AI Prompts for Founders: Investor Updates and Fundraising Prep Without Inventing Numbers

The 28th of the month is coming and the investor update is still a blank doc. Last month you promised a hiring update and a pricing test. Revenue moved but not the way you hoped, and you're tempted to bury it in the third paragraph. AI prompts for founders can help with the writing: structure, plain language, a consistent format. They cannot supply the numbers, the honesty or the judgment, and an investor who catches one invented figure will discount everything else you send.

So this pack follows one rule. Every figure, name and claim comes from you. The model arranges and questions. Where something is missing, the output says MISSING. These are untested templates, I haven't run them on a real fundraise, and the example numbers are invented. The arithmetic in the runway example was run with a short script. This post does not give legal, tax or securities advice. For those, talk to a qualified professional.

It's also different from the founder prompt list on this site. AI prompts for Indian startup founders covers pitch material, cold emails and a term sheet explainer. This post is about the ongoing work after you start talking to investors: updates, tracking, diligence questions and runway notes.

How do I write a monthly investor update with AI?

Use a fixed structure so investors can scan it and compare months. Then make the model work only from your inputs.

You are helping a founder write the monthly investor update for [COMPANY],
for [MONTH].

INPUTS (the only facts you may use):
- Last month's update, verbatim: [PASTE]
- This month's metrics, computed by me (do not recalculate or round):
  [LIST: revenue, customers, burn, cash, runway, and any other metric I
  report every month, with definitions]
- What happened (plain list, my words): [PASTE]
- What did not go to plan: [PASTE, EVEN IF UNCOMFORTABLE]
- Commitments I made last month and their status: [LIST]
- Asks (intros, hires, advice) I want to make: [LIST]

FORMAT:
1. Summary: three sentences. Lead with the most important fact,
   good or bad.
2. Metrics: a table with this month, last month, change. Use only the
   numbers I gave. Keep definitions unchanged.
3. Wins (max 3), Misses (max 3), each with one sentence of cause.
4. Status of last month's commitments: DONE / NOT DONE / PARTIAL, with
   the reason.
5. Asks: specific, one line each, with the kind of person who could help.

RULES:
- No number, customer name, quote or date that is not in my inputs.
- No adjectives that sell ("strong", "exciting", "huge"). Facts and causes.
- If the misses list is shorter than my "did not go to plan" input,
  tell me what you dropped.
- If something I said last month is missing from this month's inputs,
  list it under MISSING.

The last two rules matter. A model will quietly trim a miss because it makes the summary read better. And investors notice when a commitment from last month just disappears from the next update.

Keep the definitions fixed

If you change how you count customers, active users or revenue, say so in the update and show both figures for the month the change took effect. A model will not warn you; paste your definitions every time. If you switch from "paying accounts" to "accounts with any usage" to get a bigger number, no prompt will make that a good idea.

How do I check the runway number before I send it?

Compute runway yourself and paste the result. Here is the arithmetic I ran on invented figures (cash 84 lakh, three months of net burn):

Average net burn (3 mo): 1,086,667
Runway at average burn: 7.7 months
Runway at latest month burn: 7.7 months
MoM 1: 3.9%
MoM 2: 5.9%

These are invented numbers for illustration, not a benchmark for any company. The method is the part to copy. Use net burn (cash out minus cash in), not gross spend. State which window you averaged. If burn is rising, show runway at the latest month as well, because the average can flatter you. Then give the model the results:

Here are runway figures I computed and the method. Do not recalculate.
Method: [DESCRIBE]. Figures: [PASTE].
Write two sentences for the investor update that state runway, say what
would change it (hiring, a customer loss, a price change, a delayed
collection), and say what we plan to decide and by when.
Do not suggest that more fundraising is the answer. Do not use words like
"comfortable" or "healthy".

When to raise the subject of a short runway with your investors is your call and your board's. The point of the prompt is to make the sentence precise, not to reassure.

How do I track investor conversations without making things up?

A fundraise is a pipeline. Keep a simple tracker and let the model help you read it. This is the one place the sales mindset actually transfers. The same hygiene ideas from AI prompts for sales ops apply: dated next steps, a named contact, no vague "following up".

