Is a Bank Teller job safe from AI?

Only partly. A Bank Teller scores 43/100 (medium risk). AI can already handle a meaningful share of the tasks, so the safest path is shifting toward the judgment, relationship, and oversight parts of the role.

When will AI be able to do a Bank Teller's job?

Based on AI's measured task-completion horizon (METR, doubling ~every 4.3 months), AI reaches this role's core complexity ~3.1 years (2029). That projects the capability — actual adoption usually lags it.

How accurate is the Bank Teller AI-risk score?

It's a transparent, computed estimate — directionally useful, not a guarantee. It blends six labor and AI-exposure signals (O*NET, BLS, Eloundou task exposure, AIOE, the Anthropic Economic Index, and physical-automation data). See the methodology page for the full formula.

A Bank Teller carries a 43/100 AI replacement risk (medium). AI can already handle processing transactions; Complex problem resolution still needs a person. Of exposed work, ~100% is automation vs 0% augmentation. Capability clock: ~3.1 years (2029). (ReplacedYet AI-Risk Index, 2026 data.)

Will AI replace a Bank Teller?

AI replacement risk: 43/100 (medium risk). Moderate exposure — AI automates routine parts; judgment and relationships remain human.

Timeline: 2030–2034. Of the exposed work, roughly 100% is likely to be automated and 0% augmented. $5.9B/yr of US wages sit in highly-exposed work for this role.

AI/software exposure: 57%. Robot/physical-automation exposure: 50%.

Capability clock: AI's measured task horizon reaches this role's core complexity (~2249.2h of human work) ~3.1 years (2029) — projected from METR's ~4.3-month doubling.

Pressure Index: 55/100 (medium) — blends risk, demand trend, and real-world evidence. Job postings down 30% vs 2020.

AI tools targeting this role

Interactive Teller Machines — remote and self-service teller transactions
Kasisto — conversational banking that answers customer questions

Layoff signal: high — Digital banking and self-service machines continue to drive a long decline in teller positions.

Then vs. now: the 2013 Oxford study scored this 98%; our 2026 index scores it 43% (a fall of 55 points).

Tasks at risk

Processing transactions — ATMs and apps handle most.
Balance inquiries — Fully self-service.
Routine account tasks — Online banking automates these.

Tasks that still need a human

Complex problem resolution — In-person help for hard cases.
Relationship banking — Trust and cross-sell.

Skills that protect you

Personal banking / advising — Relationship roles.
Fraud & compliance — Specialized judgment.
Branch operations — Run the branch.

Safer adjacent careers

Personal Financial Advisor (28%) · Wealth Manager (28%) · Commodities Broker (28%) · Financial Manager (29%)

Frequently asked questions

Will AI replace Bank Tellers?: A Bank Teller carries a 43/100 AI replacement risk (medium). AI can already handle processing transactions; Complex problem resolution still needs a person. Of exposed work, ~100% is automation vs 0% augmentation. Capability clock: ~3.1 years (2029). (ReplacedYet AI-Risk Index, 2026 data.)
Is a Bank Teller job safe from AI?: Only partly. A Bank Teller scores 43/100 (medium risk). AI can already handle a meaningful share of the tasks, so the safest path is shifting toward the judgment, relationship, and oversight parts of the role.
When will AI be able to do a Bank Teller's job?: Based on AI's measured task-completion horizon (METR, doubling ~every 4.3 months), AI reaches this role's core complexity ~3.1 years (2029). That projects the capability — actual adoption usually lags it.
How accurate is the Bank Teller AI-risk score?: It's a transparent, computed estimate — directionally useful, not a guarantee. It blends six labor and AI-exposure signals (O*NET, BLS, Eloundou task exposure, AIOE, the Anthropic Economic Index, and physical-automation data). See the methodology page for the full formula.

Category: Finance · Methodology · Download the dataset

ReplacedYet AI-Risk Index. Last updated 2026-06-27. AI-estimated and directionally useful, not a guarantee.