With 3M+ people filing for unemployment benefits last week, US states are seeing their aging tech systems get overwhelmed and experts say it will only get worse
that's a huge problem https://www.wired.com/...
Context & Ripple Effects
A single week of more than 3 million unemployment filings has pushed state benefits systems — much of them still running on decades-old COBOL-era infrastructure — past their breaking point, and experts expect the load to climb as layoffs spread. The surge is not an anomaly but the leading edge of a wave: researchers have already counted 29,463 layoffs across 292 tech companies since mid-March.
The stakes go beyond slow checks. Congress's $2T relief bill extended benefits to gig workers, but those claimants report the same overloaded systems make the money hard to access and slow to arrive. And states that rushed automation before have a cautionary precedent: Michigan's flawed 2014 system falsely charged thousands of people with fraud and collected millions in fines.
First-order effects
- Millions of newly jobless workers face delayed or blocked benefit payments as state systems buckle under filing volumes they were never sized for.
- State IT departments are forced into emergency maintenance of COBOL-based platforms while simultaneously onboarding entirely new claimant categories created by the relief bill.
Second-order effects
- Gig workers, the relief bill's newest beneficiaries, become the group most exposed to the bottleneck — their benefits exist on paper but stall in the same overloaded pipelines, pressuring states to patch eligibility logic onto fragile cores.
- Under delivery pressure, states face heightened risk of repeating Michigan's automated-fraud failure, where a hastily deployed system generated false charges at scale.
Third-order effects
- The Atlanta Fed later put a price tag on this fragility, estimating delays in COBOL-based unemployment systems cut US GDP by at least $40B from March 2020 through year-end — turning legacy government IT from a back-office complaint into a measurable macroeconomic liability.
- With labor experts now warning that unemployment insurance is unlikely to cushion AI-driven job losses and Congress doing little to prepare, the 2020 overload reads as the first stress test of a safety net whose underlying architecture has not been structurally rebuilt since.
The trend: America's unemployment safety net is colliding with its own legacy IT: each economic shock exposes the same COBOL-era core, and the costs — measured in delayed payments and billions in lost GDP — keep compounding faster than states modernize.