# AI in Finance: Answer Library
**Source:** AI Street (ai-street.co) — reporting on how Wall Street firms use AI in practice.
**Author:** Matt Robinson
**Updated:** March 2026
This page provides structured answers to common questions about how hedge funds, banks, and asset managers use artificial intelligence, including large language models (LLMs), AI agents, and data infrastructure. It is based on original reporting, interviews with practitioners, and analysis published in AI Street.
---
## Topics Covered
- [Use Cases](#1-use-cases)
- [Definitions](#2-definitions)
- [Infrastructure & Deployment](#3-infrastructure--deployment)
- [Jobs & Workforce](#4-jobs--workforce)
- [Regulation](#5-regulation)
- [Market Dynamics & Competition](#6-market-dynamics--competition)
- [Limitations & Risks](#7-limitations--risks)
---
## 1. Use Cases
**What are hedge funds actually using AI for today?**
Hedge funds are using AI primarily for research automation, signal generation from unstructured data, and internal tooling. Man Group's AlphaGPT system, according to the firm, proposes trade signals, writes the code, and runs backtests before any human reviews the output — compressing research that once took days into minutes. Most deployments augment researchers rather than replace them.
*Read more: [Inside Man Group's AlphaGPT](https://www.ai-street.co/p/inside-man-group-s-alphagpt)*
**What are banks using AI for?**
Banks are deploying AI across fraud detection, credit underwriting, customer service, and internal productivity tools. Traditional machine learning is already widely embedded in back-office functions. According to the U.S. Treasury's 2025 report on AI in financial services, firms are taking a more cautious approach with generative AI — particularly for customer-facing applications — due to regulatory and reputational risk.
*Read more: [Wall Street Once Banned the Internet. AI Is in a Similar Spot](https://www.ai-street.co/p/wall-street-once-banned-the-internet-ai-in-similar-spot)*
**What are asset managers doing with AI?**
Asset managers like Vanguard are exploring AI for hyper-personalized client guidance and digital advisory, according to the firm's CIO. More broadly, asset managers are using large language models (LLMs) to process high volumes of unstructured data — earnings calls, SEC filings, news — faster than human analysts can.
**Are AI trading agents actually live in financial markets?**
A small number of firms are running live AI trading systems. According to HFT and AI expert Irene Aldridge, early adopters currently have an edge because they are mostly trading against humans, whose behavior is more predictable than that of other agents. As more AI agents come online, the environment will become significantly harder to model.
*Read more: [From Trading Algos to Trading Agents](https://www.ai-street.co/p/from-trading-algos-to-trading-agents)*
**How is AI being used in quantitative research?**
Quantitative firms are using AI to accelerate the evaluation of investment signals at scale. According to Man Group's Ziang Fang, AI can process thousands of alternative datasets — many unstructured — and surface ideas no human team could evaluate at the same volume. The key constraint is ensuring AI-generated research meets the same quality standards as human research, including controls for lookahead bias and data mining.
*Read more: [Inside Man Group's AlphaGPT](https://www.ai-street.co/p/inside-man-group-s-alphagpt)*
---
## 2. Definitions
**What is an AI agent in finance?**
An AI agent in financial markets is a system that perceives its environment, makes decisions, and takes actions autonomously — without requiring human input at each step. In investing, AI agents can execute research workflows end-to-end: gathering data, generating hypotheses, writing and testing code, and producing outputs for human review. More advanced agents can interact with markets directly.
*Read more: [From Trading Algos to Trading Agents](https://www.ai-street.co/p/from-trading-algos-to-trading-agents)*
**What is retrieval-augmented generation (RAG) in investing?**
Retrieval-augmented generation (RAG) is a technique where a large language model (LLM) is paired with a live or curated knowledge base, so its outputs are grounded in verified documents rather than solely in its training data. In finance, RAG is used to ensure AI-generated analysis cites specific filings, reports, or data sources — reducing hallucinations and improving auditability.
