∞ALTA Insight

Afraid Your AI Project Won’t Take Flight? You Should Be.

The FAA’s legacy infrastructure warnings and accelerated AI rollout

Afraid Your AI Project Won’t Take Flight? You Should Be.

On November 13, work to replace a copper telecommunications connection with fibre at a remote Federal Aviation Administration facility preceded a nationwide ground stop. According to Transportation Secretary Sean Duffy, a router began generating more than 1,000 connection attempts every four to five seconds, overwhelming parts of the air traffic control network. Communications and radar services were disrupted while the fault was isolated and service restored. (Investigators have yet to establish whether equipment or installation caused the initial malfunction.)

The disruption began during a routine upgrade to infrastructure the FAA had been trying to retire for years. At 8ALTA, we see a comparable exposure emerging in enterprises connecting AI to technology estates built over decades, often without a complete picture of what those systems can tolerate. The pace of AI adoption is increasingly outstripping the work needed to establish whether the underlying architecture is ready.

Warnings stretching across budget cycles

The FAA had been explicit about the risks of aging telecommunications infrastructure in successive FY2023, FY2024, FY2025 and FY2026 budget submissions, warning that Time Division Multiplexing (TDM) services were being withdrawn, replacement equipment was increasingly difficult to support, and obsolete components posed operational and security risks. The Government Accountability Office reinforced those concerns in September 2024, when it reported that 51 of 138 air traffic control systems were unsustainable, and again in September 2026, when it raised concerns about modernization schedules, costs and the potential for disruption during implementation.

While those reports documented broad infrastructure vulnerabilities, they did not predict the specific failure on November 13, and the FAA has said the copper-to-fibre work was separate from its larger modernization programme. Still, the incident demonstrates how much can depend on a connection that looks insignificant on an organizational chart.

A similar problem in healthcare

Nearly two decades ago, while consulting with TELUS, 8ALTA partner Ricardo Casanova helped develop a communications modernization offering informed by the needs of Alberta Health Services (AHS).

The consolidation of nine regional health authorities had left Alberta with a mixture of paging systems and established operating practices. Maintenance was becoming increasingly difficult after Glenayre, a major paging supplier, entered insolvency proceedings. Replacement components were sometimes scarce enough to be sourced through secondary markets, including eBay.

The existing paging networks remained important as cellular coverage inside hospitals was inconsistent, especially in basements and service corridors. Clinicians depended on messages reaching them regardless of where they were working.

The TELUS offering operated within the paging infrastructure and enabled doctors and first responders to receive messages on traditional pagers while also receiving them on BlackBerry devices for secure enterprise messaging. Both channels remained available during the transition and AHS was informed of progress and cooperated in discussions of its requirements and concerns.

That arrangement allowed the newer technology to earn trust in daily operation while preserving communications people relied on. It also reduced the technology change and adoption friction that typically accompanies a new service during rollout.

The AI complication

The FAA was also introducing AI into its air traffic management operations. On September 21, it launched SMART, an AI-supported planning system that combines approximately 200 streams of weather, flight, traffic and staffing data to help specialists recommend adjustments. On November 10, the FAA announced its expansion to airspace managed by all 22 en route control centers. Three days later, the nationwide ground stop brought a separate, longstanding concern about the resilience of the agency’s communications infrastructure into sharp focus.

The SMART rollout had already drawn scrutiny. In October, more than two dozen lawmakers asked the FAA to pause deployment pending answers about safety evaluations, testing reports, staff training and oversight. Their request preceded the nationwide expansion. The FAA describes SMART as a decision-support system: specialists validate its output and operational leaders consider the recommendations. The public record does not establish that SMART caused, contributed to or shared a failure path with the November 13 telecommunications outage. Air Space Intelligence, the developer, denied a connection, and officials have attributed the disruption to network equipment affected during copper-to-fibre work.

8ALTA sees a broader governance problem in the coincidence of these timelines. When an organization accelerates a new AI capability while modernizing aging infrastructure, leadership needs a defensible account of the dependencies, testing and operational boundaries of each programme. A preliminary explanation of the component that failed cannot substitute for understanding what initiated the fault and how it propagated. In healthcare, parallel operation gave a new communications service time to prove itself without withdrawing the established channel. The same discipline matters when introducing AI into systems on which essential operations depend.

