Airline network planning that tests every route on demand, feed, fleet, rivals and slots before it reaches the plan
Every route tested on demand, feed, fleet, rivals and slots before it reaches the plan.
See one case, screen by screen ↓Airline network planning is deciding what an airline flies: which cities, how often, on which aircraft and in which bank of the hub. Airline Network Planning is network planning software in which eight agents answer every route what-if on the network map — demand, connecting feed, fleet and heavy checks, rival response, slots and traffic rights, partner flows and route economics — each with its sources, while a Review agent re-checks them blind. Planners compare versions and promote one with a reason code, and the SVP Network Planning signs the season plan version.
How to choose AI for airlines: a buyer's guide to agentic AI solutions →
Updated
In airline network planning, network planners decide what an airline flies: which cities, how often, on which aircraft and in which bank of the hub. Today a new route evaluation is a relay between teams. The planner frames the idea, demand forecasting sizes the market from the DOT ticket sample, fleet planning looks for a spare tail, maintenance planning checks the heavy-check calendar, the slot team asks the coordinator, alliances add partner flows and finance prices the route. Each hand-off waits, and the SVP sees one version, not three.
The pieces disagree without anyone seeing it. The demand forecast quietly assumes no rival response, while the competitive view lives in someone else's deck. The fleet check comes after the route is promoted, so the missing widebody turns up in June, when the fix is a wet lease, a late start or a schedule change after tickets are sold — and under 14 CFR Part 260 a significant schedule change gives passengers an automatic refund if they decline it. Six months later nobody can say which ticket-sample quarter or fuel price drove the decision.
Estimated split for one new-route study at a network carrier, in working days, as a relay between teams and with the solution.
Eight agents answer every what-if on the map, disagree in the open and check each other; planners promote, the SVP signs.
Corvane Airways is building its summer 2027 network (the S27 IATA season) from its ORD hub. On Friday, October 15, 2026, a week before the SVP signs plan version S27 v4, Elena Brandt wants to test a new route: ORD–ATH, four times weekly, on a 787-9. Here is the study, screen by screen, in the working solution.
Today shows 11 changes in plan S27 v4 awaiting signature, 3 what-ifs re-run overnight after the Q2 ticket sample, 2 disagreements to look at and 7 days to sign-off. Overnight, the agents also logged that Northmere loaded a second daily JFK–ATH from June 4, 2027, and that Aurelle Air cut one-stop Chicago–Athens fares 8% for summer. The partner feed is 2 days old, and the Partner agent has marked its answers until it is current.
Review agent: “Overnight I re-ran 3 what-ifs on the new ticket sample. W-32 ORD–ATH needs you: a June aircraft conflict and one disagreement.”
In the what-if bar she types “ORD–ATH four times weekly next summer on a 787-9”. A dashed arc draws from ORD to Athens and each agent pins its answer to it: Demand $38.4M at 82%, feed from 23 cities, Fleet June short, Rivals may match at 76%, Slots to request, Partner +$2.3M beyond ATH, contribution $5.1M at 13.3%. The scenario strip under the map holds the weekly frequency, the aircraft and the start date.
“Eight agents answer on the map, each with its sources, and they tell you when they disagree. Only you can promote a scenario to the plan.”
The Demand agent forecasts 82% with no rival response; the Competition agent disagrees. The Review agent sets both side by side: 6 points apart, wider than Corvane's 4-point band. Elena asks, “What if Northmere matches?” The Competition agent answers that with 3x weekly from July 2027, version A drops to 76% and $35.2M, contribution $1.9M (5.5%), below the 8% margin floor — and that its basis is a pattern, not an announcement: Northmere matched 3 of Corvane's last 5 new ORD long-haul routes.
Review agent: “What would settle it: Northmere's summer 2027 ORD schedule load, expected with its January release.”
Fleet fit draws the 34 787-9s by week against the lines of flying, with one operational spare kept at ORD. From June 3 to June 28, N812CV is in heavy check and ORD–ATH would take the spare: 3 weeks short in June. The agent lists three fixes, cheapest first — start July 1 (version B, contribution $6.2M), use the ORD–DUB aircraft in June and let that up-gauge slip to July 1 (about $0.4M), or wet-lease a widebody with crew for June 3–28 (about $1.6M).
Fleet agent: “I never assign tails; Tom decides.”
Hub banks draws ORD's eight connecting banks as waves of arrivals and departures. CV 892 to ATH closes bank B6 at 17:55 and catches the Midwest and West Coast arrivals between 15:50 and 16:40: 23 spoke cities with at least 75 minutes to connect, and a 46% connecting share. The return, CV 893, arrives at 15:20 and reaches 27 onward cities. B5 would lose the West Coast and B7 some Midwest cities.