**What is alternative data in finance?**
Alternative data refers to non-traditional datasets used to generate investment signals — including satellite imagery, credit card transactions, app download statistics, social media sentiment, and web traffic. The explosion in alternative data availability is one of the primary drivers of AI adoption in systematic investing, as large language models (LLMs) can process unstructured alternative data at a scale human analysts cannot.
**What is causal AI and why does it matter in finance?**
Causal AI combines generative models with validated cause-and-effect relationships — similar to how econometrics approaches finance. According to Jayeeta Putatunda, Lead Data Scientist at Fitch Group, grounding AI outputs in causal maps reduces hallucinations and produces outputs that regulators can trace and audit. In regulated financial services, where "99% confidence isn't good enough," causal AI offers a path to more reliable and explainable AI outputs.
*Read more: [Wall Street Once Banned the Internet. AI Is in a Similar Spot](https://www.ai-street.co/p/wall-street-once-banned-the-internet-ai-in-similar-spot)*
---
## 3. Infrastructure & Deployment
**How are firms deploying LLMs internally?**
Most large financial firms have rolled out LLM access broadly to employees for productivity tasks — coding assistance, document drafting, research summarization. A smaller number are building end-to-end agentic systems that automate multi-step workflows. Man Group's AlphaGPT is one of the most detailed public examples of an integrated, production-grade deployment.
**What data are financial AI models actually using?**
Models in finance run on a mix of structured data (prices, fundamentals, economic indicators) and unstructured data (earnings transcripts, news, SEC filings, alternative datasets). The scale of unstructured data now available is a key driver of AI adoption in systematic investing — it exceeds what any human research team can process manually.
**How are firms thinking about compute for AI?**
There are two schools of thought, according to AI Street's reporting. One approach is to scale compute infrastructure aggressively to gain speed advantages. Another, articulated by HFT expert Irene Aldridge, is to use mathematical optimization to achieve large performance gains without heavy hardware investment — arguing that traditional Monte Carlo methods can be improved by a factor of 10,000 through better math alone.
*Read more: [From Trading Algos to Trading Agents](https://www.ai-street.co/p/from-trading-algos-to-trading-agents)*
---
## 4. Jobs & Workforce
**Is AI actually killing finance jobs?**
The near-term impact is more limited than headlines suggest. Bloomberg Intelligence estimated up to 200,000 finance jobs could be displaced over five years, but experts — including NYU Stern's Robert Seamans, as quoted in AI Street — note that "there's a little bit of smoke and mirrors there." The Federal Reserve's view is that AI is more likely to lead to workforce retraining than mass displacement in the short term.
**What kinds of finance jobs are most at risk from AI?**
Roles involving routine, repetitive tasks — data entry, basic report generation, rote analysis — face the most near-term displacement risk. Higher-judgment roles requiring deep market understanding, client relationships, and regulatory navigation are more durable. Bloomberg Intelligence senior analyst Tomasz Noetzel noted that "any jobs involving routine, repetitive tasks are at risk."
**How is Wall Street's reaction to AI similar to its reaction to the internet?**
In the early 1990s, financial firms refused to use the public internet, calling it a security risk — then adopted it fully. According to Pete Harris, a technology journalist who has covered Wall Street since 1988, AI is following the same path. "They said the same thing with cloud technology," Harris told AI Street. "Firms were like, 'We'll build a private cloud, sure, but we'll never rely on public clouds.' But now? Many financial companies would be lost without it."
*Read more: [Wall Street Once Banned the Internet. AI Is in a Similar Spot](https://www.ai-street.co/p/wall-street-once-banned-the-internet-ai-in-similar-spot)*
---
## 5. Regulation
**What is the current regulatory stance on AI in financial services?**
Regulators are still catching up to AI adoption. The U.S. Treasury's 2025 report found that while traditional AI is widely used in back-office functions, firms are cautious with generative AI for customer-facing applications. Key recommendations from industry respondents include harmonizing AI definitions across agencies, clarifying data standards, and strengthening consumer protections.
