Why teller work keeps moving, but not all at once
Teller jobs changed long before chatbots arrived. ATMs took the cash withdrawal. Mobile apps took the check deposit. The question people type is whether bank tellers will be replaced by AI next, and the honest answer is that the work is eroding task by task rather than disappearing in one step.
Two parts of the job explain why people are still behind the counter. One is handling physical money and documents: counting currency, balancing the cash drawer, and examining checks for endorsement and negotiability. Software can read a check image, but someone has to take the paper, feel the bill, and sort out a drawer that does not balance at the end of a shift.
The other is the awkward conversation. A customer disputes a fee. A signature does not match. Someone shows up with a power of attorney and a grieving relative’s passbook. Those moments mix identity checks, compliance rules, and judgment about a person in front of you. That is the part branches keep staffed, and it is the main reason the headline figure on this page sits where it does. You can read how that figure is built on our methodology page.
The pressure is real, though. The US Bureau of Labor Statistics counted about 329,480 teller jobs with median pay near $43,030 a year, and projects employment falling roughly 13% between 2025 and 2035 (BLS, 2025). Most of that decline comes from fewer branches and more self-service, not from one clever model.
What machines run, what they assist, and what stays at the window
Routine transaction processing is where software already carries the load. Deposit capture, balance lookups, loan payment posting, and ordering cards or checkbooks run through systems with little human input. Across this job’s task list, AI can handle about 0% of task time on its own. That share is our coverage read, explained under Can AI do it?
A second group is assisted rather than automated. Answering account questions and spotting transaction errors both go faster with search tools, fraud flags, and scripted prompts, but a teller still decides what to tell the customer and what to escalate. AI helps with roughly 60% of the work here.
The rest sits with people. Verifying identity and signatures, counting and reconciling cash, and explaining a hold or a dispute to an unhappy customer all stay human in our task split, which covers about 40% of task time. Those tasks need hands, eyes, and accountability in the branch.
Good to know: self-service is the bigger force in this job, so branch closures often matter more to a teller’s year than any new model release.
How strong is the evidence on tellers?
Weak, and we say so. The quality parity grade for this occupation is D, which means no study has yet tested an AI system against a qualified teller on this job’s real tasks. Because of that, we publish no parity number for tellers. A grade like this is a statement about missing measurement, not about machine ability.
What would move it? A benchmark on check examination and negotiability decisions, scored against trained tellers. A field trial of automated identity verification in branches, with error and fraud rates reported. Published audit data on cash reconciliation by automated recyclers versus staffed drawers. Until something like that exists, the parity question stays open, and we explain the grading scale under Is it better than a person?
When the work could shift again
Most likely between 2036 and 2058 (8 in 10 of our scenarios). That window is a median with a range, not a prediction about any one branch; the replacement year method sets out how it is produced.
Two things could pull the date earlier. The first is branch design: teller-free lobbies with video banking and cash recyclers cut the number of windows a bank needs, which does more than any software upgrade. The second is cost. Running software against a transaction costs far less per year than staffing a counter, and that gap is visible in the cost panel on this page.
Two things hold it back. Over half of this job’s work is physical, and the robotics it would need falls in the mobile robots tier, which is hardware that has to be bought, installed, and serviced branch by branch. Compliance is the other brake: know-your-customer rules, suspicious activity reporting, and audit trails all assume a named person made the call. Banks move slowly when a mistake is a regulatory event.
How to stay needed behind the counter
Lean into the parts of the job that the task list keeps human. Identity and signature verification, because it mixes rules with a judgment call. Cash reconciliation and drawer balancing, because someone has to find the difference and account for it. And face-to-face problem solving on holds, disputes, and fraud scares, because that is what pulls a customer into the branch in the first place.
Two skills raise the floor. Fraud and compliance literacy: knowing what triggers a report, what documents are valid, and when to escalate. And consultative selling, meaning the ability to open an account, explain a loan option, and hand off cleanly to a lender. Tellers who can do both tend to move up rather than out.
Nearby roles use the same base. New accounts clerks build on the account-opening side. Loan interviewers and clerks take the lending path. Customer service representatives cover the same problem solving across channels. You can see the wider group on our financial clerks family page and the industry picture under banking.
The score for this job is 70 out of 100 (higher is safer), with coverage at 31 out of 100. If you are weighing a move, put this job next to a target role on the compare tool, or check which roles sit on our list of jobs AI is expected to shrink.