Elena clicks Jul 1 on the scenario strip. The Review agent re-runs 4 of 8 agents for version B — Demand, Fleet, Economics and Review — and says the others' answers still hold. Version B: 31,416 passengers at 85%, $33.6M revenue, $6.2M route contribution (18.5%), $5.2M network contribution, and a free 787-9 every week from July 1 with the operational spare kept. If Northmere matches, B holds 78% and $30.2M.
Economics agent: “Version B: $33.6M revenue at 85%; contribution $6.2M (18.5%), up $1.1M on A.”
On the what-if's Agents tab each agent shows the steps it ran, the sources it read and its confidence. The Demand agent read the DOT ticket sample (10% of tickets) for Q2 2026, Corvane's bookings and one-stop history and fare shopping for Chicago–Athens, at 0.78 confidence; the Connection agent read the S27 schedule draft and DOT segment traffic, at 0.88. Each agent states what it never does and whom it hands to, and Replay run walks the steps again.
Compare puts A (as asked), B (start July 1) and C (3x weekly) side by side on flights, seats, load factor, passengers, revenue, operating cost, route contribution, what each takes from Corvane's own flights, network contribution and fleet. B is best on route contribution and network contribution and fits the fleet; A is short in June; C is tight. Under the table, the Economics and Review agents explain why: B earns $1.1M more contribution than A on $4.8M less revenue, because June is the softest summer month.
Review agent: “All three versions are re-derived blind; revenue and cost match the agents within 1.4%. My read: B, with the disagreement kept as the downside.”
Promote B opens the decision: ORD–ATH 4x weekly on a 787-9 from July 1, revenue $33.6M, contribution $6.2M, load 85%, fleet fits. A warning keeps the open disagreement in view — Demand 85% against Competition 78% if Northmere matches — and says it goes into the brief with both sources. Elena picks the reason code Demand and margin above threshold, adds “Watch Northmere's January load; re-run on release”, ticks the attestation and signs as Elena Brandt, logged to the audit trail.
“I reviewed the agents' sources and the open disagreement.” Only a Network Planner can promote.
Sign-off shows plan S27 v4 with 12 changes since v3 and $15.3M of route contribution, submitted by Elena Brandt and due October 22. The Review agent's brief: every change re-derived blind within 2%, one open disagreement on ORD–ATH kept as its downside, and no change breaks the operational-spare policy. An exception for W-31 ORD–NRT, 7 to 10 weekly, sits beside it, escalated after 3 days. Marcus can sign S27 v4 or send it back, each with a reason.
“Only the SVP signs; every signature is logged.”
The same route idea, seen by five of the people who carry it through route network planning: what the study looked like, and what it looks like now.
Eight agents answer every what-if on the map, disagree in the open and check each other; planners promote, the SVP signs.
Sizes O&D demand for the market and Corvane's share and forecasts passengers, load factor and revenue by month, from the DOT ticket sample, segment traffic, Corvane's bookings, fare history and partner sales.
Finds the spoke cities that feed the flight through its hub bank and the connecting share, from the schedule draft, the bank structure, minimum connect times and customs windows.
Checks tails by week against the lines of flying and the heavy-check calendar, with range and payload, finds the aircraft a route would take and proposes fixes, cheapest first.
Reads rivals' published schedules, fares and launch history and estimates a response and its effect on load, revenue and contribution.
Checks slot coordination, curfews, traffic rights and the customs hall from the slot calendar, coordinator rules and air services agreements, and drafts the slot request.
Counts partner flows beyond the destination, values a codeshare and drafts the request for a DOT codeshare statement of authorization.
Prices the route line by line — fuel, crew, maintenance, fees, ownership and handling — and its contribution after what it takes from Corvane's own flights.
Re-derives revenue and cost blind from all the inputs, names disagreements and what would settle them, and writes the brief for the planner and the SVP.
Promote with a reason code; confirm the fleet; approve partner drafts; sign the plan version. The agents propose; a named person decides.
The assistant answers across the solution with its sources, and acts: it opens the screen and runs the change. Every figure in an answer links to the dataset and the passage it came from.
Why did the Fleet agent flag ORD–ATH?
Because a 787-9 is missing in June. N812CV is in heavy check June 2–28, and ORD–ATH from June 3 would take the operational spare kept at ORD. Starting July 1 (version B) clears it and raises contribution to $6.2M.
What changed in plan S27 v4 since v3?
11 changes are waiting for Marcus Hale's signature, worth $9.1M of route contribution. The largest is ORD–DUB on a 787-9 at $2.9M, then ORD–EDI at $1.8M and the ORD–BCN May start at $1.4M. ORD–ATH becomes the 12th when a planner promotes it.
Which what-ifs are below the margin floor?
