Workflow guides

FAA AI and SMART: What Is Confirmed and How to Explain It

The FAA says SMART did not cause the November 13 ground stop. Its November 15 explanation identifies network instability while the root-cause investigation continues.

SJolt Editorial6 min readUpdated
Editorial diagram showing source verification, human review and a factual educational explanation
Original SJolt editorial diagram; illustrative workflow, not a product screenshot, operational dashboard or generated result.

FAA AI searches combine SMART deployment with the November 13 ground stop. In its November 15 update, the FAA explicitly says SMART was not responsible. The agency describes network instability and an ongoing root-cause investigation.

This guide uses FAA and Department of Transportation material checked on November 16, 2026, Beijing time. It explains the documented review chain and provides a practical method for making an educational explainer. It is not a flight-status service or operational aviation guidance.

Keep the deployment and incident records separate

On November 10, the FAA announced SMART’s expansion to airspace covered by all 22 centers after the initial National Capital Region rollout. It also reported early efficiency gains. Those agency-reported observations are not a nationwide guarantee or an independent evaluation.

The November 13 preliminary statement reported communications problems and a nationwide ground stop. Our original review could not attribute the disruption from that statement. The later agency explanation below adds evidence; keep the preliminary report dated rather than presenting it as the latest account.

For a dated explainer, save the exact source URL, publication date, retrieval time and a short paraphrase of the claim. Separate the event timeline, the agency’s explanation and remaining questions. Preserve the earlier evidence card when a later statement changes what can responsibly be said.

What the November 15 explanation changes

The FAA says excessive connection traffic destabilized legacy equipment on its FTI network. It identifies two locations under investigation, reports stabilization after isolating them, and says FAA and L3Harris are still investigating the root cause. It separately explains that SMART aggregates data without feeding it into air traffic control.

Use “the FAA says” when summarizing this explanation. Replace a caption implying that the SMART rollout caused the disruption, and replace an outdated statement that the agency has offered no explanation. An agency’s stated account and a completed independent investigation are different evidence categories.

Editorial fieldRevision to makeEvidence boundary
HeadlineAttribute the explanation to the FAA and date it.Do not turn a search phrase into a causal claim.
TimelineKeep the initial report and later explanation as separate entries.A later update does not change what the earlier source said.
DiagramLabel each system and cite the source for its role.Do not invent a technical connection between adjacent boxes.
Open questionsTrack unresolved details against the latest source.Do not silently promote a preliminary explanation into a final report.

For an existing graphic, review the caption, arrows and alternative text together. A corrected paragraph is insufficient if the thumbnail still suggests the old conclusion. Keep a dated correction note with the editable project so future exports use the same evidence version.

What the human review chain looks like

SMART stands for Strategic Management of Airspace, Routes and Trajectories. DOT’s September 21 launch describes an AI-supported planning tool using information such as weather, routes, traffic and staffing. It explicitly says the tool cannot replace controllers or take control of an aircraft, and that staff review its recommendations.

FAA specialists validate the data and send scheduling recommendations through the System Command Center to local leadership. Describe that review chain explicitly; an arrow straight from “AI” to “aircraft” would misrepresent it.

Layer in an educational diagramWhat to showWhat to avoid implying
InputsWeather, demand and other data categoriesA complete or live list of operational feeds.
RecommendationA proposed scheduling changeA command sent directly to an aircraft.
ReviewPeople assessing the proposalAutomatic acceptance of every output.
DecisionA recorded acceptance or rejectionThat an illustrative diagram is a real control interface.

These diagram choices are an editorial interpretation of the described process. Use arrows to explain responsibility, not to invent technical interfaces. If a source does not specify the model architecture, hosting provider or API, leave those details out instead of filling the blank with a familiar brand.

Ask questions that the available numbers can answer

An efficiency figure needs a denominator and a comparison condition. Before turning a reported saving into a chart, identify the observed period, which flights were included, how the baseline was calculated and whether other changes occurred. If the source does not supply those details, say the number is reported by the agency and avoid extrapolating it nationwide.

Use the same discipline for failure analysis. Deployment dates establish sequence; causal attribution requires evidence connecting a mechanism to the observed failure. A screenshot, a social post or a matching keyword does not supply that mechanism. Keep open questions visible rather than hiding them behind a confident headline.

text
Evidence card for an explainer
Claim: one narrowly stated fact
Source: original publisher and URL
Published / checked: dates with timezone
Evidence type: agency statement / reported metric / analysis
Scope: system, geography and time period
Unknown: cause, denominator or interface not supplied
Allowed caption: wording no stronger than the evidence

Have a second reader compare each caption with its evidence card. They should be able to identify which sentence supports the claim without relying on the illustration. If the graphic is removed, the explanation should still be accurate; if the caption is removed, the image should not imply a stronger operational claim.

Build an educational visual with an explicit boundary

For an article or classroom presentation, start with a simple three-panel layout: data categories, human review and a scheduling recommendation. Add the labels in a layout editor so spelling and arrows are controlled. Avoid runway diagrams, real flight numbers or apparent live traffic positions that could make a conceptual graphic look like an operational display.

SJolt’s Nano Banana 2 text-to-image route can create an illustrative background for this communication task. It is an image-generation workflow, not a connection to FAA systems. SJolt does not list SMART as a model in the reviewed catalog, and this article makes no claim of FAA approval or certification.

text
Concept illustration for an educational article about human review of scheduling recommendations. Three separated panels: abstract data cards, a person reviewing a document, and a proposed calendar change. Calm navy and teal palette. Leave blank areas for labels added later. No agency seals, flight numbers, airport maps, radar screens, logos or text.

Keep the source-grounded explanation outside the image prompt. The image model is being asked for composition, not to determine what caused an outage. Review the asset for invented labels and misleading symbols, then add a visible caption identifying it as a conceptual illustration. No image was generated or benchmarked for this guide.

A final editorial check before sharing

Check the exported graphic at phone size. The human reviewer should remain visually distinct from the proposed action, and the arrows should not bypass that reviewer. Confirm that the caption includes the date of the explanation and links to the deployment statement and the latest incident explanation.

Keep the project editable so an official finding can be incorporated without rebuilding the entire illustration. When updating, distinguish a newly confirmed fact from a correction to your own earlier wording. This produces a durable explainer that can absorb better evidence while preserving an honest account of what was known at publication.

Sources & further reading

Take the next idea into production.

Explore the models, test a workflow in the playground, and use the same request in your application.

Keep exploring

← All stories