The Complete Guide to Claude for Private Equity

1. How to max out your Claude setup

The teams getting the most from Claude have done three things. They connected all of their data sources, organized their data by project, and built skills for all the repeatable work. None of this needs a developer.

How teams use Claude today

We spoke with 20+ PE funds and the advisors around them. These are the patterns that came up most.

• Claude is the default tool. Most firms we spoke with run on a Team or Enterprise plan.
• Usage is heaviest from NDA to LOI. NDA redlines, CIM screens, first-pass financial models and Excel exhibits are the most common jobs.
• Everyone does their own thing. A healthcare fund’s head of origination put it plainly: “Everybody is just wildcatting and figuring out how to do their work in Claude.” Most teams lack process that is consistent across the investment team.
• Most teams have everything connected to Claude. A large-cap digital team we spoke with connects its CRM, IC deck archive and expert call libraries to Claude. Their workflow: “You put all the context in a folder, point Claude at it, and we use the skills we’ve built.”
• Most teams don’t know what they’re missing. “I know what I can use Claude for today, but what else can I be using it for?”

The building blocks

A full Claude setup has five parts. Most teams only use the first.

Building block	What it does	Plans
Chats	One-off questions on whatever you paste or upload. Every chat starts from zero	All
Projects	Workspaces with shared instructions, files and history. Claude reads them before every chat	All. Sharing needs Team or Enterprise
Connectors	Let Claude pull your email, files, CRM, documents, and market data directly	All. On Team and Enterprise an owner turns them on
Skills	Written procedures Claude follows for repeatable work, like a CIM screen or NDA redline	All. Sharing needs Team or Enterprise
Claude in Excel	Claude inside the model, memo, deck and inbox, with shared context between them	Pro, Max, Team, Enterprise
Scheduled Tasks	Automatically run a task on a set cadence. This lets you do work without having to ask	All

For a deal team of more than two people, use a Team or Enterprise plan. Sharing projects and skills is what turns one person’s setup into the firm’s shared workflow.

Context is key

Claude’s output is only as good as the context and data it has. The same CIM screen gets a far better answer when Claude knows your buy box, your portfolio, and the last three deals you passed on and why. Most of the gap between a generic answer and a useful one comes down to context.

That context lives in two places.

1. A centralized data layer. One home for deal data: CRM records, deal folders, data rooms and past CIMs, organized and permissioned. For most lower middle market firms that’s usually SharePoint or Box, connected to Claude. Larger teams tend to use PE-specific data storage tools. Either works if three things are true:

• The data is organized. Most often, we set up every deal and portfolio company in its own folder with clear naming conventions.
• Sensitive data is out of reach. This is the hard part to execute with Claude. You can set connector permissions to stop Claude from accessing sensitive HR, LP and fundraising material.
• Past CIMs, screens and IC memos are kept. One fund told us it has every CIM it has ever seen and has never used them. That’s completely untapped information that can be used in the future to help improve your Claude skills based on past decisions.

2. Shared projects. A project holds the context for one piece of work: its instructions, files and chat history. On Team and Enterprise plans you can share a project with specific teammates or the whole firm, at two levels. Can view lets people chat in the project. Can edit lets them change its instructions, files and members.

We recommend three kinds of project:

Project	Shared with	What goes in it
Firm project	Whole investment team, can view	Buy box, NDA positions, 2-3 IC memos you’d hold up as the standard, one-pager and IC templates, DRL, a short style guide
Deal project	Deal team, can edit	CIM, LOI, model outputs, call notes. Instructions with the code name, stage, key dates and the thesis in three lines
Portfolio project	Board members and operating partners	Monthly packages, budget, value creation plan, board decks

Best practices for a deal project:

1. Create a new project and name it with the deal’s code name.
2. Write the instructions: stage, key dates, the thesis in three lines, and what the partner will ask about first.
3. Add the working documents: CIM, LOI, model outputs and call notes.
4. Share it with the deal team at Can edit. Share the firm project with the whole team at Can view, so one person maintains it.

