Find warm intros to investors with your LinkedIn export and one prompt

Why this works now

Every founder raising a round hears the same advice. Get a warm intro. It is good advice, and in practice it turns into an evening of clicking through LinkedIn one investor at a time.

This year a simpler version started going round on X. Abhilash Chowdhary laid it out: export your LinkedIn connections, give the file to Claude, and let it find the paths. Paul Klay shared a Claude skill that goes further, from picking the right partner at a fund to checking for conflicts and writing the outreach. Both got the same reply over and over: how do I do this.

this is how I would find investors if I had to start from 0

cold emails get ~2% replies, warm intros get 30-40%+, I would focus on finding warm intros by doing this:

1. export your LinkedIn connections (→ Settings → Data Privacy → Get a copy of your data → Connections →…

Abhilash Chowdhary (@TheChowdhary) May 17, 2026

if you’re raising, this skill will dramatically improve your investor conversion

it finds the right investors, potential conflicts, the right partner, warm intro paths, and writes the outreach

weeks of research in one run

Paul Klay (@PaulKlayVC) September 28, 2026

What follows is the version anyone can run today with a free LinkedIn download and ChatGPT or Claude. It is honest about what the file can and cannot tell you, because that is where most of these workflows quietly fall down.

Step 1: Download your connections

On LinkedIn, click Me, then Settings and Privacy, then Data privacy, then Get a copy of your data. LinkedIn's help page has the current wording. Choose the larger archive, the one that includes connections, and request it. A download link arrives by email. It is not instant, so ask for it before the evening you plan to do this.

Inside the archive, open Connections.csv. It has one row per first-degree connection and these columns: first name, last name, profile URL, email address, company, position, and the date you connected. The email column is blank for most people, because LinkedIn only includes addresses from connections who allow it. The file may also start with a few lines of notes before the header row, which matters in a moment.

That is the whole file. No past jobs, no schools, nothing about who your connections know.

Step 2: Write down who you want to reach

The prompts work best with a list, not a wish. Write down the ten to thirty funds you would most like to talk to, with the stage and sector that makes each one a fit. If you do not have that list yet, an assistant with web search can help you build one, but check it against each fund's own site. Investor lists in AI answers go stale quickly.

If you are raising for customers rather than capital, the same method works with a list of target companies. How to build a target account list covers that side.

Prompt 1: Who do I already know at these funds

This one only needs the file. Attach Connections.csv and paste:

The attached file is my LinkedIn connections export. It may start with a few lines of notes before the header row; skip them. Columns are First Name, Last Name, URL, Email Address, Company, Position, Connected On.

Here are the investors I want to reach: [list of funds].

Find every connection whose Company matches one of these funds. Allow for name variations, such as a short name, an abbreviation, or "Ventures" and "Capital" added or dropped. Return a table with the person, their position, the fund, the date we connected, and the profile URL. Do not include anyone you are not confident works at that fund.

You will usually get a short list. Someone you know who works at the fund is the most direct path there is, even if it is an associate rather than a partner. They know who covers your sector and how the partnership likes to be approached.

Prompt 2: Which founders I know are in their portfolios

This is the prompt that finds the paths you would never spot by scrolling. Investors tend to weigh an intro from a founder they have backed above almost anything else; Arian Ghashghai puts it first in his ranking of how to reach a VC.

If you're raising VC for the first time, there is a consensus hierarchy of outreach to VCs:

1. warm intro from a founder in that VC's portfolio (the more money that founder has made for the VC, the stronger the referral)

2. warm intro from an employee in that VC's portfolio…

arian ghashghai (@arian_ghashghai) June 16, 2026

Turn on web search, keep the file attached, and paste:

For each fund in my list, find its portfolio on the fund's own website. Then find every connection in my file whose Company matches a portfolio company.

Return a table with the person, their position, the portfolio company, the fund that backed it, and the link to the page where you found the investment. If you could not find a fund's portfolio page, or a match depends on a guess, put it in a separate list called "could not verify". Do not invent investments.

The last two lines matter. Assistants fill gaps with something plausible, and a made-up portfolio company is worse than a missing one. Asking for the source link lets you check each match in a few seconds.

Prompt 3: Rank the paths and draft the ask

Now put the two lists together:

Combine both tables into one list of intro paths, one row per fund. For each fund, rank the paths: founders the fund backed first, then people who work at the fund, then everyone else. Note the date we connected next to each person.

For the top path at each fund, draft a short note from me asking for the introduction, and a separate paragraph about my company that they could forward without editing. Keep the note under 100 words and make it easy to say no.

The forwardable paragraph is the part that gets intros made. A connector who has to write the email themselves often never gets round to it. How to ask for an introduction covers choosing who to ask, and the ChatGPT prompt for a warm introduction email has more prompts for the email itself.

What the CSV cannot tell you

Every path on that list has the same weakness. A LinkedIn connection is not a relationship. A reply in Paul Klay's thread put it well: one founder had four routes to the same partner, three of them LinkedIn-only, and only the real relationship replied within two days. The export cannot tell you which of your connections would recognise your name, because it has no record of a single conversation. The date you connected is a weak hint at best.

It also misses whole categories of path. Past colleagues who have since moved, people you went to school with, and anyone two steps away are invisible, because the file only holds current company and title.

There are a few ways to close that gap. LinkedIn itself will show you second-degree connections at a firm if you open its page and filter its people, one firm at a time. OpenVC's Warm Intro Finder connects Gmail and LinkedIn to show warm paths to investors in its database. Intelligence.com maps second-degree paths across a verified relationship graph.

NetworkOS takes a different route: it reads your email and calendar alongside your LinkedIn connections, so it knows who you have actually written to and met, and how recently. It works inside Claude and ChatGPT, so the same questions in the prompts above run against your real history instead of a CSV, and when you find the path it can send the opt-in ask from your own Gmail. The trade-off is that it needs access to your Gmail, and it can only see relationships that left a trace there or on LinkedIn. Someone you only ever met in person, you still have to tell it about.

Before you send anything

Read the final list once with a cold eye.

• Would each connector recognise your name without looking you up?
• Is every portfolio match backed by a link you have opened?
• Have you asked no more than one person per fund? Two people forwarding the same deck to the same partner looks disorganised.
• Is the forwardable paragraph something you would be happy to see pasted unchanged under someone else's name?

If a fund has no path you trust, a short, specific cold email is better than a weak warm one. The point of the exercise is to find the intros worth asking for, not to ask for all of them.