Industry Guide: Investment Banking
How to Use Claude in Investment Banking: A Guide for Deal Teams and Analysts
Fewer all-nighters on formatting, more hours on the deal.
This guide shows bankers how to put Anthropic's Claude to work on the production load of an M&A or capital markets deal: pitch books, comparable company analysis, precedent transactions, CIMs, teasers, buyer lists, and model reviews. It covers what Anthropic has published for financial services, what named firms have said in Anthropic's own posts, and where a banker's judgment cannot be delegated.
Knowing how to use Claude in investment banking means handing it the heavy production work and keeping the checks: drafting pitch book pages and company profiles, building comps and precedent tables from data you supply, drafting CIM sections and teasers from diligence material, assembling buyer lists, and auditing merger and LBO models for broken formulas. Claude produces the first draft quickly. The analyst and associate still verify every number, and the senior bankers own the advice given to the client.
Published
Key Takeaways
- The strongest investment banking uses are pitch book pages, comps and precedent tables, CIM and teaser drafts, buyer lists, and model review.
- Anthropic has published Agent Skills for comparable company analysis, DCF models, and teasers, plus a pitch builder and model builder agent template.
- Every figure in a client deliverable still gets tied to a source: filings, a data provider, or the client's own numbers.
- Junior bankers who learn to review AI output well become faster, not redundant. The judgment layer is what the job trains.
- Client and deal information belongs only in a workspace your bank has approved for it.
Copy and Paste
Prompts You Can Use Today
Company profile for a pitch book
You are an analyst on a sell-side M&A team. Using only the filings and investor presentation excerpts pasted below, draft a one-page company profile for a pitch book: business overview in three sentences, segments with revenue share, key customers and geographies as disclosed, recent strategic moves, and three points relevant to a potential sale. Put the source document and page next to every figure. Do not use outside knowledge. Material: [paste].
Restricting Claude to the pasted material keeps stale or unsourced figures out of the book.
Comps table and commentary
Here is a table of trading data for a peer set: company, enterprise value, revenue, EBITDA, and growth for the last fiscal year and the next two forecast years. Calculate EV to revenue and EV to EBITDA for each year, the mean and median for the set, and flag any company whose multiple is more than one standard deviation from the median. Then write four sentences of commentary a managing director could read aloud. Show your formulas. Data: [paste].
Asking for formulas means you can rebuild the numbers in Excel and confirm them.
CIM section draft
Draft the Industry Overview section of a confidential information memorandum for the company described below. Use only the market data and management materials I provide. Structure it as: market definition, size and growth drivers, competitive landscape, and why the company is positioned to benefit. Write in the neutral, factual tone of a sell-side CIM. Mark any claim that needs a source with [SOURCE NEEDED]. Material: [paste].
The [SOURCE NEEDED] marker keeps the draft honest before it goes to the client for review.
Buyer list with rationale
Based on the target company profile below, draft a buyer universe organized as strategic buyers and financial sponsors. For each potential buyer, give a one-line rationale tied to the target's products, customers, or geography, and note any obvious reason they might not engage. Mark every buyer you are not confident about as unverified so the team can check it against our database. Target profile: [paste].
Claude's knowledge has a cutoff, so the buyer list is a starting point to check against your firm's database.
Merger model review
Review the attached accretion and dilution model as an associate checking an analyst's work before the MD sees it. Trace the purchase price, sources and uses, financing assumptions, synergies, and pro forma EPS. List every hardcoded input inside a formula, every inconsistency with the transaction assumptions tab, and any sign error or broken link. Do not fix anything. Return a numbered issue list with cell references.
Keeping review and fixing separate means you decide what changes and learn where errors tend to hide.
From Anthropic
What Anthropic offers for Investment Banking
Claude for Financial Services
Anthropic's financial services offering, announced in July 2025, combines Claude models, Claude Code, and enterprise access with pre-built connectors to data providers such as FactSet, S&P Global, PitchBook, and Morningstar, plus implementation partners. Anthropic says data is not used for training by default.
· Source
Advancing Claude for Financial Services
An October 2025 update adding Agent Skills for comparable company analysis, discounted cash flow models, due diligence data packs, company teasers and profiles for pitch books and buyer lists, earnings analyses, and initiating coverage reports, alongside a beta of Claude for Excel and connectors including LSEG and Moody's.
· Source
Agents for financial services
Ten agent templates released in May 2026, including a pitch builder, meeting preparer, model builder, and valuation reviewer, shipped as plugins in Claude Cowork and Claude Code, with Anthropic's Excel, PowerPoint, and Word add-ins made generally available.
· Source
In Practice
How Investment Banking Teams Use Claude
According to Anthropic's published customer story
RBC Capital Markets
In Anthropic's October 2025 financial services update, RBC Capital Markets' head of AI and digital innovation said Claude stands out at integrating multiple data sources and automating workflows.