A few rules make projects work:

• One owner per project. Edits get made once and everyone picks them up.
• Keep full data rooms out. Once a project’s knowledge nears the context limit, Claude switches to retrieval, starts to hallucinate, and becomes less accurate. Note: Anyone in the project can also query the documents in the project context. That’s a data layer problem, covered in Section 3.

Connect your tools

Connectors let Claude talk to your other tools directly, so nobody drags files into chat. This accelerates the rate at which Claude can get context and give you a better answer.

These are the most common connectors PE teams use:

Tool	What Claude can do with it	Setup notes
Microsoft 365: SharePoint, OneDrive, Outlook, Teams	Search deal folders, read email threads and calendar, find Teams messages	Owner enables it, then an Entra Global Administrator grants one-time consent. Read-only, and Claude only sees what each user can already see
Google Workspace: Drive, Gmail, Calendar	The same for firms on Google	Owner enables it. Each person connects their own account
Box	Search deal folders and internal data rooms	Scope access to deal and portfolio folders before you connect
PitchBook	Companies, deals, investors, funds, LP commitments and co-investor mapping	Needs a PitchBook license with MCP access. An owner adds it under Admin settings > Connectors
Financial and market data sources (ex: S&P Capital IQ, Daloopa, Morningstar, Moody’s)	Comps, financials, credit and market data, pulled straight into Claude or Excel	Uses your existing license
CRM (Hubspot, Salesforce, PE-specific CRMs)	Check whether the firm has seen a company, pull relationship history, log deals	Start read-only. Turn on write access once the team trusts it
Deal origination & contact data (ex: Apollo)	Turn a target list into an outreach list	Read the vendor’s data-sharing terms first. One fund’s IT team vetoed this connector over them

To set them up:

1. An owner turns on the connectors the firm will use in the organization’s admin settings.
2. For Microsoft 365, your Global Administrator grants consent once for the whole tenant.
3. Each person connects their own account from Claude’s connector settings.
4. Start with connectors having read-only access. You can turn on write actions, like drafting emails or updating the CRM, once the team is comfortable.

Microsoft 365 connector · PitchBook connector · Claude for Financial Services

Claude in Excel

Claude in Excel is the use case associates told us surprised them most. It runs as a sidebar inside the workbook and is generally available on Pro, Max, Team and Enterprise plans.

What it does:

• Builds models from scratch or fills your templates.
• Reads the whole workbook and answers with cell-level citations you can click.
• Changes assumptions while keeping formulas intact, so the model still flows.
• Pulls data in through other data connectors.
• Shares context with Claude in Word, PowerPoint and Outlook, so one conversation can run from the model to the memo to the IC deck.

Where PE teams use it:

• First-pass LBO. “Populate this LBO template with the banker’s projections and 3.5x leverage at close.”
• Model review. “Does the balance sheet balance in every period, and do historical EBITDA figures tie to the QoE?”
• Customer data. Cohorts, retention and concentration from the seller’s customer export.
• Trial balances. 36 monthly exports stacked into one model-ready tab.

To install it, find “Claude for Microsoft 365” on Microsoft AppSource, click Get it now, then open Excel, activate the add-in and sign in. To roll it out to the firm, an admin deploys it from the Microsoft 365 Admin Center under Settings > Integrated apps > Add-ins. Claude for Excel docs

Set up and use skills

A skill is a written procedure Claude follows for a repeatable task. It prevents you from having to write the same instruction every time you want to do a task. It also can create shared workflow, so every associate follows the same process with Claude.

How a skill works.

A skill is just a file that has a reusable prompt. There are two ways to use skills.

1. If you ask for a task, “Screen this CIM” and Claude sees you have a CIM screening skill, it’ll use it automatically.
2. You can explicitly use skills as well. In Claude, type a “/” and your skill name in order to instantiate it.

Skills and projects work together. The skill holds the instructions. The project holds the data and context.

To install a skill (any paid plan, with code execution and file creation turned on):

1. Go to Customize > Skills.
2. Click +, then Create skill, then Upload a skill.
3. Upload the ZIP. The skill appears in your list and can be toggled on or off.

