FAA空管AI評測：8.75億美元砸新創，能解決缺人危機？ | FAA Air Traffic AI Review: $875M Bet vs. Controller Crisis

9月 18, 2026

By Kit 小克 | AI Tool Observer | 2026-09-18

🇹🇼 FAA空管AI評測：8.75億美元砸新創，能解決缺人危機？

美國聯邦航空總署（FAA）砸下 8.75億美元，把美國的空中交通管制 AI 系統外包給一家波士頓新創 Air Space Intelligence（ASI），而不是找履歷更漂亮的 Palantir 或 Thales。這筆12年合約要打造代號 SMART 的系統，目標是在飛機起飛前就預測塞機、疏導航班，聽起來很美好，但問題是：AI空管能解決的到底是哪一種問題？

SMART系統在做什麼：用AI預測塞機

SMART（Strategic Management of Airspace, Routes, and Trajectories）是跑在FAA新資料骨幹FMDS（Flow Management Data and Services）上的一套AI模組，會即時分析航班時刻表、天氣、機場容量、空域狀況等資料，在飛機還沒離地前就標出可能塞爆的航路。部署時程先從華盛頓特區都會區開始，2028年前目標全面上線。

為什麼是ASI，不是Palantir或Thales？

Palantir跟Thales都是政府標案老手，資源、履歷都比ASI深厚，但這次FAA選了在隱身模式開發近一年的ASI。業界看法不一：有人認為這代表空中交通管制 AI的技術門檻其實沒那麼高，小團隊也能做出可用產品；也有人擔心少了備援供應商——輸掉標案的兩家公司剛被FAA說「不夠好」，未來想找備胎會很尷尬，採購圈稱這是「被鎖死」。

誠實看待：AI解決得了管制員荒嗎？

真正的病灶是人力，不是演算法：FAA目前短缺約3000名合格管制員，是20年來最嚴重的缺口。今年稍早FAA甚至下令40個主要機場減班，用的是「配給僅有的人力」，不是「排程算得更聰明」。
AI能做的是緩解，不是治本：SMART能提早示警壅塞、幫既有管制員做更好的決策，但它生不出新的合格管制員——那需要好幾年的訓練與招募。
時程拉得很長：12年合約、2028年才全面上線，中間任何一次系統延遲、標案爭議或政治風向轉變，都可能讓時程再拖。FAA過去在IT現代化上不是沒有翻車紀錄。

對一般人的意義

如果你常搭美國國內線，SMART上線初期（先在DC都會區）不會馬上讓誤點消失——管制員短缺這個根本問題還在。但長期來看，這是一個值得觀察的案例：政府砸大錢導入AI基礎建設，到底是解決問題，還是把預算花在包裝更漂亮的舊問題上。AI空管會不會成功，2028年全面上線後才知道。

好不好用，試了才知道。

🇺🇸 FAA Air Traffic AI Review: $875M Bet vs. Controller Crisis

The FAA just handed a Boston startup, Air Space Intelligence (ASI), an $875 million, 12-year contract to build an AI air traffic control system called SMART — beating out Palantir and Thales, two companies with far deeper government contracting resumes. The pitch: predict congestion before planes even leave the gate. The real question: does AI air traffic control actually fix what's broken?

What SMART Actually Does

SMART (Strategic Management of Airspace, Routes, and Trajectories) runs on top of FMDS (Flow Management Data and Services), the FAA's new data backbone replacing the old Traffic Flow Management System. It ingests airline schedules, weather, airport capacity, and airspace conditions in real time to flag likely bottlenecks before aircraft depart. Rollout starts in the Washington, D.C. metro area this fall, with full deployment targeted for 2028.

Why a Startup Beat Palantir and Thales

ASI had been building this in stealth mode for nearly a year, and it won over two incumbents with much bigger balance sheets. Reactions are split: some see it as proof that AI air traffic control doesn't require a giant defense contractor to solve. Others point out the FAA just told its two most obvious backup vendors they weren't good enough — leaving no real Plan B if ASI stumbles at scale.

The Honest Problem: AI Doesn't Fix a Staffing Shortage

The root cause is people, not routing math. The FAA is short roughly 3,000 certified controllers — the worst gap in over two decades. Earlier this year it ordered flight cuts at 40 major airports because it had to ration the controllers it had, not because the schedules were dumb.
SMART mitigates, it doesn't cure. Better congestion forecasting helps existing controllers make sharper calls, but it can't produce a single newly certified controller — that takes years of training and hiring.
The timeline is long. A 12-year contract with full rollout in 2028 leaves plenty of room for delays, procurement disputes, or political shifts. The FAA's track record on IT modernization isn't spotless.

What This Means If You Fly

If you fly U.S. domestic routes, don't expect SMART's D.C.-area debut to fix delays overnight — the controller shortage that actually causes most of the chaos is still there. But this is worth watching as a test case for whether pouring AI budget into government infrastructure solves real problems or just repackages old ones with a shinier interface. Whether this bet on AI air traffic control pays off won't be clear until the 2028 full rollout.

You won't know until you try it.

Sources / 資料來源

延伸閱讀 / Related Articles

AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

搜尋此網誌

AI小貼士