Boardroom trust
88A CEO selling the company of their career hires a person they have known for years, not a model's output; the mandate decision is emotional, reputational and personal in ways no interface reaches.
Investment banking is unusually exposed to AI at the bottom and unusually protected at the top. The analyst layer's core output — first-draft models, pitch pages, diligence summaries, market updates — is text and structured data, exactly the material generative systems handle best, and every major bank has deployed internal AI assistants since 2023. The senior layer's product is trust, judgment and accountability, which no client yet buys from software.
The likeliest future is therefore not replacement but compression: smaller analyst classes producing the same output, a steeper apprenticeship in which juniors supervise machines instead of building everything by hand, and unchanged — possibly intensified — competition for the relationship franchise at the top. The open question is where the next generation of trusted seniors comes from if the training grind that formed the current one is automated away.
High
Share of the work a machine could do
A majority of current junior-banker hours go to tasks AI already performs credibly: assembling presentations, drafting first-cut models and comps, summarizing data-room documents, producing market-update materials. What resists automation is the senior franchise — winning mandates on trust, negotiating against humans, and signing regulated opinions someone must be liable for. The job survives; the pyramid beneath it narrows.
Scored from the tasks, not the job title. Lower is safer.
Jobs AI cannot take →What machines cannot take
Negotiation against humans
82Reading the other side's constraints, timing a concession, bluffing credibly across a table — adversarial, unrepeatable situations with hidden information remain firmly human ground.
Accountability & regulated opinions
75Fairness opinions, prospectus liability and licensing regimes all require an identifiable, insurable human or firm to be legally responsible; no regulator accepts a model's signature.
Confidential judgment
70The most valuable deal knowledge — who might sell, which board is fracturing, what a rival bid really was — lives in conversations that never reach any dataset a model can train on.
Syndication & placement relationships
62Placing billions of new securities still runs on human networks of reciprocal trust between banks and investors, built over cycles — though electronic bookbuilding keeps eroding the edges.
What they already take
Pitch-book assembly
85Drafting standard pages — company profiles, market overviews, credential slides — is already substantially automated by internal AI tools at major banks, cutting what took an analyst a night to minutes.
Diligence document review
80Machine reading of data-room contracts, filings and financials for red flags now outpaces junior human review on speed and consistency; law and accounting firms deploy the same systems on the same deals.
Market updates & internal memos
78Weekly sector updates, earnings summaries and deal-committee first drafts are text generation from structured inputs — precisely the workload banks report moving to AI assistants first.
First-cut models & comps
72Template-driven three-statement models, comp spreads and league tables build increasingly from software; the judgment calls — adjustments, normalizations, what the number implies — still get checked by hand.
How the work is changing
The analyst pyramid narrows
Banks openly discuss smaller incoming classes as AI absorbs production work; the apprenticeship shifts from building everything by hand to supervising machine output — faster to competence for some, a lost training ground for others.
Boutiques take the advisory share
Independent advisory firms — Evercore, Centerview, PJT, Rothschild & Co — have spent two decades winning fee share from full-service banks on the argument that advice unbundled from lending is cleaner; the trend strengthens as execution work commoditizes.
Private markets eat the pipeline
With companies staying private longer and private credit passing roughly $1.7 trillion in assets by 2024, the classic IPO-and-public-M&A machine now competes with sponsor-to-sponsor deals, continuation funds and direct lending — reshaping which bankers matter.
The hours finally face limits
After the 2021 Goldman analysts' leaked survey and renewed scrutiny in 2024, JPMorgan capped most junior weeks at roughly 80 hours and rivals added tracking and protected days — modest ceilings, but the first structural concessions in a generation.
New jobs branching off
Private-credit originator
Structuring loans at direct-lending funds — the fastest-growing corner of finance — using classic banking skills outside the regulated banking perimeter, often for former M&A and leveraged-finance bankers.
Secondaries & private-capital adviser
Advising on stakes in private funds, GP-led continuation vehicles and NAV financing: a specialty that barely existed in 2010 and now supports dedicated teams at every major advisory firm.
Transition & infrastructure finance specialist
Raising and structuring capital for energy transition, data centers and infrastructure — the deal category behind GIP's $12.5 billion sale to BlackRock — blending project finance, policy fluency and classic coverage banking.
Deal-technology & AI diligence specialist
Building and supervising the AI systems that now draft materials and read data rooms: a hybrid banker-engineer role inside banks and the growing vendor ecosystem around M&A execution.
Outlook
The historical parallel bankers themselves cite is the trading floor: electronic execution eliminated thousands of broking and market-making jobs after the 1990s, yet the banks' revenues and their senior franchises survived by moving up the value chain. Advisory banking is now beginning the same migration — production automates, judgment concentrates, and the people paid the most are those whose names win the mandate.
The structural risk is generational. Every current managing director was formed by the grind the industry is now automating; if analysts no longer spend years inside the numbers, banks must invent new ways to grow people who can price a company from first principles and hold a boardroom. Firms that solve that apprenticeship problem will own the next era of the franchise.
What does not change is the underlying demand. Companies will keep merging, splitting, failing and raising capital, and someone trusted will stand between the parties when they do — as someone has since Florence. The seats will be fewer, the tools unrecognizable, and the product, trusted intermediation at scale, exactly what it was in 1397.
Keep exploring
More in Business & Finance
🚀
Entrepreneur
The person who starts the company — spotting the gap, bearing the risk and answering for payroll, from Assyrian caravan financiers to venture-backed founders.
AI-resistant 88 🧾Accountant
The keeper of the books: heir to a craft so old it invented writing itself, now negotiating with the software built to automate it.
AI-resistant 35 📣Marketer
The professional who creates demand — from Pompeii's painted walls and P&G's 1931 brand-man memo to the auction-driven feeds of the digital era.
AI-resistant 38 🗺️Product Manager
Decides what a company should build next, and why — turning customer needs, business goals and engineering limits into one shared plan nobody else fully owns.
AI-resistant 50 💱Economist
The scholar of scarcity — from Adam Smith's pin factory to the central-bank decision room, still asked to predict what no model fully captures.
AI-resistant 62