Will AI Replace Financial Analysts? The 2026 Risk Picture
Financial analysts and securities traders (ISCO 33) score 7.5/10 on AI exposure - high, and the disruption is not pending. Bloomberg Terminal's AI capabilities, FactSet Cognit, and earnings call AI summarisers are already eliminating the routine work of junior analysts: scraping filings, building comps, writing first-draft research notes. BLS OEWS May 2025 counts 330,000 US financial analysts at a median of $99,890 per year.
Key findings
- Financial analysts (ISCO 33): AI exposure 7.5/10 - high; junior analyst functions already being automated
- 330,000 US financial analysts, median $99,890/yr (BLS OEWS May 2025)
- Globally: approximately 5 million workers in ISCO 33 (ILO ILOSTAT 2024)
- Highest risk: junior analysts doing comps, data scraping, earnings summaries (8.5/10)
- Safest: portfolio managers and relationship bankers with fiduciary client accountability (5.0-5.5/10)
What AI is already doing in financial analysis
The disruption to financial analysis is not a future scenario. Bloomberg launched Bloomberg AI in 2024, integrating large language model capabilities directly into the Terminal used by virtually every institutional analyst globally. FactSet's Cognit product generates first-draft equity research notes from structured data. Earnings call AI - deployed by platforms including Sentieo and AlphaSense - automatically transcribes, summarises, and flags management tone shifts across thousands of quarterly calls simultaneously.
These are functions that previously required analyst time: reading transcripts, building financial models from 10-K filings, scanning for peer comparable data. The ISCO 33 score of 7.5/10, derived from ILO ILOSTAT (CC BY 4.0) and cross-referenced against Frey-Osborne (2013), OECD (2019), and IMF (2024) task-automation research, reflects the high proportion of routine data processing in the role's task bundle.
Goldman Sachs disclosed in 2023 that AI had automated a significant proportion of its initial public offering prospectus generation. Morgan Stanley's OpenAI integration provides advisers with instant access to 100,000 research documents. The institutional adoption of AI in financial analysis is documented, named, and commercially deployed - not speculative.
AI exposure by finance role
| Role | AI score | US workers (BLS) | Median wage |
|---|---|---|---|
| Junior equity / credit analysts (sell-side) | 8.5/10 | included in 330k | $75,000 |
| Financial analysts (broad BLS category) | 7.5/10 | 330,000 | $99,890 |
| Insurance underwriters | 7.5/10 | 116,000 | $79,840 |
| Securities and commodities traders | 7.5/10 | 62,000 | $98,030 |
| Budget analysts | 6.5/10 | 49,000 | $82,260 |
| Portfolio managers (buy-side) | 5.5/10 | included above | $131,000 |
| Personal financial advisers (retail) | 5.0/10 | 335,000 | $99,580 |
| Investment bankers (M&A, DCM) | 6.0/10 | incorporated | $140,000+ |
Source: BLS OEWS May 2025 (employment and wage data). AI scores are WorldJobsData estimates per ISCO-08 group 33. Sub-role scores reflect task-level analysis within the group.
Why junior analysts are the exposed tier
The task bundle of a junior analyst in a sell-side equity research team is: pull financial data from databases, build or update a discounted cash flow and comparable company model, read and summarise earnings call transcripts, draft the first section of a research note, and check competitor pricing. Every one of these tasks is now either fully automatable by existing deployed AI systems or substantially automatable with human review.
The task bundle of a senior portfolio manager is different in kind: build and defend a thesis with the fund's investment committee, manage client expectations during drawdowns, decide whether a macro event invalidates a position, and maintain relationships with management teams that provide non-public insight (within legal limits). These tasks involve accountability, judgment under uncertainty, and relationship capital. They are not task-list execution - they are fiduciary decisions that require a named, accountable human.
The finance industry has a historical analogue: algorithmic trading eliminated the floor trader. It did not eliminate the quantitative researcher who designs the algorithm. AI is eliminating the analyst doing structured data retrieval. It is not yet eliminating the analyst who constructs a structural thesis on why an industry is mis-priced - though that function is also being augmented.
Safest finance roles from AI
| Role | AI score | Why lower risk |
|---|---|---|
| Portfolio managers (institutional) | 5.5/10 | Fiduciary accountability, committee decision-making |
| Relationship bankers (corporate) | 5.0/10 | Multi-year trust, non-public information access |
| Personal financial advisers | 5.0/10 | Client trust, behavioural coaching, liability |
| Forensic accountants | 4.5/10 | Adversarial investigation, legal testimony |
| Venture capital investors | 4.5/10 | Network access, founder judgment, portfolio governance |
What this means for the next 3 years
The 1-3 year window in financial analysis is the compression of junior analyst headcount at major institutions. Goldman Sachs, JPMorgan, Morgan Stanley, and their equivalents are running efficiency programmes that explicitly cite AI as the driver. The entry-level analyst cohort that has historically numbered 200-400 per major bank per cycle is being reduced. This is not a forecast - it is a documented trend visible in public hiring data and disclosed in earnings calls.
The medium-term implication is that the path from junior to senior analyst compresses for those who remain. An analyst who can use AI to do the work of three junior analysts while adding the judgment of a senior analyst is valued higher. The credential shift is from "can build a model" to "can interpret a model and defend a thesis."
See financial sector AI risk by country
Finance and insurance professional concentration varies significantly across 206 countries. Countries with large financial services sectors have more exposure concentrated in this category.
Open the Explore Tool →Was this analysis useful?