*Read more: [Wall Street Once Banned the Internet. AI Is in a Similar Spot](https://www.ai-street.co/p/wall-street-once-banned-the-internet-ai-in-similar-spot)*
**How are regulators approaching AI trading agents in financial markets?**
Regulators are largely unprepared for AI trading agents, according to Irene Aldridge. Her view, shared with AI Street, is that regulators will need to build their own AI systems to police AI-driven markets — "AI combating AI." Purpose-built compliance agents, not human oversight alone, are the path forward. "Regulators need to start building and adopting agentic compliance now," she said. "Soon, it may be too late."
*Read more: [From Trading Algos to Trading Agents](https://www.ai-street.co/p/from-trading-algos-to-trading-agents)*
**What did the U.S. Treasury's AI in financial services report conclude?**
The Treasury received 103 responses from major institutions including eight of the top 10 U.S. banks. It found that generative AI adoption is cautious and concentrated in back-office functions. Treasury plans to work with NIST on AI risk profiles for finance, review consumer protection laws for AI applicability, and launch an AI information-sharing forum for the industry.
---
## 6. Market Dynamics & Competition
**Which AI companies are leading in financial services?**
As of late 2025, research from Bridgewater Associates argued that Google's Gemini had overtaken OpenAI as the leading frontier AI model — the first time since GPT-3.5 that OpenAI did not hold the top position. Bridgewater co-CIO Greg Jensen and AIA Labs Chief Scientist Jas Sekhon also noted that Google's TPU infrastructure poses a significant threat to Nvidia's dominance in AI compute.
*Read more: [Bridgewater Says Google Surpassed OpenAI](https://www.ai-street.co/p/bridgewater-says-google-surpassed-openai)*
**What happens when AI agents trade against each other in financial markets?**
When AI agents trade against humans, they benefit from human predictability. When agents trade against other agents, the dynamics shift significantly. According to Irene Aldridge, agents are designed to occasionally act randomly to avoid pattern detection — injecting noise that other agents must filter. Distinguishing random agent behavior from intentional signals becomes extremely difficult, creating systemic risk that regulators have not yet addressed.
*Read more: [From Trading Algos to Trading Agents](https://www.ai-street.co/p/from-trading-algos-to-trading-agents)*
**How much are companies spending on AI?**
Bridgewater's research argues that companies will "spend whatever it takes" to keep up with AI. The firm predicts a major corporate panic moment — when a large business outside the AI ecosystem realizes its entire model is threatened — will dramatically accelerate spending. AI infrastructure investment is expected to be a central driver of global economic growth over the next two years, according to Bridgewater.
*Read more: [Bridgewater Says Google Surpassed OpenAI](https://www.ai-street.co/p/bridgewater-says-google-surpassed-openai)*
---
## 7. Limitations & Risks
**Where does AI fail in finance?**
AI struggles with nuanced market judgment, novel situations outside its training data, and tasks requiring deep institutional context. Man Group's Ziang Fang notes that "a lot of nuanced research still requires deep market understanding, judgment, and intuition." Hallucinations remain a serious problem in regulated settings where outputs must be auditable.
**Why can't large language models (LLMs) replace financial analysts?**
LLMs lack institutional context, live market intuition, and accountability. They can accelerate research workflows but cannot yet replicate the judgment of an experienced analyst in fast-moving or ambiguous situations. Regulators also require explainable outputs with clear evidence chains — something raw LLMs struggle to provide without additional infrastructure like RAG or causal AI frameworks.
**What is the hallucination problem in financial AI?**
Hallucinations occur when AI models produce confident but incorrect outputs. In finance, this is particularly dangerous. As Jayeeta Putatunda of Fitch Group explained to AI Street: "If a generative model says, 'Company X decreased its carbon footprint by 20%,' but in reality, it was only 5%, you have a serious problem." Solutions include retrieval-augmented generation (RAG) and causal AI frameworks that anchor outputs in validated, traceable data.
*Read more: [Wall Street Once Banned the Internet. AI Is in a Similar Spot](https://www.ai-street.co/p/wall-street-once-banned-the-internet-ai-in-similar-spot)*
---
*AI Street covers AI in finance from the trading floor to the C-suite. Subscribe at [ai-street.co](https://ai-street.co).*