Two, against the 8% floor. W-29 ORD–GRU loses $1.1M as a response to Northmere's 5 weekly, against a $2.6M loss with no response. W-26 ORD–DEN was rejected at a 4.1% margin.
Show me where the agents disagree
There are two open disagreements. On W-32 ORD–ATH, Demand forecasts 82% and Competition 76% if Northmere matches. On W-31 ORD–NRT, Demand wants 10 weekly but the Fleet agent finds no aircraft.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds the agents flag against, the hub's own rules and who may do what are settings the planning systems administrator keeps. Changing a threshold re-flags the what-ifs it touches.
| Setting | Default | Choose from |
|---|---|---|
| Minimum route contribution margin | 8% | Percent; what-ifs below it are flagged |
| Load-factor floor | 75% | Percent |
| Disagreement band | 4 points of load factor or 10% of revenue | Points and percent |
| Minimum feed | 5 cities | Spoke cities |
| Data freshness | 30 days for bookings, one quarter for the ticket sample | Days and quarters, by source |
| Minimum connect times at ORD | 50 min domestic, 75 min domestic to international, 90 min international to domestic with customs | Minutes, by connection type |
| Operational spare at ORD | 1 aircraft per widebody type | Aircraft |
| Roles | Planner promotes, SVP signs, Fleet confirms, Alliance Lead approves partner drafts | Role matrix; the agents' row is locked |
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 season · “target” = the design goal, measured in the live solution · “estimated” = our estimate (the days in “where a new-route study's days go”) · 14 CFR Part 260 (automatic refunds after significant schedule changes, from October 28, 2024) and the 2007 US–EU Air Transport Agreement are named as context. People, Corvane Airways, Northmere Airways, Aurelle Air, Kyma Air, tails and flight numbers named on this page are fictional: characters and sample data in the working solution.
Airline network planning is deciding what an airline flies: which cities, how often, on which aircraft and in which bank of the hub. Airline Network Planning is network planning software in which eight agents answer every route what-if on the network map — demand, connecting feed, fleet and heavy checks, rival response, slots and traffic rights, partner flows and route economics — each with its sources, while a Review agent re-checks them blind. Planners compare versions and promote one with a reason code, and the SVP Network Planning signs the season plan version.
In airline route planning with this solution, the planner types the what-if in plain words, such as “ORD–ATH four times weekly next summer on a 787-9”. The agents size origin and destination (O&D) demand from the DOT ticket sample and bookings, split local from connecting traffic through the hub bank, check the fleet, estimate a rival's response, check slots and traffic rights, count partner flows and price the route line by line. The planner then changes frequency, aircraft or start date, and only the agents the change touches re-run.
The Connection agent places the flight in its departure bank and counts the spoke cities whose arrivals make the minimum connect time. In the sample, a 17:55 departure at the end of bank B6 catches 23 spoke cities and a 46% connecting share; the planner can try bank B5 or B7 and see the feed change. The agent never moves another flight to make a connection work.
The Competition agent estimates a rival's likely response from its published schedules, fares and launch history, and shows the effect: in the sample, 76% load and $35.2M if the rival matches 3x weekly, against the Demand agent's 82%. It states a probability with its basis, never a rival's intent as fact. When the two differ beyond the disagreement band, the Review agent shows both and what would settle it.
Yes. The Fleet agent reads the fleet register, lines of flying and the heavy-check calendar for every what-if, finds the tail the route would take and proposes fixes — move the start, use another route's aircraft, or wet-lease. The Fleet Planning Manager confirms or overrides the finding; the agent never assigns a tail. Catching the gap before sale avoids the late schedule change that can trigger refunds under 14 CFR Part 260.
The Slots and rights agent checks slot coordination, curfews, traffic rights and the customs hall and drafts the slot request; in the sample, US–EU service needs no new route authority under the 2007 US–EU Air Transport Agreement. The Partner agent values the codeshare and drafts the request for a DOT codeshare statement of authorization. People review and send both; agents never file with a coordinator or DOT.
Yes. Agents compute, build versions and draft; they never promote, sign, confirm or file, and that setting is locked. Only a Network Planner promotes a version, with a reason code and an attestation that covers any open disagreement, and only the SVP Network Planning signs a season plan version. Every agent run and decision is on an audit trail that cannot be edited.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your hub and stations, bank structure, fleet register, schedule system, ticket and traffic data, partner exchange and roles — and it goes live once every quality gate has passed. We will walk you through it on one of your own route ideas first.
One more point of revenue per flight, and no seat change posted without an analyst's yes.
1-min demoAirline Fare FilingEvery filing shopped before it publishes, so a mistake fare never reaches a shopper.
1-min demoCompetitive Fare MonitoringEach morning, the five rival moves worth money, with the captures that prove them.
1-min demoWe'll run Airline Network Planning on one of your own route ideas, your hub banks and your fleet plan.