Industry Guide: Asset Management
How to Use Claude in Asset Management: A Guide for Buy-Side Investment Teams
Read more of the market, and decide with better notes.
This guide is for portfolio managers, analysts, and investment operations teams at asset managers and hedge funds who want to put Anthropic's Claude to work on research, earnings season, IC memos, portfolio reviews, and quantitative workflows. It covers what Anthropic has released for financial services and what named firms have said in Anthropic's own posts.
Learning how to use Claude in asset management means pointing it at the reading and writing that crowd out analysis: summarizing earnings calls and filings into model updates, drafting initiation and thesis notes, building IC memos and portfolio reviews from holdings data, writing Python for data analysis, and preparing client and compliance documents. Claude produces structured first drafts quickly. Portfolio managers keep the investment decision, and every figure that reaches a client, committee, or regulator is checked against the source.
Published
Key Takeaways
- The main buy-side uses are earnings reviews, research notes, IC memos, portfolio reviews, and analyst tooling in Python and Excel.
- Anthropic has published Agent Skills for earnings analyses and initiating coverage, and agent templates including an earnings reviewer and valuation reviewer.
- Anthropic's posts quote Citadel, Bridgewater, and Walleye Capital on how their teams use Claude.
- Claude speeds up research, but it does not generate alpha on its own. Your process and judgment do.
- Holdings and client data belong only in a workspace your firm has approved for it.
Copy and Paste
Prompts You Can Use Today
Earnings call to model update
You are a buy-side analyst covering the company below. Using the earnings call transcript and press release pasted here, list: every change to guidance with the old and new figures, management's explanation for each, the three most important analyst questions and how management answered them, and any line in my model summary that now looks wrong. Quote the transcript for each point. Do not use outside knowledge. Material: [paste transcript, release, and model summary].
Pasting your own model summary is what turns a generic call summary into a model update list.
Thesis tracker update
Here is our written investment thesis for a current holding, with the three pillars and the conditions that would make us sell. Here is the latest quarter's news, filings summary, and our notes. For each pillar, say whether the new evidence strengthens it, weakens it, or is neutral, and why. List anything that touches a sell condition. Keep it under three hundred words and flag any point where you are inferring rather than reading from the material. Material: [paste].
A written thesis with explicit sell conditions is what makes this prompt useful. Write one if you do not have it.
IC memo draft from research
Draft an investment committee memo recommending a new position, using only the research notes, model outputs, and valuation summary below. Sections: recommendation and sizing rationale, business summary, thesis, valuation, key risks and what would prove us wrong, catalysts, and portfolio fit. Write for a committee that reads quickly and asks hard questions. Do not add figures that are not in my material; mark gaps as [NEEDS DATA]. Material: [paste].
Ask for the risks section first if your committee tends to focus there.
Portfolio review commentary
Using the holdings, weights, benchmark weights, and period returns pasted below, write a quarterly portfolio review for a client committee: performance versus benchmark, the main contributors and detractors with a one-line reason each, changes made during the quarter, and positioning going forward. Keep the language plain and factual, avoid forecasts, and flag any number you calculated rather than read from the data. Data: [paste].
Check the attribution against your performance system before anything reaches a client.
Python for a research question
Write a Python script using pandas that loads the CSV described below, calculates rolling ninety-day volatility and the correlation of each holding to the benchmark, and produces a table of holdings sorted by contribution to portfolio volatility. Add comments explaining each step, handle missing prices explicitly, and include a short test on a small fake dataset so I can confirm the logic before running it on real data. CSV columns: [describe].
Asking for a test on fake data first lets you check the logic without exposing real holdings.
From Anthropic
What Anthropic offers for Asset Management
Claude for Financial Services
Anthropic's financial services offering, announced in July 2025, combines Claude models, Claude Code, and enterprise access with connectors to data providers such as FactSet, Morningstar, S&P Global, Snowflake, and Databricks. Anthropic says data is not used for training by default.
· Source
Advancing Claude for Financial Services
An October 2025 update adding Agent Skills for earnings analyses, initiating coverage reports, comparable company analysis, and DCF models, plus connectors including Aiera for earnings call transcripts, LSEG for market data, Moody's for credit data, and MT Newswires for news.
· Source
Agents for financial services
Ten agent templates released in May 2026, including an earnings reviewer, market researcher, valuation reviewer, and statement auditor, shipped as plugins in Claude Cowork and Claude Code, with Excel, PowerPoint, and Word add-ins generally available.
