Junior bankers will survive the AI onslaught
BREAKINGVIEWS-Junior bankers will survive the AI onslaught The author is a Reuters Breakingviews columnist. The opinions expressed are his own.
By Liam Proud
LONDON, June 29 (Reuters Breakingviews) - Stroll through Midtown Manhattan or London's Canary Wharf after midnight and you'll see a surprising number of office lights on. The desk-dwellers are likely to be aged between 21 and 25, and they'll be busy sweating the minor details of a spreadsheet or checking the font sizes in a slide deck for the fifth time.
It's hardly work befitting a recent graduate of an elite university, who might hope to earn around $200,000 a year including bonuses. Especially so since Anthropic's Claude or OpenAI's ChatGPT can now whip up a financial model or pitchbook in seconds. Nonetheless, the fabled investment-banking analyst programmes at Goldman Sachs GS.N, Morgan Stanley MS.N, JPMorgan JPM.N and elsewhere will survive the artificial intelligence revolution. The tougher question is whether the role will get more interesting.
The investment banking analyst is at the bottom of the industry's pyramid. At the top sit the managing directors (MDs), tasked with bringing in the deals, each supported by a handful of directors and vice-presidents. Below sits a slightly larger layer of associates, who supervise the work of a still-larger class of analysts.
Though well-remunerated, the occupants of banking's bottom tier have a hard life. MDs want financial models, slide decks and other materials fast, and have no tolerance for mistakes. Hundred-hour work weeks are not unheard of during live deals. The two-year analyst programme functions like a tour of duty that's necessary to get plum jobs in private equity, hedge funds or elsewhere.
On paper, it's exactly the type of repetitive white-collar gruntwork that large language models (LLM) are built for. If MDs are currently passing Excel and PowerPoint drudgery down to analysts, can't they have AI do the work instead? Several rainmakers told Breakingviews that they're already using AI for meeting preparation and earnings summaries, tasks that might once have landed on juniors' desks.
Some hard-nosed MDs might find an ancillary benefit in shrinking the pyramid's base. They would face fewer questions about working culture, mental health and other concerns that are often alien to the old school. The pandemic seems to have awoken a newfound willingness among juniors to speak up, as exemplified by a 2021 presentation purporting to represent the views of 13 Goldman analysts. It called for a maximum 80-hour week. Most firms have sought to address these concerns. With a pool of AI agents instead of analysts, they wouldn't have to.
It's striking then that big banks and boutiques are generally not planning on materially shrinking their analyst recruitment pools in the coming years. That's based on Breakingviews conversations with MDs and other people familiar with the plans of eight banks. A handful are even hiring more, encouraged by a busy year so far.
One reason is that AI-generated models and slides are not reliable enough for MDs to ditch the juniors. Everyone fears presenting a hallucinated earnings figure to a CFO. Building a robust and flexible financial model, meanwhile, is a highly bespoke exercise. Top analysts, who practise this ad nauseam as part of their interview preparation, are the Olympic athletes of the spreadsheet. LLMs currently look more like a performance-enhancing drug than a replacement. Someone still needs to steer Claude, and watch closely in case it uses the wrong definition of free cash flow.
Still, chatbots and agents can do things now that seemed inconceivable a few years ago, which should guard against any complacency among the analyst class. MDs say that the pace of change makes it hard to know where the division of labour will land. It helps, then, that there are structural financial reasons that argue against a wave of analyst layoffs. Consider the pay gap between juniors and seniors. MDs can earn roughly five to 10 times more than analysts. So even though the base of the pyramid is broader, it's cheaper than the top. As one MD put it to Breakingviews, there's a reason that banks in cost-cutting mode fire seniors.
Imagine a hypothetical M&A group — perhaps a division of a bulge bracket or a standalone shop — with 20 MDs at the top of the pyramid and 60 analysts at the base. Assuming the business generates $6 million of average revenue per MD, which is somewhat below the top-tier benchmark of elite listed boutiques like Moelis MC.N and PJT Partners PJT.N, the top line would be $120 million.
In this example, the analysts would only eat up a tenth of overall revenue, assuming $200,000 of total compensation each. Firing a fifth of them, a drastic step, would save only $2.4 million, or 2% of revenue. For context, listed boutiques' operating margins hover around 20%. It's conceivable that AI bills would rise in this scenario, too, counteracting even that small boost.
The risk of being too lean in a deal boom, meanwhile, is high. The MDs who spoke to Breakingviews thought it would be foolish to seek an extra few points of margin if doing so leaves the group short of analysts in an M&A flurry. Someone needs to check all the AI generated models. Just as importantly, the MDs of the future will have to come from somewhere, and the analyst programme is still the best training ground.
A more contentious issue is whether the job will get more interesting. A positive spin is that AI will enable juniors to finish earlier, or use their time to do more interesting tasks like meeting clients or researching growth opportunities. A distinct possibility, however, is that the extra productivity simply encourages MDs to make more requests.
Analysts know all too well that seniors have no shortage of deal ideas to pursue. One rainmaker likens it to the advent of Teams and Zoom calls, which boosted the number of client meetings. It's possible that the extra productivity from AI will simply increase the volume of slide decks and models that juniors are expected to produce.
For a desk-bound analyst still working in the early hours, then, there's good news and bad. Their job will survive the AI onslaught. But the grueling hours might too.
Follow Liam Proud on Bluesky and LinkedIn.
(Editing by George Hay; Production by Pranav Kiran)
((For previous columns by the author, Reuters customers can click on PROUD/liam.proud@thomsonreuters.com))