INVESTOR TRACKER (my data; columns: investor code, stage [intro asked /
intro made / first call / partner meeting / diligence / term discussion /
passed], last contact date, what they asked for, what I owe them, next
step with date, source of intro):
[PASTE]

TODAY: [DATE]

Produce:
1. Rows where I owe something and the date has passed.
2. Rows with no dated next step.
3. Rows with no contact in more than [N] days that are not marked passed.
4. For each, one short follow-up email draft that restates what I owe or
   what they asked for, in under 80 words.

RULES:
- Do not predict who will invest or pass.
- Do not summarise an investor's interest. Quote the tracker.
- Do not invent anything an investor said. If the tracker is blank, say
  so.
- Do not draft emails that make claims about traction, forecasts or
  terms. Leave a placeholder [CLAIM TO CONFIRM] for me to fill or cut.

The last rule keeps you away from the part of investor communication that needs professional input. What you say about forecasts, traction and the terms of a raise can have legal consequences that vary by country. Have a lawyer who knows securities rules set the guardrails once, and then keep your drafts inside them.

How do I prepare for diligence questions?

Investors in later conversations will ask for evidence behind your claims. Use the model to find the gaps in advance, from your own deck and update history.

I am preparing for investor diligence. Here are the claims in my current
deck and last three updates, verbatim: [PASTE]

For each claim containing a number, date or comparison:
1. State the claim.
2. List the document I would need to show to support it (for example:
   a bank statement, billing export, signed contract, cohort query).
3. Say whether the three updates are consistent with each other on
   this claim. Quote any difference.

Then list ten questions an investor might ask about this material, each
tied to a specific line in my text. No questions about topics my text
does not mention.

Do not answer the questions. Do not assess whether the claims are good.

Do not ask it to draft answers about legal structure, share classes, IP ownership or compliance. Those go to your lawyer or accountant, and a confident wrong answer in a data room is worse than a delay.

Can AI help me write the hard emails?

Yes: bad news about a missed milestone, a request for a bridge conversation, or a thank-you to an investor who passed. Give it the facts, tell it the tone you want, and keep the decision yours.

Draft a short email from me to [INVESTOR CODE / ROLE].
Situation: [WHAT HAPPENED, WITH DATES]
What I know and don't know: [LIST]
What I plan to do: [ACTION AND DATE]
What I am asking for: [ONE THING, OR NOTHING]

Under 130 words. Lead with the news. No apology beyond one specific
sentence. No spin. Do not mention terms, valuations, share numbers or
promises about future performance.

If an email touches money, terms or a promise about the future, show it to your lawyer or lead investor before it goes out.

A worked example (illustrative)

An invented scenario to show the output shape, not a real run. A founder pastes: last month's update promised "ship usage-based pricing and hire one engineer". This month's inputs say pricing shipped on the 18th, the hire did not close, one customer cancelled, cash is 84 lakh, and net burn rose.

What a good update summary looks like (shape):

Summary: Usage-based pricing shipped on 18 Oct. The engineering hire did
not close; the candidate declined on 24 Oct. One customer cancelled
(reason in Misses). Runway is 7.7 months at the three-month average burn.
Commitments from September: pricing - DONE; engineer hire - NOT DONE
(offer declined, search reopened, target date to be set).
Misses: (1) Hire not closed: candidate declined; we are sourcing again.
(2) Customer cancellation: [REASON MISSING from inputs].
Asks: introductions to backend engineers with early-stage experience.
MISSING: this month's revenue figure; definition of "active customer".

The model flagged what it didn't have instead of filling it in. That's the behaviour you're paying for.

Failure modes to check for

What not to automate

• The numbers. Compute and define them yourself. You are the one signing off on them.
• The decision to raise, how much, and on what terms. Talk to your co-founders, advisers and a lawyer.
• Anything about securities, share issuance or legal structure. Ask a professional, every time.
• The relationship. An investor who passed on you in this round may fund the next. Write that email yourself.
• Confidential documents in unapproved tools. The AI use policy template is a starting point for deciding what is allowed.

For the habit of constraining the model to your own inputs, read the lesson on avoiding hallucinations. For structuring what you paste, see giving context. If you share updates on LinkedIn too, prompts for LinkedIn content covers that, with a note: public posts about a raise are a legal question as well. The prompt library has more templates, including a status report you can adapt.