· Source
In Practice
How Asset Management Teams Use Claude
According to Anthropic's published customer story
Citadel
According to Anthropic's May 2026 financial services post, Citadel's head of core engineering said analysts use Claude in Excel to build and update coverage models and pressure-test their work.
According to Anthropic's published customer story
Bridgewater Associates
In Anthropic's July 2025 financial services announcement, Bridgewater's AIA Labs said Claude powered the first versions of its Investment Analyst Assistant, which generated Python code, created data visualizations, and worked through complex financial analysis.
According to Anthropic's published customer story
NBIM (Norges Bank Investment Management)
Anthropic's customer story reports that NBIM, which manages Norway's sovereign wealth fund, saw weekly time savings per employee on Claude-assisted analytical and operational tasks and broad adoption across departments within two months.
According to Anthropic's published customer story
Walleye Capital
In Anthropic's May 2026 financial services post, Walleye Capital's chief executive and chief investment officer said all of the firm's employees use Claude Code.
Illustrative walkthrough
Earnings night for a covering analyst
- The analyst pastes the transcript, the press release, and a short summary of the current model into the coverage project.
- Claude lists guidance changes, management explanations, and the model lines that now look wrong, quoting the transcript for each.
- The analyst checks each quote, updates the model, and rejects one suggested change that misread a segment definition.
- Claude drafts a short note to the portfolio manager from the analyst's updated view.
- The portfolio manager reads the note and decides whether the position changes.
Asset Management and Claude: the Published Numbers
| Figure | Context | Source |
|---|---|---|
| 20% time saved weekly per employee | NBIM, on Claude-assisted analytical and operational tasks, per Anthropic's customer story | Source |
| 600+ active users across all departments within the first two months | NBIM's rollout, per Anthropic's customer story | Source |
| 100% of Walleye employees use Claude Code | Stated by Walleye Capital's CEO and CIO in Anthropic's May 2026 financial services post | Source |
Core Workflows
Claude Asset Management Use Cases on the Buy Side
These are the jobs where AI for asset management saves the most analyst time.
AI for earnings analysis
Transcripts, releases, and filings summarized into guidance changes, management commentary, and a list of model lines to revisit.
AI for buy-side research
Sector overviews, competitor maps, initiation drafts, and thesis updates built from filings, transcripts, and your own notes.
IC memos
An AI investment memo generator is only useful with your research as input. Claude drafts the committee memo in your format and marks gaps.
AI for portfolio management
Portfolio reviews, attribution commentary, exposure summaries, and rebalance rationales drafted from holdings data you supply.
Quant and data work
Python and SQL for screens, risk calculations, and data cleaning, written with comments and tests so an analyst can check the logic.
Operations and compliance
Investment policy checks, LP statement reviews, and client reporting drafts, with every figure traced back to the system of record.
Research
How Asset Managers Use AI for Research and Earnings Season
Ask how asset managers use AI and the most common answer is research throughput. An analyst covering a few dozen names cannot read every transcript, filing, and competitor release closely in earnings season. Claude reads them in full and produces structured summaries: guidance changes, management tone, the questions analysts pushed on, and what it means for your model. Claude AI investment research works best when you give it your own model summary and thesis, so the output is about your position rather than a generic recap.
Anthropic has built directly for this work. Its October 2025 financial services update added Agent Skills for earnings analyses and initiating coverage reports, plus connectors including Aiera for earnings call transcripts, LSEG for market data, Moody's for credit data, and MT Newswires for news. In May 2026 it released agent templates including an earnings reviewer, market researcher, and valuation reviewer.
Generative AI in asset management is only as reliable as the checks around it. The pattern that holds up is to ask for quotes and page references, restrict Claude to supplied material when accuracy matters, and have the analyst confirm every figure that changes a model or a recommendation.
Hedge Funds
AI for Hedge Funds: What the Named Examples Show
AI for hedge funds attracts the loudest claims, so it helps to stick to what firms have actually said on the record. In Anthropic's May 2026 post, Citadel's head of core engineering described analysts using Claude in Excel to build and update coverage models, separate signal from noise, and pressure-test their work. Walleye Capital's chief executive said every employee at the firm uses Claude Code.
Earlier, Bridgewater's AIA Labs said Claude powered the first versions of its Investment Analyst Assistant, which generated Python code, created data visualizations, and worked through complex financial analysis tasks. The common thread is not that AI makes investment decisions. It is that analysts get through more analysis, and engineers and quants ship tools faster.
