How is Artificial Intelligence reshaping the role of Investment Banking?

Inside 8 global financial institutions (JPMorgan, Morgan Stanley, and Bank of America, MUFG, Itaú, KBW, Santander and SMBC)

By Felipe Gonzalez

Artificial intelligence is becoming ubiquitous across global finance, and few sectors of the industry are changing faster than investment banking. Investment banking at its essence has always been two jobs merged together: an apprenticeship of analysis at the bottom and a craft of a relationship at the top, the trust that wins one. Around both sits a culture that decides who gets hired and what they need to know. Artificial intelligence is now pressing on every layer of that structure at once, and no two banks are absorbing it the same way. Many professionals are now left wondering: will AI replace investment banking and in what ways will this profession get reshaped?

Over one week in July, I put the same questions to senior bankers at eight global banking institutions (MUFG, Itaú, KBW, Santander, JPMorgan, SMBC, Morgan Stanley, and Bank of America): how is artificial intelligence reshaping the role of the investment banker, and the financial culture built around that role? And ultimately, what still requires human judgement?

Their answers converged on three patterns. Which banks embrace AI is governed by who controls the data, not by national culture. Its effect on jobs so far is a change in what banks hire for, not yet how many they hire. And at all eight the consensus was all ultimately the same: trust.

The critical disparity across the eight banks is one of restrictions and accessibility. Some institutions restrict AI because confidential information cannot be placed in a data center that is not their own – most notably MUFG, where the concern is sharpened by the geopolitics between the United States and Japan. A managing director there commented how the bank technically has Copilot but described it as of little use, so much of the work remains manual; the bank, she acknowledged, is falling behind.

The clearest evidence that control, not culture, decides the matter sits across town at SMBC, which is also Japanese. Its Copilot model is trained on the bank and its own data: ask it about a coworker and it knows. And it operates with almost no restrictions at all.

Other banks never faced the ceiling in the first place. With proprietary systems or close partnerships, confidential information never leaves their control: JPMorgan through its internal AI platform, Bank of America through a specialized tool, Santander through its partnership, Itaú with no restrictions of any kind. But access is only half the divide; the tools themselves are not equal. The bankers I spoke with consistently praised the frontier systems while dismissing basic Copilot deployments as the bare minimum.

A senior banker at KBW tied this divide to money. Building or owning a serious AI system is expensive, and this industry has been concentrating for decades: the number of FDIC insured institutions in the United States has fallen from more than 12,600 at the end of 1994 ("Quarterly Banking Profile Graph Book") to roughly 4,278 today ("FDIC-Insured Institutions Reported Return on Assets"). In his view, AI will accelerate the shrinking, because the largest banks can afford the most powerful systems and pull further ahead with them. However one twist noted is that some of the newest competitors are not banks at all; companies such as Apple and X are pushing into financial services and they arrive with AI of their own already assembled.

So far, the clearest effect in jobs and hires is not a smaller intake but a different one. At Santander, a banker recalled a recent meeting with Wells Fargo and quoted the takeaway directly, "They didn't reduce the hire size, but what did change was the criteria." The skills that once was the essence of investment banking, i.e. financial modelling, no longer are quintessential for the role, as the software handles it. At KBW, the forecast went further. The same senior banker described a proposal to cut analyst classes roughly in half, from thirty to fifteen, and expects much of the routine accounting style work to disappear almost completely. What firms screen for in first year analyst has shifted accordingly: critical thinking, curiosity and individuality. As AI assimilates the monotonous parts of the job, hiring will heavily focus on the qualities artificial intelligence lacks.

The rest of the eight fill out an honest range. MUFG is hiring fewer juniors even as demand grows. JP Morgan has seen little change so far but expects office sizes to fall over the long term. Bank of America reports little effect at all so far. The general consensus, however, is that AI is not firing bankers yet. That being said, it is rewriting the entry requirements for hires.

On one point, all eight banks converged: the part of the job AI cannot touch is the relationship. Trust, the bankers agreed, is the core of the profession, and it cannot be automated. One put it flatly: the only thing that cannot change is relationships. AI is useful for the multifaceted, tedious layers of the work, drafting presentations, running financial models, but ultimately the client is not buying the model. The client is buying the person who stands behind it, and that person still has to genuinely know the material.

Morgan Stanley is the purest case of that philosophy. The bank uses relatively little AI and has no plans to build a system of its own; there, a banker described the technology's role in strictly practical terms. A financial model that once occupied a team for days can now be produced by one person in five to ten minutes. Everything past that compression, in the firm's view, a banker hones alone. The knowledge has to be real, because trust depends on it.

The rate at which this profession is reshaping varies institution by institution, but there is no question that the profession is evolving. Which banks embrace AI is governed by who controls the data, not culture; what they hire for has changed even where headcount has not; and what remains scarce, everywhere, is trust. As the KBW banker put it, "The profession will have a rough patch, but it will be beneficial long term when everything adapts."

Sources & Citations

"FDIC-Insured Institutions Reported Return on Assets of 1.26 Percent and Net Income of $80.5 Billion in First Quarter 2026." Federal Deposit Insurance Corporation, 27 May 2026, www.fdic.gov/news/press-releases/2026/fdic-insured-institutions-reported-return-assets-126-percent-and-net.

"Quarterly Banking Profile Graph Book: Second Quarter 2004." Federal Deposit Insurance Corporation, 2004, www.fdic.gov/system/files/2024-07/qbpgr_61.pdf.