According to Anthropic's published customer story
Mizuho
Anthropic's May 2026 announcement of its financial services agents quotes Mizuho's banking COO describing preparation time being turned into time for ideas.
According to Anthropic's published customer story
Crunched
According to Anthropic's customer story, Crunched builds Excel tooling on Claude aimed at investment banking, private equity, and consulting users, and its customers report large time savings on modeling, financial spreads, and company write-ups.
Illustrative walkthrough
A pitch book company profile, sourced and checked
- The analyst pastes filing excerpts and the latest investor presentation into the deal project.
- Claude drafts the one-page profile with the source document and page next to every figure.
- The analyst ties each figure to the source and corrects a segment figure taken from the wrong period.
- The associate edits the strategic points to match the story the managing director wants to tell.
- The page goes into the book only after the standard tie-out and review.
Illustrative walkthrough
Reviewing a merger model before the MD meeting
- The associate asks Claude for an issue list on the analyst's accretion and dilution model, with no edits.
- Claude lists hardcoded inputs, a financing assumption that differs from the assumptions tab, and a broken link.
- The analyst fixes each issue and writes a short note on what changed.
- Claude runs a second review to confirm the fixes and flag anything new.
- The associate walks the managing director through the final outputs and owns the numbers.
Investment Banking and Claude: the Published Numbers
| Figure | Context | Source |
|---|---|---|
| financial spreads 8+ hours to 1 hour | Reported by customers of Crunched, an Excel tool built on Claude, per Anthropic's customer story | Source |
| company write-ups 8 hours to 20 minutes | Reported by customers of Crunched, per Anthropic's customer story | Source |
| >50% time savings on Excel modeling | Reported by customers of Crunched, per Anthropic's customer story | Source |
Core Workflows
Claude Investment Banking Use Cases Across a Live Deal
These are the jobs that consume most analyst and associate hours, and where AI for investment banking earns its place.
Pitch books
Company profiles, situation overviews, and market pages drafted from filings and data you supply. An AI pitch book generator is only as good as the sources you feed it.
Comps and precedents
AI for comps analysis means calculating multiples from your data, flagging outliers, and drafting commentary, with formulas you can rebuild and check.
CIMs and teasers
AI for CIM drafting turns diligence material into structured sections and anonymous teasers, with gaps marked for the deal team to fill.
Buyer lists
Strategic and financial buyer universes with a rationale for each name, ready to be checked against your firm's relationship data.
Model review
An AI merger model check traces purchase price, financing, synergies, and pro forma EPS, and lists hardcodes and broken links for the associate to fix.
Meeting and process prep
Client meeting briefs, process letters, management presentation outlines, and responses to buyer questions drafted from the deal record.
The Production Layer
How Investment Banks Use AI on the Work Juniors Know Best
Most investment banking hours go into production: turning data and filings into pages, checking that every number agrees across a hundred slides, and rebuilding tables when a comp set changes. That is exactly the work Claude handles well, as long as the inputs are clean and the output is checked. The pattern that works is narrow prompts with a fixed output format: one profile, one table, one section at a time.
Anthropic has leaned into this work directly. Its October 2025 financial services update described Agent Skills for comparable company analysis with valuation multiples and operating metrics, discounted cash flow models with WACC calculations and sensitivity tables, and company teasers and profiles for pitch books and buyer lists. In May 2026 it added agent templates including a pitch builder, model builder, meeting preparer, and valuation reviewer, available as plugins in Claude Cowork and Claude Code.
The discipline is sourcing. A bank's name is on every page that leaves the building. Ask Claude to cite the document and page for each figure, restrict it to material you provide, and keep a person responsible for tying out the book before it goes to the client. AI does not change the standard; it changes how quickly you reach it.
Careers
AI for Junior Bankers: Will AI Replace Investment Banking Analysts?
The question of whether AI will replace investment banking analysts comes up in every analyst class now, and the honest answer is that the job is changing shape rather than disappearing. The formatting and first-draft work that used to fill late nights is increasingly done by tools. What remains is the part of the job that was always the real training: understanding the business, knowing which number is wrong, and explaining a valuation to a skeptical client.
AI for junior bankers is most valuable as a reviewer and explainer. Ask Claude why a sensitivity table moves the way it does, what drives the difference between two multiples, or what a buyer will push back on in a CIM section. Those answers build judgment faster than formatting ever did, provided you check them against what your seniors say.
Analysts who learn to direct and verify AI output become the people teams rely on for speed and accuracy at once. Analysts who paste output into a book without checking it will be caught quickly, because the review culture in banking was built to catch exactly that.
