■ CRITICAL RISK ■ Technology
Yes — the irony of being automated out of a job by the very machines you tended. Automated job scheduling, self-monitoring systems, and cloud migration have been dismantling the operations room for thirty years, and AIOps is sweeping up what's left.
“Cloud computing killed the mainframe room. And the operator.”
Our AI replacement risk score — how we score jobs
The classic operator ran the machine room: mounting tapes, loading print jobs, watching consoles for abends, restarting failed batch jobs, executing the nightly schedule in the right order, and calling the on-call programmer at 3 a.m. when the payroll run died. It was shift work built around the rhythm of batch processing, and for decades every bank, insurer, and government agency staffed it around the clock.
The dismantling came in layers. Workload automation software took over job scheduling and dependency management long ago — the schedule runs itself and retries its own failures. Virtual tape libraries ended the tape-mounting ballet. Monitoring suites watch the consoles better than tired humans and page the right engineer directly, skipping the operator middleman. Then cloud migration removed entire data centers from company premises, and the operations that remain concentrate in a handful of providers running highly automated facilities with skeleton crews. The newest layer, AIOps, uses machine learning to detect anomalies and remediate routine incidents automatically — explicitly targeting the judgment calls that were the operator's last claim.
The twist that keeps this interesting: mainframes themselves aren't dead. Core banking, airlines, and government systems still run on them, and the people who deeply understand z/OS are retiring en masse, creating real scarcity for mainframe systems programmers and administrators. But that's a skilled engineering role, not an operations role. The console-watcher position our 92 describes is nearly extinct, while the veteran who can debug a COBOL batch failure at the system level has become, briefly and lucratively, hard to find.
Automatability: our editorial assessment of current and near-term AI capability
Mostly already done — this decline started in the 1990s with workload automation and accelerated through the cloud era. What's finishing it now is AIOps absorbing incident response and the consolidation of remaining mainframe estates into outsourced, heavily automated operations centers. Remaining operator seats disappear steadily this decade; simultaneously, demand for genuine mainframe engineering skills spikes as the veteran generation retires.
Both are true, about different jobs. Operator roles — console watching, batch babysitting, media handling — have been automated nearly to extinction. Mainframe engineering roles — systems programmers, COBOL developers, DB2 specialists — face a genuine shortage as veterans retire while banks and governments still run the platforms. The machine room died; the skill set became scarce.
AIOps applies machine learning to IT operations data: detecting anomalies, correlating alerts, predicting failures, and auto-remediating routine incidents. It matters because incident triage and first response were the last substantial human tasks in the operations center. Vendors pitch it explicitly as reducing operations headcount, and large enterprises are deploying it now.
Yes, and it's the most common escape route. Operations discipline — change control, incident response, understanding batch dependencies and failure modes — maps directly onto cloud operations and site reliability work. The gap to close is tooling: Linux, scripting, and a major cloud platform's certification path. Operators who make the jump often outperform developers at production discipline.
Selectively, yes — but learn engineering, not operations. COBOL maintenance, z/OS systems programming, and mainframe-to-cloud migration expertise command strong rates precisely because supply is collapsing faster than the workload. It's a contrarian bet with a shrinking but wealthy market: core banking systems will need these skills for years, and almost nobody under fifty has them.