← All articles
NEWS

The Real Cost of a Bank Teller Conversation (And What Replaces It)

The cost of a conversation. A closer look at teller and mobile service costs, and how to measure whether avatar-guided self-service actually saves money.

A number gets repeated often enough in banking circles that it stops getting checked: a teller or call-centre interaction costs around $4, a mobile interaction costs about a dime. It's a real number, from a real source, but the version that circulates has drifted from what the source actually says, and the drift matters for how you use it.

What Bain Actually Found

The figure traces to Bain & Company's own analysis, based on a survey of roughly 5,300 US banking consumers conducted in 2016, with a separate savings estimate elsewhere in the piece for a typical US bank running 1,000 branches. Bain's own wording on the cost gap itself: each mobile interaction incurs a variable cost of about 10 cents, versus $4 for a teller or call-agent interaction. That's one combined figure for tellers and call agents together, not two separate numbers, and it's US-specific data from 2016, not a current global benchmark.

What That Means, and What It Doesn't

Nearly a decade of inflation, branch consolidation, and digital investment separate that survey from today, so treating $4.00 and $0.10 as this year's numbers for your own institution would be a mistake. What almost certainly hasn't changed is the direction: a conversation that requires a person's time costs more per interaction than one a system handles on its own, and that gap is why banks keep investing in digital channels even as branch visits decline. The specific multiple is dated. The underlying economics aren't.

Why the Gap Exists, and Where It Gets Complicated

A teller or call-agent interaction costs more because it requires a person's full attention for the length of the conversation, with no ability to serve another customer at the same time. A mobile interaction scales without that constraint, the marginal cost of one more balance check approaches zero once the system is built. That's a capacity argument, not a complicated one, and it's the whole reason self-service exists. What it doesn't capture is that a lot of what happens in a teller or call-agent conversation isn't simple, it's a customer working through a decision, asking a follow-up question, or needing something explained in a way a static screen doesn't handle well. A cheaper channel that can't do that pushes the customer back to the expensive one anyway.

Where a Conversational Avatar Sits Between These Two Costs

One hypothesis worth naming, untested by Bain's data or by anything else we found: a conversational avatar inside the mobile channel could handle more of the multi-step, needs-explaining conversations that currently push a customer back toward a teller or call agent, without carrying the per-interaction cost of a person's time. Whether it actually closes that gap, and by how much, depends on how well it's scoped to the specific questions that currently escalate, not on the technology alone. That's a pilot to run, not a savings figure to promise in advance.

Teller/Call-Agent vs Self-Service Mobile vs Avatar-Guided Mobile

The point of this comparison isn't that avatar-guided mobile wins outright. It's that the choice isn't only between a $4 conversation and a 10-cent form. There's a real question in the middle: which of the conversations currently happening at the expensive end could actually be handled by something more capable than a static screen, without requiring a person's time.

A Hypothetical Worth Walking Through

Take a purely illustrative example, not a real bank's figures: a call centre fields a steady stream of calls asking why a specific transaction was flagged or delayed, a question that takes a person two or three minutes to answer but requires no judgment call, just an explanation of what the system already knows. If even a quarter of those calls could be handled by an avatar inside the app instead, the bank isn't just saving the per-call cost. It's freeing call-centre time for the calls that do need a person's judgment. Whether that share is large enough to justify the build is exactly the kind of thing worth measuring in a scoped pilot rather than assuming.

What to Measure Before You Believe the Savings

  • Call and chat volume for the specific question types the avatar is scoped to handle, before and after.
  • Resolution rate within scope, versus how often it hands off to a person.
  • Customer satisfaction for the scoped interaction type specifically, not the channel as a whole.
  • Actual build and maintenance cost against the call volume it's redirecting away from a live agent.

Before quoting a savings number internally, pull your own call centre's cost per call and your own mobile app's cost per active user, and compare those to the gap you're actually trying to close. A decade-old survey from a different country's banking market is a reason to ask the question, not the answer to it. A number that took a research team a real survey to establish shouldn't get repeated as if it's still true of your own institution just because it's the one most people in banking already know.

Full source list

Bain & Company, "The Bank Branch and Call Center Traffic Jam": https://www.bain.com/insights/bank-branch-call-center-traffic-jam/

Avatars Global film

15 MINUTES. A LITTLE POSSIBILITY.

Let's find
a time to talk.

Tell us when you're free. We'll confirm a time with you by email.

When works for you?

We'll use your details to arrange this conversation. Privacy policy