Why the cage still runs on people
A casino cage is a bank inside a gaming floor. Cage workers exchange chips and tokens for cash, cash checks, issue and redeem markers, count and balance their drawer at the end of a shift, and keep the paperwork that gaming regulators expect. The money moves fast, in person, and someone has to be accountable for every dollar of it.
That accountability is the sticking point. Counting cash is arithmetic, and machines are good at arithmetic. Deciding whether a patron’s identification matches the person at the window, whether a check should be cashed, or whether a run of transactions looks like structuring is judgment, and a licensed human signs for it. Casinos also sit under strict cash-reporting rules, so the person completing a currency transaction report is part of the control, not just the clerk filling in a form.
The other half is the counter itself. Patrons at a cage window are often tired, sometimes losing money, and occasionally arguing. Handling a dispute over a chip count, calming a guest, or calling a supervisor at the right moment is work AI can prompt but not perform.
What software handles, what it assists, and what stays at the window
Machines already own the counting. Cash recyclers, chip-counting trays and transaction systems total and reconcile faster than a person, and reporting software can draft the regulatory records a cage produces each day. Of the work AI touches, 5% is the share it can take outright. Our coverage score, which asks how much of the total task time AI can handle today, reads 22 out of 100; how coverage is measured explains what goes into it.
Assistance covers the middle ground. Identity checks, flagging unusual patterns for a human to look at, and pre-filling forms for review all speed the job without removing the signature. That assisted slice is 23% of the automatable work.
The rest sits with people: 72% of task time. That is the window itself. Serving a patron face to face, settling a disagreement over a transaction, deciding when something needs a supervisor or a report, and holding personal responsibility for a cash bank are all duties a casino assigns to a named, licensed employee.
What the evidence shows
There is no published head-to-head test of AI against trained cage workers on their own tasks. That is why the evidence grade for quality parity is D, and why this page carries no parity number for the job. A grade at that level means not measured, not measured and found wanting.
What would settle it is specific: a study timing accurate drawer balancing and shift reconciliation with and without automated counting, an audit comparing suspicious-activity flagging by compliance software against experienced cage staff, and error rates on currency transaction reports drafted by software versus completed by people. Until something like that is published, the honest answer is that the counting is proven and the judgment is untested. You can read how we treat untested claims on the quality parity method page, and the full method at our methodology.
When the cage could change
Most likely after 2036 (8 in 10 of our scenarios). For what that window does and does not mean, see how we build the replacement year.
Two things could pull the date earlier. Ticket-in, ticket-out kiosks and cash recyclers keep spreading across floors, so routine chip and voucher redemption increasingly happens without a cashier. And compliance tooling is getting better at drafting the reports that used to take a cage worker real time each shift.
Two things hold it back. Gaming regulation ties cash accountability to licensed individuals, and swapping that for a machine is a rules change, not a software update. Physical cash still has to be moved, secured and audited, and the robotics tier that could do that work on a casino floor is mobile machines rather than cheap fixed hardware. Employment in the occupation is also already thin: the Bureau of Labor Statistics counts about 14,430 cage workers in the US with median pay of $37,580, and projects a small decline through 2035 (BLS, 2025). Shrinkage here looks like fewer windows staffed, not an empty cage.
How to stay needed in the cage
Lean into the parts of the job a kiosk cannot cover. Take the customer-facing work seriously: disputes, check-cashing decisions and marker transactions are where experience shows. Own the compliance side, including currency transaction reports and suspicious-activity escalation. And build a reputation for clean drawer accountability, because trust with cash is why casinos keep people in these roles.
Two skills pay off. The first is regulatory knowledge: Title 31 reporting, anti-money-laundering basics and your jurisdiction’s gaming rules. The second is supervision, since shift leads and cage managers decide how the automated tools get used.
What to do: ask your cage manager which reports your system already drafts for you, then learn to audit them rather than retype them.
If you want to look sideways, the closest work is at Tellers, Gambling Change Persons and Booth Cashiers and First-Line Supervisors of Gambling Services Workers. You can also browse the wider financial clerks family, the arts and entertainment sector, our list of jobs most exposed to AI, or put two jobs side by side.