How to share skills across the firm:

• Share with people: open the … menu next to the skill and choose Share. Recipients get view-only access and toggle it on.
• Publish to the firm: open the skill and click Publish to org. Choose whether it’s available to install, installed by default or required.
• Admin setup: an owner turns on skills and sharing under Organization settings > Plugins & skills > Policy.

How to write your own private equity Claude skills

Start with a repeatable process. Let’s use NDA redlining as an example. Write out a list of instructions for Claude on how to redline an NDA and give it context on the NDA criteria your firm will accept.

Next, ask Claude to “turn what I just wrote into a skill”. If you want a head start, our pack of Claude skills comes free with this guide.

The most common ways private equity is using Claude

These are the workflows we’ve seen working at real funds, each matched to the skill in our free pack that runs it.

Workflow	What we’ve seen	Skill
NDA redline	Banker’s NDA marked up against the firm’s positions, back out in about five minutes	nda-redliner
CIM screen	Every CIM run through a templated screen, with a different template for platforms and add-ons	deal-screen
Weekly deal triage	A batch of teasers ranked against the buy box so the team reads the best ones first	cim-triage
House one-pager	CIM plus model turned into the firm’s one-pager, then finished in PowerPoint	one-pager
First-pass LBO	“Assume 3.5x leverage at close” on the banker’s projections, with an IRR and MOIC grid	quick-lbo
DRL check	Each request marked fulfilled, partial or missing against what the seller uploaded	drl-tracker
Pre-QoE tie-out	P&L tied to tax returns and bank data, add-backs tested before the QoE firm starts	pre-qoe-scan
Morning brief	“What are the three things I have to focus on today, and who should I call?” from the inbox and calendar	daily-deal-brief
Sector thesis	Market size, fragmentation, PE ownership and deal activity for a new vertical	thesis-market-map
Target list	Trade show lists turned into an enriched target universe with owner profiles	target-universe
Contract review	Change of control and assignment terms pulled into one table, tied to revenue	contract-risk

The full team setup, step by step

This is the setup we’d put in place at any lower middle market fund. The Claude side takes an afternoon. The IT side, mostly scoping access, can take a few weeks, so start it first.

1. Move to Team or Enterprise. Sharing projects and skills needs one of them. Admins also control which connectors are on and whether Claude can write or only read.
2. Map what data Claude should see. Live deals, dead deals and portfolio folders in. HR, LP and fundraising out. Get folder permissions right before connecting anything, because connectors follow them.
3. Connect your core tools. Microsoft 365 or Google first, then your file store, CRM and market data like PitchBook.
4. Install Claude in Excel for everyone who touches a model or a memo.
5. Build the firm project. Buy box, NDA positions, two or three model IC memos, your templates, DRL and style guide. Share it with the whole team at Can view.
6. Create a project per live deal. Code name, instructions, working documents, shared with the deal team at Can edit.
7. Install the skill pack. Start with the skills for your highest-volume work: NDA redlines, CIM screens and one-pagers.
8. Write your own skills for anything the team does repeatedly. Publish them so anyone at the firm can use them.
9. Review after 30 days. Which skills get used and which manual step should become an agent. Section 3 talks about the next step after maxing out Claude: autonomous agents that do work without you having to ask.

2. The PE Claude skills suite

We’ve built 29 Claude skills that cover the deal cycle from origination to portfolio monitoring. Each one follows your firm’s materials when it finds them in the project, and falls back to sensible lower middle market defaults when it doesn’t. They install in one upload and work on any paid plan.

Origination

• thesis-market-map: Researches a sector for a thesis: market size, fragmentation, who owns the players and deal activity
• target-universe: Builds and enriches a target list with owner profiles and warm intro paths
• founder-outreach: Drafts founder and owner outreach, or intro requests, in the firm’s voice

Screening

• nda-redliner: Redlines a banker or seller NDA against your positions, with an issues list and cover note
• cim-triage: Ranks a batch of CIMs and teasers against the buy box so you know which to read first
• deal-screen: Screens one CIM against your criteria: verdict, bull and bear case, first-call questions
• cim-pressure-test: Tests banker claims, add-backs and projections against what the numbers support
• one-pager: Turns a CIM or model into your firm’s one-pager in its exact format
• quick-lbo: Builds a first-pass LBO and IRR and MOIC sensitivity in Excel on your standard terms

