The Human Moat In The Age of AI

We read an interesting article in the The Wall Street Journal which covered the discussion of a growing dilemma on Wall Street: Banks are rapidly adopting AI tools that can build financial models, draft presentations, summarize filings, analyze transactions, and generate first-pass deal materials in a fraction of the time previously required by junior bankers.

And the key question is…what are the implications?

For decades, investment banking has operated on a relatively simple apprenticeship model. Analysts spend countless hours building models, formatting presentations, conducting research, and preparing materials for clients. Associates review their work. Vice presidents coordinate teams and managing directors bring in business.

The system has often been criticized for its inefficiency and brutality. Yet it serves an important purpose. The analyst who spent three years tearing apart financial statements eventually became the associate who could evaluate a business instinctively. The associate eventually became the vice president who could advise management teams. The vice president eventually became the managing director who could walk into a boardroom and convince a CEO to pursue a billion-dollar acquisition.

The tedious work is really not just production but training. Now AI threatens to remove much of that training ground.

This creates a fascinating paradox. The more effective AI becomes at performing junior-level tasks, the less need there is for junior bankers. But if banks hire fewer junior bankers, where will the next generation of senior bankers come from?

Wall Street’s concern is not really about modeling. Modeling has always been a commodity skill. The concern is about rainmakers.

Every investment bank ultimately lives and dies by its ability to develop trusted advisors who can originate deals, maintain relationships, and persuade clients to act. AI can build a discounted cash flow model. It cannot yet persuade a hesitant CEO to sell a family business. It cannot read the emotions around a boardroom table during merger negotiations. It cannot build trust over twenty years of interactions. At least not yet.

The next five years are therefore likely to produce a strange outcome. I don’t believe investment banking jobs disappear but I think they split into two very different categories.

The first category consists of highly repeatable analytical work: Financial modeling, presentation creation, market screening, comparable company analysis, industry research, diligence coordination, document drafting, and much of the work traditionally performed by analysts and associates will increasingly be automated. The number of people required to produce the same output may decline substantially. A team of five may become a team of two equipped with powerful AI systems. Goldman Sachs’ David Solomon has already noted that work which once required multiple people and weeks of effort can increasingly be completed in minutes.

The second category consists of relationship-intensive work. Strategic advice, negotiations, trust building, client management, judgment, storytelling, and transaction leadership become more valuable because they are harder to automate. In fact, their relative importance may increase as technical execution becomes commoditized.

Ironically, AI may make the best bankers even more powerful.

Think of what happens when a managing director who previously managed ten clients can suddenly manage thirty because AI handles much of the analytical burden. The top performers become more scalable. Their reach expands. Their economics improve. The gap between elite bankers and average bankers may widen dramatically.

Historically, investment banks focused on larger clients because servicing smaller companies was uneconomic. If AI reduces the cost of analysis, presentation preparation, and transaction execution, banks may be able to profitably serve thousands of middle-market and lower-middle-market companies that were previously ignored. Some observers have suggested that AI could dramatically expand the addressable market for advisory services by lowering delivery costs.

If that happens, we may actually see fewer bankers per transaction but more transactions overall.

My base case is that analyst and associate hiring declines by perhaps 20% to 40% over the next five years at major banks. The remaining junior bankers will likely oversee AI systems rather than perform every task manually. Their role will increasingly resemble editor, verifier, and orchestrator rather than Excel and PowerPoint ninja.

What is worrying is the long-term talent pipeline: Imagine a generation of bankers who never built a model from scratch, never spent nights debugging assumptions, never learned through repetition. They may become excellent AI operators but weaker financial thinkers.

Considering what happens in other professions, we know that pilots still learn to fly manually even though autopilot systems exist. Surgeons still learn anatomy despite robotic assistance. Investment bankers may need a similar approach. Some experiences shouldn’t be outsourced because they build judgment.

So, I think moving forward, leaders in the bank should create an environment where junior bankers spend less time moving logos on PowerPoint slides and more time participating in client meetings. They should learn negotiation earlier. They should understand industries more deeply. They should receive training in communication, psychology, persuasion, and strategic thinking. AI can generate information. Humans still create conviction.

The traditional route of analyst to associate to vice president to managing director was built around information processing. Future career ladders may instead be built around judgment accumulation. The banker of 2031 may spend far less time building models and far more time interpreting them.