None of these examples claim that AI improved returns, and no honest guide can promise that. Treat Claude as leverage on your research process. If the process has an edge, AI helps you apply it to more names. If it does not, AI will not supply one.
Choosing Tools
Claude vs ChatGPT for Asset Management
The fair way to settle Claude vs ChatGPT for asset management is to test both on your own work. Take a transcript you have already summarized, a memo you have already written, and a Python task you have already solved. Run the same prompts in each tool and compare accuracy, citation behavior, code quality, and how each handles missing data. Repeat when either tool changes.
For most firms the practical question is integration and governance: which tool connects to the data providers you already pay for, which data terms apply to holdings and client information, and what compliance has approved. Anthropic lists connectors to FactSet, S&P Global, Morningstar, LSEG, and others, and states that its financial services offering does not use data for training by default.
Many teams end up using more than one tool for different jobs. What matters is that each use is approved, that output is checked, and that the team shares the prompts and review habits that work.
Getting Started
Setting Up an Investment Team for AI Research
Write theses down
Claude can only test a thesis it can read. Written pillars and sell conditions make every thesis update prompt sharper.
Start with earnings season
Use the earnings prompt on a quarter you already covered. Compare the output with your notes before trusting it on live names.
Test code on fake data
Have Claude write tests and run them on a small fake dataset first. Only then point the script at real positions.
Share what works
Keep a team prompt library for recurring jobs so analysts produce consistent notes and reviews.
Limits
What AI Cannot Do for an Investment Team
Claude can misread a table, attach a figure to the wrong quarter, or state an outdated fact from its training data with confidence. In research that feeds a trade, and in reporting that reaches clients or regulators, every figure must be confirmed against the source or the system of record.
Claude does not see live prices or your order management system unless your firm has connected them. Anything it says about current markets without a connected data source is at best out of date. Keep it on analysis of material you provide.
Holdings, client information, and unpublished research are sensitive. Use only the workspace your firm has approved, follow your information barrier and recordkeeping rules, and keep personal accounts out of investment work entirely.
Turn Research Habits Into AI Habits
Knowing these prompts and using them well during earnings season are different things. The free Nightschool AI curriculum is hands-on from the first lesson, so the habits are in place when volume peaks.
Frequently Asked Questions
How do hedge funds use AI?
Hedge funds use AI for research throughput, coverage model updates, code for quantitative work, and internal tools. In Anthropic's posts, Citadel describes analysts using Claude in Excel to build and update coverage models and pressure-test work, and Walleye Capital says all of its employees use Claude Code. Investment decisions remain with portfolio managers.
How do asset managers use AI?
Asset managers use AI to summarize earnings and filings, draft research notes and IC memos, write portfolio reviews and client reporting, and build analysis tools in Python and Excel. Anthropic's customer story on NBIM, Norway's sovereign wealth fund manager, reports broad use across departments and weekly time savings per employee.
What AI tools do hedge funds use?
Most combine a general assistant such as Claude with coding tools, market data providers, and internal systems. Anthropic lists connectors for FactSet, S&P Global, LSEG, Moody's, Aiera, and others, and agent templates for earnings review and valuation review. Which tools a fund uses depends on its data terms and compliance approvals.
Can AI generate an investment memo?
Claude can draft an investment memo in your committee's format from your research notes and model outputs, and mark where evidence is missing. It should not generate the thesis itself. A memo drafted from your material and edited by the analyst is useful. A memo generated from nothing is a liability.
Will AI replace hedge fund analysts?
AI is taking over much of the reading and first-draft work analysts used to do by hand, so analysts can cover more names and spend more time on judgment. The job of forming a differentiated view and defending it to a portfolio manager remains human work.
How much does AI improve hedge fund returns?
There is no reliable public evidence we can cite that AI assistants improve returns on their own, and you should be skeptical of claims that they do. The documented benefits in Anthropic's customer stories are time savings and faster tooling. Whether that becomes better performance depends on the fund's process.
How should an investment team check AI research?
Ask for quotes and page references, restrict Claude to supplied material when accuracy matters, and confirm every figure that changes a model or recommendation against the source. For code, require tests and run them on fake data before using real positions.
Is it safe to use AI with holdings data?
Only inside a workspace your firm has approved, with data terms your compliance team has reviewed. Anthropic states that its financial services offering does not use data for training by default. Follow your firm's recordkeeping and information barrier rules regardless of the tool.
Keep going
Related Guides and Courses
Claude in Asset Management, by role
Cover More Names Without Cutting Corners
Hands-on lessons built around real research work.