How to Prompt AI to Write an Investor Update

The fastest way to get a mediocre AI-written investor update is to hand a model a vague instruction like "write my monthly update" and let it fill in whatever sounds good. The fix is a prompt that bans vague superlatives outright, forces every number to come from you, requires bad news to be stated as plainly as good news, and sets aside a real section for what you actually need from investors. Below is a scaffold you can copy, the reasoning behind each rule, and a walkthrough of how to run it.

Why AI-written updates make investors suspicious

Investors who read a lot of these notice the pattern fast. Founders who have studied why AI-generated pitch material tends to fall flat point to the same tell every time: confident language that never lands on a specific number. Phrases like "strong momentum" or "great traction" read as filler the moment a reader has seen a hundred of them, because they carry no information a reader can act on.

An update that says "we made great progress on sales" isn't false, it's just empty. The updates that actually get replies are the ones where every claim has a number attached, and where the writer admits plainly when there isn't one yet.

What to gather before you write the prompt

AI can structure and word-smith an update. It cannot invent your numbers. Pull these together first, even as rough fragments:

• Current MRR or revenue, and the change from last month

• Cash in the bank and current runway in months

• Headcount changes, hires made or open

• Two or three concrete wins, each with a number attached (a customer name and contract value, a launch and its first-week usage, a metric that moved)

• One or two real problems, stated as facts (a missed target, a churned account, a delayed launch)

• What you specifically need from investors this month, and by when

If you don't have a number for something, tell the model that directly in your prompt instead of leaving a blank. Left alone, a language model will reach for a qualitative substitute to cover the gap, and that substitute is exactly the kind of filler that makes an update forgettable.

None of this matters much if you haven't settled the more basic question of whether raising money is the right move for your company right now.

A prompt that refuses to fill gaps with fluff

Paste your raw notes and numbers under this prompt. The instructions matter more than the example structure, they're what keep the output honest. The same rule-based approach works for prompting AI to write a runbook or any other document where accuracy matters more than polish.

You are drafting a monthly investor update from the raw notes and numbers below.

Rules:
- Every number in the update must come from the notes I gave you. Never estimate, round dramatically, or invent a figure.
- If I did not give you a number for something (churn, NPS, pipeline value, whatever), do not substitute a qualitative phrase like "strong growth" or "solid progress." Either omit the claim entirely or write "no figure tracked this month."
- Never use hype language: "crushing it," "explosive growth," "game-changing," "we're thrilled." State facts plainly, in plain sentences.
- Bad news gets stated as directly as good news, in the same tone. Do not attach a silver lining to a miss unless I gave you that silver lining as a fact.
- Structure the update as: Recap (2-3 sentences), Metrics (a short list, numbers only, no adjectives), Highlights (2-4 bullets, each with a number), Lowlights (1-3 bullets, real problems, not humble-brags disguised as problems), Asks (1-3 specific, answerable requests with a timeframe), Next month (1-2 sentences on what changes).
- Keep the whole thing under 400 words. Investors skim these between meetings.
- Write in first person plural, short sentences, no nested bullets.

Here are my notes:
[paste your raw numbers, wins, problems, and what you need here, fragments are fine]

The asks section most founders skip

Y Combinator's own guidance for investor updates treats the ask as one of the three load-bearing sections alongside metrics and a recap, not an afterthought at the bottom. Most AI-drafted updates either omit it or default to something like "always open to intros," which is unanswerable and gets ignored.

Tell the model what a real ask looks like in your notes, then let it tighten the wording, not invent the substance. "Looking for a warm intro to a Series A lead at a fintech-focused fund, we start that process in about six weeks" gets a reply. "Any intros appreciated" does not, because there's nothing specific for a reader to act on.

A prompt that requires "specific, answerable requests with a timeframe" for the asks section, as in the scaffold above, is doing more useful work than any amount of polish on the highlights paragraph.

Making the model report bad news instead of spinning it

Most business writing a language model has been trained on leans upbeat, so left unprompted it will round a miss into "a learning experience" or bury a churned account inside a sentence about pipeline growth. The instruction that fixes this is blunt: bad news stated in the same tone as good news, no silver lining unless you supplied one as fact.

In practice that means notes like "lost our second-largest account, they cited price" should produce a lowlight that says exactly that, not "we're refining our pricing strategy based on customer feedback." The second version isn't a lie, but it hides the fact that matters, and investors who read a lot of updates will notice the hedge.

How to run this in practice

1. Pull your numbers first. Don't let the model estimate or look anything up, feed it the real figures.

2. Drop your notes under the prompt as rough fragments, full sentences aren't necessary.

3. Run it, then read the output specifically for any claim without a number behind it. Cut or fix anything that slipped through despite the rules.

4. Sharpen the asks so each one names a person type, a company type, or a decision, with a rough date attached.

5. Send it, and note which asks actually get a reply. Feed that back into next month's notes, since it tells you what kind of ask this list of investors actually acts on.

If the update leads to a call, and that call produces new commitments, turning meeting notes into action items keeps them from disappearing before the next update.

Feeding a model your real material instead of a topic and letting it improvise is the core discipline behind writing a good AI prompt in the first place, not just for investor updates.

Common mistakes when prompting AI for investor updates

• Letting the model estimate MRR or growth rate from partial numbers instead of giving it the actual figure

• Skipping the asks section, or leaving it generic enough that no one can act on it

• Burying the one number that matters, a missed target or a lost account, inside a highlights paragraph instead of naming it in lowlights

• Not telling the model explicitly to avoid vague language, then being surprised when it defaults to "strong quarter" and "positive momentum"

• Sending an update that's mostly narrative with no metrics section, which forces a busy reader to hunt for the numbers

FAQ

What should a monthly investor update include?

A short recap of the month, a metrics section with real numbers and no adjectives, two to four highlights with numbers attached, one to three honest lowlights, and one to three specific asks with a timeframe. Most founders can fit all of this under 400 words.

How honest should an investor update be about bad news?

Completely. State it in the same plain tone you'd use for good news, without a forced silver lining. Investors who get updates regularly can tell when a miss has been dressed up, and it costs more trust than the miss itself would.

Can AI write an entire investor update on its own?

No. It can structure, tighten, and draft the prose, but it cannot know your MRR, your churn, or what you actually need from investors this month. Feed it your real notes and explicit rules against filling gaps with vague language, and treat the output as a first draft to edit, not a finished update.

What's a good ask to put in an investor update?

One that names something specific enough to act on: an intro to a named type of customer or investor, help closing a particular hire, or feedback on a decision you're making, ideally with a rough date attached. "Always open to intros" isn't an ask, it's a placeholder.

How long should an investor update be?

Short enough to read between meetings, generally under 400 to 500 words. Investors are reading many of these; the ones with a clean metrics section and specific asks get read in full, the long narrative ones often don't.