will AI replace financial analysts?

Fifteen of the 26 tasks in your role have zero AI penetration according to O*NET task data. That's not a rounding error. These are the tasks that define whether you're a real analyst or a spreadsheet jockey, and they all share something in common: they require you to be present, trusted, and accountable.

The most defensible work you do is the stuff that happens in rooms. Visiting company facilities to assess them as investments. Sitting across from a CFO and reading whether they're hiding something in the way they answer a question. Collaborating with lawyers and accountants on deals where the details aren't in any database yet. Attracting corporate clients by building relationships over years. An AI can't walk the factory floor with you. It can't tell you the CEO looks exhausted and the operations manager won't meet your eye.

Supervising and mentoring junior analysts is also fully yours. Training someone to develop judgment, not just run models, is a deeply human act. The same goes for advising clients on capitalization strategy, where timing, relationships, and risk tolerance are personal and contextual in ways no model can fully capture. If you're specialising in green finance or ESG instruments, that space is still being defined by people, not algorithms. The criteria are contested, the standards are shifting, and clients need a human to help them navigate the ambiguity. That's where your judgment earns its keep.

The biggest problem with AI in financial analysis isn't that it's wrong sometimes. It's that it's confidently wrong in ways that are hard to catch. Tools like Bloomberg Terminal's AI features and ChatGPT can hallucinate earnings figures, cite non-existent filings, or produce plausible-sounding summaries of companies that quietly misstate key ratios. In a field where a single wrong number can drive a bad trade or a flawed report, that's a serious liability.

AI also can't read primary sources the way you can. It processes text, but it doesn't catch the tone shift in an earnings call transcript where a CEO dodges a question three times. It doesn't notice that a company's 10-K has quietly changed its revenue recognition language from the prior year. Pattern recognition and language models are not the same as the forensic skepticism a good analyst develops over a decade of reading filings.

There's also a regulatory accountability gap. When an AI-assisted report gets a recommendation wrong, who's liable? The firm's compliance frameworks still require a named human analyst to sign off. That's not a technicality. It means the human judgment layer isn't optional, and regulators like the SEC have been explicit that AI-generated investment recommendations carry the same disclosure and accountability standards as human ones. The tool doesn't take the blame. You do.

The analytical tasks that used to eat your week are now largely automated. Pulling earnings data, screening securities by quality metrics, scanning financial publications and government agency releases for relevant developments, these are things AI handles quickly and at scale. Bloomberg Terminal now has a built-in AI layer that summarises earnings calls, flags material changes in filings, and cross-references data across markets. FactSet's AI tools can generate first-draft equity research notes with financial ratios, comps, and sector context already populated.

For report generation, tools like Visible Alpha and Koyfin have made it possible to build initial models and charts in minutes rather than hours. Drawing charts and graphs, once a meaningful time sink, is now nearly automatic. Interpreting price and yield trends, identifying risk signals in macro data, even producing a first draft of a written sector report, all of this sits in the high-penetration zone. Microsoft Copilot for Finance, integrated into Excel and PowerPoint, can build variance analyses and presentation slides directly from your data without manual formatting work.

For client-facing materials, Gamma and Beautiful.ai are being used to generate presentation decks from structured data and bullet points, which cuts the time to produce a client update from a half-day to under an hour. The AI isn't making the investment call. But it's handling the scaffolding around the call: the data gathering, the initial interpretation, the chart building, the draft narrative. That's a real and genuine shift in where your hours go.

The sequence of your day has changed more than the nature of your judgment. You're spending less time in spreadsheets running screens and more time stress-testing conclusions that AI drafted for you. The work isn't slower, but the error-checking phase is now more important than the build phase.

What's genuinely gone, or nearly gone, is the data-gathering ritual. Analysts used to spend hours each week pulling figures from filings, cross-referencing news sources, and building initial financial models from scratch. That time has compressed significantly. According to a 2024 survey by CFA Institute, analysts using AI-assisted research tools reported saving an average of 6 to 8 hours per week on data aggregation and initial report drafting.

What hasn't changed at all is the client relationship work. The calls, the meetings, the face-to-face assessments, the collaborative work with lawyers and bankers on live deals. None of that is faster or different because of AI. You still spend roughly the same time on it. The difference is that it now represents a larger share of your total week, because the surrounding admin has compressed. That shift is actually a career improvement for most analysts. The work that's left is more interesting.

The BLS projects 5.7% growth for financial analysts between 2024 and 2034, which translates to roughly 25,100 job openings per year against a current base of 368,500 employed analysts. That's modest but positive growth. The question worth asking is whether that growth holds as AI takes on more of the analytical workload.

The honest answer is that headcount growth will probably slow at the junior end. The tasks that used to require a team of three analysts, one pulling data, one building models, one drafting reports, can now be handled faster with one analyst and the right tools. Banks and asset managers are already restructuring junior analyst pipelines. Goldman Sachs and JPMorgan have both publicly discussed AI tools cutting the time required on initial research work by 30 to 50%. That doesn't mean fewer analysts long-term, but it probably means fewer entry-level seats doing purely mechanical work.

At the senior end, demand is holding. The Anthropic Economic Index rates financial analysts as an "amplified" role, meaning AI is increasing output per analyst rather than replacing the function. Portfolio managers, sector leads, and client-facing analysts who can take AI-generated analysis and turn it into a defensible recommendation are in more demand, not less. The growth in assets under management globally, forecast to exceed $145 trillion by 2025 according to PwC, means there's more financial complexity requiring human oversight, even as the tools get better.

The AI exposure score of 76% is likely to rise slightly over the next five years, not stay flat. The tasks currently in the "speeds up" category, like client presentation creation, will probably cross into full automation territory as multimodal AI gets better at reading context and formatting outputs for specific audiences. Scenario modelling, which still requires human input to frame the assumptions, could also shift as AI gets better at generating plausible alternative scenarios from structured inputs.

But a genuine threat to the core of this role would require AI to do things it currently can't. It would need to build trusted relationships with institutional clients. It would need to walk a facility, read a management team, and make a judgment call that isn't in any dataset. It would need to be legally accountable for a recommendation. None of that is close in a five-year window, and it's genuinely unclear in a ten-year one. The most realistic future is a smaller number of analysts doing higher-value work, not a collapse of the profession. If you're doing the relationship and judgment work now, your position in that future is stronger than it sounds.

The clearest move you can make is to deliberately shift your time toward the zero-penetration tasks. That means volunteering for facility assessments, client advisory work, cross-functional deal teams, and mentoring responsibilities. These aren't just safe from AI. They're the tasks that build the reputation and judgment that make you irreplaceable at a specific firm, to specific clients.

The second move is to get fluent with the AI tools without becoming dependent on them. Knowing how FactSet's AI note generator works, where it makes errors, and when to override it is a real skill. Firms are starting to value analysts who can supervise AI outputs, not just produce their own work. That's a different skill set than traditional analysis, and it's worth building now. Consider the CFA Institute's AI in Investment Management certificate, which launched in 2024 specifically to address this gap.

If you're earlier in your career, specialise in areas where data is thin, contested, or private. ESG and green finance instruments are a good example. The criteria for what counts as a green investment are still being argued over by regulators, asset managers, and NGOs. That ambiguity is where human judgment earns its premium. Private markets, early-stage company analysis, and cross-border deals in emerging markets all involve information gaps that AI struggles with. The further you get from clean, structured, public data, the safer your position. Build expertise in those spaces before the data catches up.