuel slip uplift that disagrees with the logbook, is flagged on the flight it belongs to. Compare data lines up zero fuel, takeoff and landing weight, centre of gravity, block, takeoff and trip fuel, uplift, arrival fuel and passengers across the final load sheet edition, performance cards, NavLog, fuel slip, logbook and the recorded flight. Each figure has its own tolerance, such as 0.3% MAC for centre of gravity, and every value is labelled with the document it came from.
Document AI sorts each upload, including angled phone photos, scans and PDFs, into its document type for the missing-document check. It reads the weights, centre of gravity and fuel figures that Compare data lines up, along with V-speeds, flex temperature, hold loads and handwritten logbook block times. A failed model request leaves the document queued for retry rather than mislabelled, and every captured document is kept with preview and download.
Actual burn comes from the fuel and OUT, OFF, ON and IN times on each flight's signed NavLog and is compared with planned trip fuel from the same flight plan, by aircraft, route and cruise speed profile. Each aircraft with at least ten measured flights gets a recommendation: raise its bias when burn runs more than 1% above plan, lower it when burn runs more than 1% below, otherwise keep it. Taxi-out fuel is reported by station and aircraft type.
A low-fuel arrivals list shows past flights whose fuel at the gate, read from the signed NavLog, was below planned alternate plus reserve fuel, with flights below reserve fuel alone marked critical and listed first. Each row shows the route, date, captain, first officer, fuel in, required fuel and margin, and when there are none the page states how many flights were checked.
Delay propagation shows how departure delays carry into arrivals, alongside time made up between departure and arrival, turn times by station and day-of-week patterns, all measured only on flights with complete OUT, OFF, ON and IN times. Block time is measured against the signed flight plan, and departures and arrivals against schedule at 15 minutes. Connected to the operator's crew and operations system, FLIGHT Ai uses that system's scheduled block times as the on-time baseline.
Crew pages show each pilot's position, flights, hours and destinations, compare monthly flying with the average for the same position, and include a printable monthly duty and rest report from imported duty records. Passenger analytics report load factor on flights with known loads by route, aircraft, day and hour. A fleet fuel analysis PDF with a page for each aircraft, and CSV exports, carry the figures into reports.
A live map shows the fleet's positions, flown tracks, ETAs and the crew on board. Upcoming flights carry METAR and TAF for the destination and alternate, with the TAF period at the ETA highlighted. Network weather adds SIGMET, G-AIRMET and PIREP layers from NOAA's Aviation Weather Center, and an ACARS inbox lets users acknowledge messages and reply.
An MCP endpoint lets an authorized AI assistant run read-only queries on the flight database, preview documents in the review queue and tag them. Each query is a single read-only statement with a time limit and a row cap, and every tag the assistant applies requires a written reason.
FLIGHT Ai uses AI where rules cannot decide: sorting and reading varied paperwork, and predicting from flight history. NavLog parsing, cross-checks, fuel bias and schedule analytics are deterministic calculations you can trace to the flights behind them.
Document AI handles phone photos, scans and PDFs and is instructed to leave out any value it cannot read clearly. Documents it cannot match to a required type go to a review queue, where a person previews and tags each one and the time of the tag is recorded. The photos most recently tagged by hand become reference examples for later classifications.
A model trained on the operation's own history predicts fuel burn for the coming week's flights, with an error band and the difference from the plan, and its error on the newest flights it did not train on is compared with the flight plan's own error on those flights. A second model ranks upcoming flights by their risk of departing more than 15 minutes late, beside its backtest ranking score (AUC) and the base late rate. The models retrain automatically about once a day, and a model status view shows each one's current version, its backtest metrics and whether the fuel model beats the plan.
An anomaly model scores recent flights on fuel burn per nautical mile and per hour, taxi-out and taxi-in time, departure delay and passenger load. Each flagged flight lists the features that stood out beside their typical values. Where plain statistics answer the question, FLIGHT Ai uses them: taxi allowances are the median, 75th and 90th percentile of observed taxi times at each station, and route profiles place each route pair above or below the median across routes for fuel per nautical mile and for load factor.
Agree a representative sample and a baseline from your current process. These are proposed evaluation measures, not published performance results.
Representative return documents and a held-out period of flight records.
Check extracted fields against reviewed values and forecasts against a simple baseline.
Field accuracy; missing-data coverage; forecast error; reviewer correction time.
Yes. FLIGHT Ai synchronizes flights, aircraft, crew and attached files from ForeFlight Dispatch on a short, configurable cycle, then parses NavLogs and sorts and reads new return documents automatically, so its checks and analytics stay current. It keeps its own database, dashboard and accounts, and it does not require AURORA Ai or any other FlightAtom product.
Each flight returning to the operator's base is checked against the required return-document set. Document AI sorts each upload, including phone photos, scans and PDFs, by type and reads its fields. A Compare data view then lines up weights, centre of gravity, fuel, uplift and passengers across the documents and the recorded flight and flags differences beyond tolerance. A user can email the captain and first officer a list of anything missing.
From the fuel readings and OUT, OFF, ON and IN times recorded on each flight's signed NavLog. Kilograms and pounds are reconciled and physically impossible readings are discarded. Planned trip fuel comes from the same signed flight plan, so plan and actual are compared on the same flights.
Document AI classifies return documents and reads their fields. Trained machine-learning models predict fuel burn, rank upcoming flights by late-departure risk and flag unusual flights with the features that stood out. NavLog parsing, document cross-checks, fuel bias recommendations, low-fuel arrivals, schedule analytics, route profiles and taxi allowances are deterministic calculations.
The fuel and late-departure models are each tested on recent flights they did not train on. The fuel model's error is compared with the flight plan's own error on those same flights, and the late-departure model shows its backtest ranking score (AUC) and the base late rate. Models retrain automatically about once a day, and a model status view shows each model's current version and backtest metrics. Accuracy is measured on each operator's own flights rather than quoted as a general figure.
Yes. FLIGHT Ai has an MCP (Model Context Protocol) endpoint. An authorized AI assistant can run read-only queries on the flight database and help work the document review queue. Each query is a single read-only statement with a time limit and a row cap, and every tag the assistant applies requires a written reason.
No. FLIGHT Ai provides operational records, analysis, document review, and decision support. Authorized personnel retain responsibility for approved planning systems, release decisions, and required calculations.
No. FLIGHT Ai is operations software for airlines and aircraft operators. It analyses flight records, fuel, schedule performance, and return documents for operations teams. It does not search for or sell passenger flights.
As your fleet, routes, rules or teams change, we adapt FLIGHT Ai to match, and you approve each change before it goes live.
A focused walkthrough, with the people and decisions that matter to your team.