The Federal Aviation Administration is turning on a new artificial intelligence system called SMART over the skies of Washington, DC, as early as this week. It is the first live test of a nationwide plan that could eventually cover the entire United States, all 29 million square miles of airspace the agency manages.
SMART does not fly planes and does not talk to pilots. It is a forecasting tool. It looks at airline schedules, weather, how many planes an airport can handle, and current airspace conditions, then warns air traffic controllers about a jam or conflict before it happens. Right now controllers typically see trouble coming about 15 minutes ahead. SMART is built to stretch that warning window out to two hours.
The company behind it, Air Space Intelligence, is a small Boston startup, not a giant defense contractor. It beat out much bigger and more famous competitors, including Palantir and the French company Thales, for a 12 year contract worth 875 million dollars. That works out to roughly 73 million dollars a year, a small slice of the 12.5 billion dollars Congress already set aside to fix the country's air traffic system. Air Space Intelligence already sells a version of this technology to Alaska Airlines, Delta, United, and even the US Air Force, so the FAA is buying a tool with a track record rather than an untested idea.
Why start in Washington, and why start small? Two reasons. First, DC airspace is some of the most congested and closely watched in the country, especially after the deadly midair collision between an American Airlines regional jet and an Army helicopter near Reagan National in January, which killed 67 people and forced the FAA to permanently restrict helicopter routes near the airport. Second, the FAA is being deliberately cautious. Airline officials told reporters they were nervous about how much control the AI would actually have, and the agency responded by promising that SMART will not change any existing procedures for controllers or airlines. It only offers suggestions through systems controllers already use.
That caution matters more than the AI itself. The FAA is running this rollout while still short on air traffic controllers, still repairing outages caused by decades old equipment, and still under intense public scrutiny after a fatal crash. Layering a new AI tool on top of that would be reckless if it changed how humans make decisions on day one. Instead, the agency is using AI to buy controllers more lead time to do the same job they already do, not to replace their judgment.
For any business that depends on reliable flights, whether that is a manufacturer waiting on air freight or an executive who flies every week, the honest takeaway is patience. This is a 12 year contract with a phased national rollout. It will not fix delays this year or likely next. But it is a real signal that Washington is finally treating air traffic control as urgent infrastructure rather than a problem to patch quietly.