Data room

• data-room-organizer: Reorganizes a messy data room into your folder structure, mapped to the DRL
• data-room-index: Indexes the data room by workstream and flags duplicates and conflicting figures
• drl-tracker: Tracks the DRL against uploaded files and drafts the follow-ups
• data-gap-finder: Finds what the thesis depends on that the data room doesn’t support

Diligence

• thesis-interviewer: Interviews the deal team to write a testable thesis with kill criteria
• pre-qoe-scan: Ties the P&L to tax returns and bank data and tests add-backs before the QoE
• revenue-quality: Cohorts, retention, churn and customer concentration from customer data
• contract-risk: Extracts change of control, assignment and termination terms and ties revenue to them
• tech-dd-light: Light tech diligence: systems, cyber basics, data privacy, AI risk and upside
• diligence-meeting-prep: Preps questions for management, expert and banker calls, and processes the transcript after
• diligence-tracker: Tracks workstreams, owners and deadlines from LOI to close

Investment committee

• what-matters: Cuts long diligence findings to the 5-10 make-or-break issues
• finding-verifier: Checks each finding against source documents and marks it proven or unsupported
• ic-memo-drafter: Drafts the IC memo in your template from the one-pager, model and findings
• dd-writing-style: Rewrites findings and memos in house style, with cited facts and no AI filler

Portfolio

• value-creation-planner: Builds a value creation and 100-day plan with an EBITDA bridge
• portfolio-monitor: Reviews monthly packages against budget, checks covenants and drafts updates

Firm setup

• daily-deal-brief: A morning brief from your inbox and calendar: deals needing attention and who to call
• connector-access-setup: Plans how to connect Claude to Box, SharePoint or a CRM without exposing HR or LP data
• ai-use-case-map: Maps where AI can help across the firm or portfolio, with a 90-day plan

3. Guide to autonomous agents

Most firms that take Claude seriously hit a ceiling. They’ve connected their tools, built shared projects and installed skills, and Claude is fast. But it still waits to be asked.

The next step is autonomous agents built into the operations of the fund. These agents pick up work as it arrives, work together across the deal process, and do the first pass without being asked. The deal team reviews what they produce and makes the judgement calls.

What are autonomous agents?

An autonomous agent is AI with a job, the access to do it, and a trigger that starts it. A chat answers one question when you ask it. An agent runs a multi-step process on its own. For example, it can find the new CIM you got emailed, read the deck, check it against the firm’s investment thesis, and create the deal in the CRM with a summary of what it found.

Agents also work together. One picks up a CIM from the inbox and files it. A second screens it against the buy box. A third checks the portfolio for add-on fit and drafts the brief. Each agent does one job well, and together they run a workflow end to end.

The AI-native operating system for PE

Put enough agents in place and the fund runs on an AI-native operating system: the firm’s data, rules and workflows in one place, with agents doing the first pass on everything that comes in. Three pieces make it work.

• A centralized data store. The inbox, CRM, deal folders, data rooms and past deals in one permissioned place. An agent can only act on what it can reach.
• An operating system for the firm. The firm’s rules written down once: thesis, buy box, portfolio list, NDA positions, templates and skills. It’s what makes an agent’s output look like your team’s work.
• Triggers. An email lands, a file is uploaded, a date passes. The agent runs the right skill and writes the result back to where the team already works.

Every trigger runs through the same data layer and firm rules. The agent produces drafts, but humans always do the final review. Each decision goes back into the data layer, so the AI improves itself over time.

Examples of AI agents on an investment team

CIM analysis

A banker sends a CIM. Today an associate downloads it, reads it, screens it and logs it. With an agent:

1. The CIM lands in the deals inbox. The agent saves it to the deal folder and checks the CRM for whether the firm has seen the company before.
2. It runs the firm’s screen against the buy box, benchmarks the deal against past CIMs, and checks the portfolio for add-on fit.
3. It creates or updates the CRM record with the screen attached.
4. It sends you a brief.
5. You decide: take the first call, pass, or ask for more. If you pass, the reply to the banker is already drafted.

