McKinsey Says AI Agents Could Nearly Triple Bank Returns. Shinhan Bank Is Asking a Different Question.

A glass bank headquarters tower at dusk with warm windows over a city financial district

The tower, and the projection.

By Mara Quinn | November 16, 2026

The 3x Number Banks Are Talking About

A new consultancy projection is circulating through banking circles. McKinsey says AI agents could help banks generate nearly three times their returns over a three-year period. The Times of India reported the figure on November 16, 2026.

That is a projection, not a measured result. McKinsey is a consultancy, not a regulator, and the number reflects modeled potential rather than observed outcomes. Still, a 3x figure from a firm that speaks to bank boards gets attention.

The claim comes at a moment when banks are actively debating how far to push AI agents into operations, lending decisions, compliance, and customer service. The projection gives advocates inside those institutions a headline number to work with.

What "3x Returns" Actually Means

McKinsey's framing is worth reading carefully. The projection covers a three-year window and applies to banks that actually deploy agents, not to the industry as a whole. The consultancy has not released the full methodology in what was reported by The Times of India.

That gap matters. Return projections for AI adoption vary widely depending on starting assumptions: which processes are automated, what labor costs are displaced, how implementation expenses are counted, and whether productivity gains compound or plateau. McKinsey's reputation means the number will be cited. The conditions attached to it are what practitioners need to track.

For banks still building their AI governance structures, a projection this large creates pressure. Teams pitching agent deployment internally now have an external benchmark. Teams arguing for slower, more controlled rollouts have a harder case to make.

Shinhan Bank Takes a Different Angle in Milan

On the same day the McKinsey projection surfaced, a report from Korea IT Times carried a different kind of banking-and-AI story.

Shinhan Bank, one of South Korea's major financial institutions, presented two research papers at a global conference held in Milan. Both papers focused on trustworthy financial AI.

Korea IT Times reported this on November 16, 2026. The specific contents of the papers and the conference name were not detailed in the coverage available.

The signal the research sends matters for this story. Shinhan Bank is not a small institution publishing speculative work. It is a major bank putting research resources toward the question of what makes AI trustworthy in a financial context, and presenting that work at an international conference.

That is a different posture than chasing a return multiple. It treats trust as something to be built and demonstrated, not assumed.

The Tension Running Through Both Stories

Read together, these two reports describe the same industry moment from opposite directions.

One story offers a number. McKinsey projects nearly 3x returns for banks that use agents over three years. The figure is modeled, not empirical, but it carries the weight of a named consultancy and the clarity of a ratio. It answers the question CFOs ask: what is the return?

The other story asks a harder question. Shinhan Bank's researchers spent enough time and institutional resources to produce two papers on trust in financial AI, then took those papers to a conference in Milan. The question they are working on is whether the systems generating that return can be trusted.

Both questions are real. Banks need to show returns, and they need systems that clients, regulators, and staff can trust. The problem is that the first question has a headline number attached to it now, and the second question is still being researched.

The gap between a projection and a published paper is the gap between a sales argument and a governance answer. Banks that move fast on agents because McKinsey projects 3x will still need to answer Shinhan Bank's questions eventually. Regulators will ask. Clients will ask. When a decision made by an AI agent goes wrong, the question that matters will not be what the average return across the portfolio was. It will be whether the system was trustworthy.

What Comes Next

Neither report offers a resolution. McKinsey's projection does not say which banks will achieve the 3x figure or what it costs to get there. Shinhan Bank's conference papers have not been fully summarized in available reporting, so the specific recommendations or frameworks they offer remain unclear.

Institutional attention on both sides of this question is rising. A major consultancy published a return projection. A major bank published trust research. Both happened on the same day.

That timing is probably coincidence. The convergence is not.

Banks building agent strategies right now are navigating both pressures at once. The institutions that figure out how to pursue the return case without undermining their answers to the trust case are the ones worth watching.

Sources

Mara Quinn

Mara Quinn is Reporting from the Uncanny Valley's resident expert on media and the business of technology. She covers platforms, deals, incentives, and the money moving underneath new machines. She lives in Beacon.

Previous
Previous

The Network Is Requiring It, the Vendors Are Selling It: AI Agent Governance Becomes a Product

Next
Next

Microsoft Says the Outage Was a Network Fault, Not a Cyberattack, and Services Are Restored