AI Finance Brief › Finance Prompts › ChatGPT Prompts for Investment Banking
ChatGPT Prompts for Investment Banking
IB coverage / M&A / capital markets analysts & associates
Investment banking analysts and associates build the same artifacts on repeat under deadline — buyer universes, comps, precedents, CIM outlines, accretion/dilution checks. A large language model won't source the numbers for you, but with the right structure it will turn your pasted data into a clean first draft in minutes. Below are copy-paste prompts built around actual coverage, M&A, and capital-markets workflows. Paste them into ChatGPT, Claude, Copilot, Gemini, or Perplexity — then check every figure.
3 free prompts you can run right now
Strategic-buyer longlist + diligence question tree
Early-stage sell-side or a pitch: you need a defensible universe of logical acquirers and the diligence questions each raises.
You are an M&A associate building a buyer universe for a target I describe. Produce two things: A) A strategic + financial buyer LONGLIST. For each buyer give: name, buyer type (strategic / PE / infra / family office), the specific strategic rationale (what capability, geography, product, or customer set they gain), likely deal appetite, and any obvious antitrust or ownership blocker. Group by 'most logical', 'plausible', and 'wildcard'. Do not pad the list — quality over count. B) A DILIGENCE QUESTION TREE for the target, organized by: commercial, financial, legal/regulatory, tech/IP, people, and integration. Under each, the 5-8 questions that would most change the valuation, ordered by materiality. Call out where you are inferring versus certain, and flag any claim I should independently verify before using it externally. Here is the target: [DESCRIBE TARGET: sector, size, geography, what it does, why it might sell] Rules: Do not invent, estimate, or extrapolate any figure — if a number is not in what I give you, write "not provided" and flag it. Mark every claim I should verify externally before relying on it. Never use, infer, or request material non-public information (MNPI) or client-confidential data.
Precedent-transactions screen + adjusted-multiples framing
You need a defensible set of precedent M&A deals for a target and a clean read on what buyers actually paid.
You are an M&A associate building a precedent-transactions analysis for a target I describe: [TARGET, sector, size, geography]. Produce: 1. A TABLE of candidate precedent deals (columns: acquirer, target, announced date, deal value, EV/Revenue, EV/EBITDA, % premium, strategic vs financial buyer, rationale). Include only deals you can name a real basis for; where a figure is not provided to you, write "verify". 2. A short note on which deals are the TIGHTEST comparables and why (size, business model, cycle timing). 3. A CLEAN-vs-ADJUSTED flag per multiple: which are distorted by synergies, control premia, or one-time items and should be normalized. 4. The 2-3 questions to resolve before quoting any multiple to a client. Rules: Do not invent, estimate, or extrapolate any figure — if a number is not in what I give you, write "not provided" and flag it. Mark every claim I should verify externally before relying on it. Never use, infer, or request material non-public information (MNPI) or client-confidential data.
Comparable-companies universe + clean-vs-adjusted multiples grid
Building trading comps and you want a peer set you can defend plus multiples flagged clean vs adjusted.
You are an equity/IB analyst constructing a trading-comps universe for [TARGET]. Do the following in order: 1. Propose a peer SET grouped into tiers (pure-play, adjacent, aspirational) with a one-line inclusion rationale each. 2. Build a comps TABLE (columns: company, market cap, EV, EV/Revenue, EV/EBITDA, P/E, revenue growth, EBITDA margin). Where I have not given you a figure, write "pull" — never fabricate. 3. Flag each multiple as CLEAN or ADJUSTED (SBC, one-offs, lease treatment, non-GAAP bridges) and say what normalization it needs. 4. Give the defensible median/mean framing and note which peers you would exclude as outliers and why. Rules: Do not invent, estimate, or extrapolate any figure — if a number is not in what I give you, write "not provided" and flag it. Mark every claim I should verify externally before relying on it. Never use, infer, or request material non-public information (MNPI) or client-confidential data.
Get today's prompt free — every weekday
One fresh, desk-ready prompt for banking, trading, or the buy side lands in your inbox each morning. No news, no fluff. Free.
Instant access + the free brief. Unsubscribe anytime.
Want prompts matched to your exact role and coverage? Set your prompt profile →
7 more Banking prompts in Pro
The full Banking set — plus every other desk's prompts, personalized to your role — is part of Pro. Here's what's inside:
M&A sell-side process timeline + workstream tracker
Kicking off or running a sell-side process and need the phase map plus who-owns-what so nothing slips.
You are a deal-team associate. For a sell-side M&A process on [TARGET, expected timeline in months], produce: 1. A phased TIMELINE (Preparation, Marke
CIM outline + one-page teaser draft
You have a data pack on a target and need a structured CIM skeleton plus a blind teaser you can send.
You are an IB associate drafting marketing materials for [TARGET]. I will paste the key facts/data pack. Produce: 1. A CIM OUTLINE with numbered secti
Accretion / dilution first-pass sanity check
You want a quick, structured read on whether a deal is accretive before building the full model.
You are an M&A analyst running a FIRST-PASS accretion/dilution check. I will give acquirer and target [EPS, net income, shares, offer price, % cash vs
Football-field valuation range synthesis
You have several valuation methods and need to synthesize them into a defensible range with a clear story.
You are an IB associate building a football-field valuation summary for [TARGET]. I will paste the ranges from each method (DCF, trading comps, preced
LBO feasibility + debt-capacity quick screen
You want to know fast whether a target could support an LBO and roughly what return a sponsor might see.
You are a leveraged-finance analyst running a QUICK LBO feasibility screen on [TARGET: EBITDA, growth, margin, capex, expected entry multiple, prevail
Management-presentation question bank
You have a management meeting or diligence session coming and need a sharp, non-generic question set.
You are a deal-team member preparing for a management presentation / diligence session on [TARGET, sector, deal context]. Produce a QUESTION BANK orga
Data-room index red-flag scan + diligence priority list
You have a data-room index (or document list) and need to triage where the diligence risk concentrates.
You are a diligence lead triaging a data room for [TARGET]. I will paste the index/document list. Produce: 1. A PRIORITY TABLE grouping documents by r
Frequently asked
Can ChatGPT build a comps table?
It can structure and format one, but it cannot reliably pull live multiples on its own — treat any figure it retrieves as unverified. The dependable workflow is to paste your own peer set and raw figures (from your data provider or filings) and have the model rank the set, flag cheap/expensive names on a growth-adjusted basis, and explain the spread. You still verify every multiple against the source. Educational only, not financial advice.
Is it safe to use AI for M&A and banking work?
Only with your firm's data-handling and AI-use policy in front of you. Never paste material non-public information (MNPI), client-confidential deal data, or PII into a consumer AI tool. The prompts here are designed to work on public filings and on data you are cleared to use, and they instruct the model to say 'not disclosed' rather than invent a number. You own the output and its accuracy.
Which AI model is best for investment banking prompts?
There is no single winner — the prompts here are model-agnostic and work in ChatGPT, Claude, Copilot, Gemini, or Perplexity. Models with a larger context window handle a full CIM or 10-K in one paste more comfortably. What matters far more than the model is the prompt structure and the discipline of verifying every output against the primary source.