The brief looks like this (illustrative):

New deal: Project Ridgeline. Commercial HVAC services, Texas. $6M EBITDA.

80% fit against thesis. Possible add-on to [Portfolio Co], which covers the neighbouring Oklahoma market.

Highlights: 60% of revenue from recurring service contracts. Founder-owned, no prior sponsor. Margins above the portfolio average.

Lowlights: Top customer is 18% of revenue. Add-backs are 22% of adjusted EBITDA. Projections assume two new branches in year one.

Recommendation: Take the first call. Ask about the top customer’s contract term and the branch plan.

NDA redline

Once you flag interest, the banker sends an NDA. With an agent:

1. The agent picks up the NDA from the email thread.
2. It compares it to the firm’s NDA positions and template.
3. It produces a tracked-changes redline, an issues list, and a two-line note for the banker on anything you can’t accept.
4. It saves a draft reply in your inbox with the redline attached.
5. You review, adjust and send. One fund’s NDA skill already does the redline in about five minutes. The agent removes the download, upload and email steps around it.

Red flag review on every new document

Diligence doesn’t start at the full data room. Before the LOI, sellers send documents in batches: a few years of financials, a customer list, a management deck. Most teams review each batch by hand, if at all. One firm we worked with received about 20 new documents after the first review was done.

With an agent, every batch triggers a new pass:

1. New documents land in the deal folder or data room.
2. The agent indexes them and checks them against everything already received, including the CIM.
3. It flags what’s new: figures that don’t tie between documents, risks the CIM didn’t mention, and requests still outstanding.
4. It updates the running issues list and the DRL, and drafts follow-ups to the seller.
5. You get a short note on what changed since the last pass.

By the time the LOI is signed, you already have a running list of the issues that matter and the questions to ask. This is one of the core workflows in Fulcra. It runs a fixed set of checks across every workstream each time new documents arrive, and traces every finding back to the page it came from.

The CRM updates itself

Most firms’ CRMs are only as current as the last associate who updated them. With agents connected, the CRM keeps itself current:

• Every inbound deal and banker email gets logged to the right company and contact.
• Deal stages move when the email thread says they should. A signed LOI moves the deal to exclusivity.
• New companies are checked against everything the firm has seen before.
• Contacts are enriched with ownership, size and how to reach the owner.
• Follow-ups are scheduled with bankers and owners the team hasn’t spoken to in 90 days.

Some firms do this on top of their existing CRM. Others build a lightweight CRM around the agents, so the system is designed to be updated by AI from the start.

Portfolio monitoring and reporting

The same pattern works after close. When a portfolio company’s monthly package arrives, the agent pulls the KPIs, compares them to budget and plan, checks covenants and flags variances worth a call. Each quarter it rolls the portfolio up into a draft LP report and board pre-reads in your format, with the questions worth asking. The operating partner reviews, edits and sends.

More examples of autonomous agents worth automating

Trigger	What the agent does	What you decide
7am every day	Morning brief from the inbox, calendar and CRM	Who to call and what to answer first
A batch of teasers arrives	Ranks them against the buy box	Which ones to read
Management call ends	Updates the diligence tracker and logs open questions from the transcript	What to ask next
QoE report delivered	Verifies each finding against source documents and ranks what matters	The IC position on each issue
Every Friday	Drafts DRL follow-ups for anything outstanding	Whether to send or escalate

Want a hand taking the next step for your fund?

We help investment teams get more out of Claude. Usually that starts with the setup in this guide.

Some teams want to go further and build agents into how the fund runs. We build custom AI tools for that too, including the examples in this post:

• CIM review and screening
• AI-native CRMs
• Due diligence tooling
• QoE-lite and financial review
• NDA redlining and contract review

Interested in understanding the next step for your fund? We offer consultations to help you evaluate where your fund is at with AI and show you where the opportunities are. If you’re interested, get in touch, and we’d be happy to compare notes.

Want a head start on your Claude setup?