IMAGE: ISTOCK
A GENERATION ago a bright graduate could expect to spend their first two years in finance doing heavy routine work: building spreadsheets, summarising earnings calls, drafting prospectuses and similar tasks. This process acted as both initiation and training. Today that process has faded away. At Goldman Sachs the acceptance rate for internships has fallen to 1.16%. The bank last year received a quarter of a million applications for fewer than 2,900 places. The entry-level job, as it was once known, has disappeared.
The reason is a lasting structural change rather than a cyclical slump. Generative artificial intelligence has moved from a novelty to an everyday work tool. Firms now say these systems can complete about 95% of the early work on documents such as IPO prospectuses, a task that previously kept junior analysts busy for days. Productivity research points to a 33% improvement for each hour of AI use. In practical terms one analyst can now do the work once spread across several people. The result is a sharp separation between growth in assets and growth in staff numbers. BlackRock, while increasing assets under management, has carried out three rounds of job cuts since early 2025, each removing around 1% of its global workforce. The roles being cut are usually broad, process-driven jobs that once served as entry points to the industry.
The technology itself has changed. Early large language models acted as passive tools. The new generation consists of “agentic” systems that can think, plan and complete connected tasks independently. Goldman Sachs uses Claude, an AI built by Anthropic, as a digital colleague for trade reconciliation and compliance checks. Bridgewater Associates is trying something more ambitious: creating an “artificial investor” at its AIA Labs, designed to study why markets move, rather than simply spotting patterns in data. In private equity firms say AI can raise qualified deal flow by 40% and reduce due-diligence time by more than a third. The routine analytical work—the training ground for juniors—is being automated at scale.
What does this mean for aspiring financiers? Career paths are splitting in two. The traditional apprenticeship model is breaking down. Applicants now have to arrive with skills once learned on the job: comfort with Python, command of valuation methods and the ability to question AI outputs. As one industry watcher puts it, patience for training has narrowed; graduates must add value from the first day. Recruitment has therefore shifted towards candidates trained in data science, physics and engineering, alongside finance.
Even so, the human role grows in importance rather than shrinking. A new idea, known as the “Last 5 Percent” rule, has taken hold. If AI can deliver 95% of an analysis, the value moves to the final 5%: ethical judgement, strategic narrative and understanding of clients. The role changes from number-cruncher to overseer of exceptions and creator of reports. This makes qualifications like the Chartered Financial Analyst designation more valuable, as they focus on ethics and difficult judgement—areas where algorithms are weak.
The industry’s gatekeepers have closed the entry point. The basic work that taught generations of junior professionals is now done by machines. Those who join the industry will handle more judgement and oversight, but there will be fewer of them. The path into finance still exists, but its bottom step has been removed. ■