Choosing Tools
Claude vs ChatGPT for Investment Banking, and Claude Skills for Investment Banking
The Claude vs ChatGPT for investment banking comparison should be run on your own deliverables. Take a company profile, a comps table, and a model you have already checked, and give each tool the same prompt and inputs. Compare accuracy, citation behavior, formatting effort, and how each responds when data is missing. Both tools update frequently, so rerun the test when your workflow changes.
Claude skills for investment banking refers to the Agent Skills and plugins Anthropic has published for financial work, covering comps, DCF models, teasers, and pitch materials, together with connectors to data providers such as FactSet, S&P Global, PitchBook, and LSEG. Where your bank has enabled them, they reduce setup work. Where it has not, the prompts on this page work in a standard Claude chat with documents you are allowed to upload.
The deciding factor in most banks is not model quality. It is what compliance has approved, what data terms apply to client information, and which data providers connect to the tool. A tool your bank has approved for live mandates beats a tool you can only use on public filings.
Getting Started
Four Habits That Make AI Output Bankable
Source every number
Ask for the document and page next to each figure, and restrict Claude to the material you pasted. Unsourced numbers do not go in the book.
One page at a time
Prompt for one profile, table, or section with a fixed format. Narrow prompts produce output you can check quickly.
Review, then fix
When auditing a model, ask for an issue list without changes. You decide what gets fixed and learn where errors hide.
Keep a prompt library
Save the prompts that match your bank's formats. Shared prompts make a whole deal team faster and more consistent.
Limits
Where Bankers Should Not Rely on AI
Claude can produce a fluent page with a wrong number in it. It can misread a scanned table, confuse fiscal and calendar years, or carry an outdated figure from its training data. None of that is acceptable in a client deliverable or a fairness opinion context. Tie out every figure against the source, and treat Claude's output as a draft that has not been checked.
Advice is a human responsibility. Recommending a process, a buyer, or a price involves relationships, market reads, and regulatory duties that stay with the bankers on the deal. Claude can help you prepare for those conversations but cannot have them for you.
Confidentiality is absolute. Mandates, client financials, and buyer interest are material non-public information. Use only the workspace your bank has approved, follow your information barrier rules, and never paste deal material into a personal account.
Build the Skill Before the Next Staffing Email
Prompting well under deal pressure is a trained skill, not a trick. The free Nightschool AI curriculum is hands-on from the first lesson, so the habits are there when the book is due.
Frequently Asked Questions
Will AI replace investment banking analysts?
AI is replacing parts of the analyst workload, especially formatting, first drafts, and table building. It is not replacing the need for people who understand the business, catch the wrong number, and support senior bankers with clients. Analysts who learn to direct and verify AI output tend to become more valuable to their teams.
Is investment banking still a good career with AI?
The core of banking is advice, relationships, and execution judgment, and those remain human work. What is shrinking is the manual production that used to dominate analyst years. People entering banking now should expect to spend more time on analysis and review and less on formatting, which many will find a better apprenticeship.
Can AI build a pitch book?
AI can draft much of a pitch book: company profiles, market pages, comps tables, and situation overviews, and Anthropic has published a pitch builder agent template. It cannot decide the story the MD wants to tell or guarantee every figure is correct. Treat the output as a first draft that the team sources, checks, and shapes.
Does Claude or ChatGPT do better financial modeling?
It depends on the model, the inputs, and which version of each tool you use. Test both on a model you have already checked and compare accuracy, formula transparency, and how each handles missing data. For client work, the tool your bank has approved and connected to its data providers is usually the practical answer.
How do investment banks use AI?
Banks use AI for pitch materials, comps and precedent analysis, CIM and teaser drafts, buyer lists, meeting prep, and model review, as well as internal research. Anthropic's own posts quote RBC Capital Markets on Claude integrating data sources and automating workflows, and Mizuho on preparation time turning into time for ideas.
Is AI quietly replacing entry-level analysts in investment banking?
AI is changing what entry-level analysts do more than whether they exist. Production tasks are moving to tools, and review, analysis, and client-facing preparation make up more of the job. The analysts at most risk are those who never learn to check AI output critically.
How do I check a model that AI helped build?
Treat it like an analyst's first draft. Trace sources and uses, the financing, and the output line by line, look for hardcoded numbers inside formulas, check that the balance sheet balances, and compare key assumptions to the source documents. Asking Claude for an issue list on its own model is useful, but a person still signs off.
Is it safe to use AI on live mandates?
Only inside a workspace your bank has approved for client information, under your information barrier rules. Mandates and client financials are material non-public information. Anthropic states that its financial services offering does not use data for training by default, but your bank's own policy decides what is permitted.
Keep going
Related Guides and Courses
Claude in Investment Banking, by role
Be the Analyst Who Is Fast and Right
Hands-on lessons you can apply to your next book.