Many business applications were designed around predictable transactions, whether through their databases, data structures and integration layers that have evolved through years of patches, acquisitions and incremental upgrades. They can remain dependable within their established workloads even when the architecture underneath has become difficult to maintain.

A chatbot may appear to retrieve information effortlessly while depending on records that are fragmented, inconsistently defined or difficult to reconcile. An autonomous agent can generate multiple queries, retry failed requests and initiate subsequent actions before an employee has noticed a problem. A database built for a narrow set of application calls may struggle when automated activity increases the volume and variety of requests.

An incorrect record may influence several downstream decisions when an agent treats it as authoritative. A timeout can trigger retries that place additional load on an already struggling service, while an agent with poorly defined permissions may repeat an action across far more records than an employee could process in the same period.

The operational blast radius grows when automated activity propagates failures before anyone intervenes, particularly where experienced employees once supplied the workarounds.

Preparing the foundations

8ALTA recommends treating infrastructure readiness as a condition of granting AI access to critical operations. Technical leaders need to understand the systems supplying its data and executing its requests, including unsupported components, capacity limits, inconsistent records and dependencies between applications. Testing should establish what happens when a service slows down, returns incorrect information or becomes unavailable, with accountable owners for the systems and decisions affected.

Controls must be proportionate to the authority granted. Reading a record presents a different risk from modifying it, triggering a payment or changing an operational schedule. Rate limits, restricted permissions, monitored retries and reliable ways to stop automated activity are essential where failures could spread.

Modernization can proceed in stages. Gateways and integration layers may help preserve continuity while older components are replaced, but they cannot indefinitely compensate for hardware and software that vendors no longer support. The organization still needs a funded path toward a sustainable architecture.

The FAA’s infrastructure warnings accumulated over several budget cycles, while questions about the readiness of its new AI-supported planning tool were raised before nationwide expansion. No established finding connects SMART to the ground stop, but the overlapping programmes illustrate why 8ALTA argues that AI readiness must include the systems beneath the application, the boundaries of its authority and the evidence used to approve deployment. Organizations funding new AI capabilities while postponing investment in the infrastructure they depend upon may discover the cost of that decision during live operations.

8ALTA advises leadership teams on the architecture, operational resilience and modernization decisions that determine whether enterprise AI can be deployed responsibly and sustained at scale by our Certified AI Vendor Partners.

Sources

U.S. Department of Transportation, November 14, 2026, FAA ground-stop briefing: https://www.transportation.gov/briefing-room/icymi-us-transportation-secretary-sean-p-duffy-provides-update-restoration-airspace
FAA FY2023 Budget Submission: https://www.transportation.gov/sites/dot.gov/files/2022-04/FAA_Budget_Estimates_FY2023.pdf
FAA FY2024 Budget Submission: https://www.transportation.gov/sites/dot.gov/files/2023-03/FAA_FY_2024_President_Budget_508.pdf
FAA FY2025 Budget Submission: https://www.transportation.gov/sites/dot.gov/files/2024-03/FAA_FY_2025_Budget_Request_508-v5.pdf
FAA FY2026 Budget Submission: https://www.transportation.gov/sites/dot.gov/files/2025-05/FAA_FY_2026_Budget_Estimates_CJ.pdf
GAO, September 2024: https://www.gao.gov/products/gao-24-107001
GAO, September 2026: https://www.gao.gov/products/gao-26-107992
Ricardo Casanova, AHS communications modernization case study: https://www.toasty.media/casanova_case_study.pdf
FAA, September 21, 2026 SMART launch: https://www.faa.gov/newsroom/trumps-transportation-secretary-sean-p-duffy-delivers-state-art-air-traffic-control
FAA, November 10, 2026 SMART expansion and fact sheet: https://www.faa.gov/newsroom/statements/general-statements
Senator Chris Van Hollen, October 9, 2026 congressional request on SMART: https://www.vanhollen.senate.gov/news/press-releases/van-hollen-joins-mcclain-delaney-colleagues-in-call-to-pause-faas-ai-tool-at-washington-area-airports