Founders do an unreasonable amount of writing. Every week there's a customer interview summary, a weekly investor update, a pitch deck narrative, a job description, a GTM plan, a board memo, a slack town-hall message, and roughly a dozen other artifacts that each deserve more thought than you have hours for. The compounding problem is that the quality of founder writing quietly shapes outcomes — investor updates color future check decisions, JDs color the candidate pool, GTM briefs color how the team executes.

AI is genuinely useful here, but only if you prompt it like a founder with context, not like a tourist asking for help. "Write me an investor update" produces something no investor would believe. A structured prompt that ingests your metrics, your progress this week, your asks, and your honest read of the risks produces a draft an investor would actually trust. This guide collects those kinds of prompts: 18 templates across customer research, fundraising, hiring, GTM, and operational writing that founders do over and over.

Every prompt below is designed to be copy-pasted with your real context filled in. They work with ChatGPT, Claude, Gemini, or any capable model — and pair well with Promptwright, where you can keep your entire founder prompt library versioned and shareable with your team.

Why founder prompts need to be different from generic ones

Most online "AI for startups" guides recycle the same handful of one-liners: "ask AI to write your business plan", "ask AI for marketing ideas", "ask AI to summarize customer feedback." These are fine for a teenager exploring an idea; they're useless for an actual founder running a real company. Three things distinguish prompts that work for founders:

Founder prompts must protect confidential information patterns. You're feeding AI real metrics, customer names, salary bands, and deal terms. That doesn't mean you can't use AI — it means you should pasteurize the obvious identifiers when using a tool that isn't under your control, and you should always read the model's terms for sensitive inputs.

Founder prompts must be honest about uncertainty. A lot of founder writing is half-truth by omission — you don't know whether a customer will renew, you don't know whether your hiring budget holds, you don't know whether the next round closes. A prompt that asks AI to "draft an optimistic update" will produce fiction. A prompt that asks AI to "draft an update that is honest about three known risks but framed constructively" produces writing that builds trust with investors over time.

Founder prompts must produce editable drafts, not finished copy. The point is to save 70% of the writing time, not to outsource your judgment. Every prompt below is structured to give you a strong first draft with the hard structural thinking done — and then you ship the version you actually stand behind.

Customer research and访谈 synthesis

Customer interviews are the highest-leverage activity a founder does and the worst-documented one. Most founders conduct interviews, scribble notes, and lose 80% of the insight within a week. AI turns transcripts into patterns you can act on.

Prompt: Customer interview synthesis with structured framework

Act as a senior product researcher who has interviewed hundreds of early-stage customers. Below is a transcript of a customer interview. Produce:

1. **Interviewee context** (3-4 bullets): role, company, relevant context, and how they came to be in our pipeline.
2. **The job they're hiring for** (1-2 sentences, in Jobs-to-be-Done language): what underlying problem are they trying to solve, regardless of how they frame it?
3. **Current workaround** (2-4 bullets): what they actually do today, including manual steps, spreadsheets, or competing tools — even if they didn't explicitly name them as the problem.
4. **Pain points** (bulleted, ordered by severity as expressed): each pain as a one-line quote-style statement in the interviewee's voice, followed by an interpretation in 1 sentence.
5. **Buying signals**: any explicit or implicit indication of willingness to pay, switch, or advocate.
6. **Dissent and concerns**: anything they pushed back on, were skeptical of, or said wouldn't work for them.
7. **Concrete language**: 5-8 short phrases or terms the interviewee used that we should consider adopting verbatim in our messaging.
8. **Open questions for follow-up**: 2-3 things worth circling back on in a future conversation.
9. **Comparison to other interviews**: if I describe patterns from prior interviews below, flag whether this interview confirms, complicates, or refutes those patterns.

Be honest about what the customer did NOT say. Don't extrapolate.

Interview transcript:
[paste]

Prior interview patterns (if any):
[paste or leave blank]

The point of this structure is to force separable observations (what they said, in their words) from interpretation (what we think it means). That separation is what makes synthesis across many interviews tractable.

Prompt: Pattern synthesis across multiple interviews

Below are 5-8 short structured summaries from customer interviews I've conducted in the past two weeks. Each summary has the same structure: interviewee context, job-to-be-done, current workaround, pain points, buying signals, concerns, concrete language, open questions.

Help me synthesize patterns. Cover:

1. **Recurring jobs-to-be-done**: which underlying problems show up in at least 3 interviews? Quote the interview numbers where each appears.
2. **Recurring current workarounds**: which workarounds or competing approaches show up in at least 3 interviews? Quote the interview numbers.
3. **Pain point frequency table**: a table with columns Pain point | # interviews | Severity (High/Med/Low) | Verbatim phrasing example | Likely ICP segment. Sort by frequency.
4. **Dissenting or unique insights**: anything one interviewee said that contradicts the pattern but feels important — flag separately, since one-off insights often point to a segment edge or a hidden assumption.
5. **Three candidate positioning statements** — each one sentence, with the interview numbers that support it most strongly.
6. **Three holes** — what do we still not know that I should test in my next 5 interviews?

Interview summaries:
[paste them, numbered]

Run this every 5-10 interviews, and pattern recognition compounds fast. The synthesis forces you to count instances, which keeps you honest about what's actually recurring vs what felt important because it was the most recent.

Pitch deck narrative and slide-by-slide drafting

A deck is an argument, not a slide template. AI helps with the argument if you give it the structure.

Prompt: Pitch deck narrative arc

Act as an early-stage investor with 15+ years of experience. Help me pressure-test my pitch narrative BEFORE we get to slide design.

I'll paste my raw thinking below. Produce:

1. **One-tightest-sentence version of my thesis**: the kind of sentence that, if an investor believes it, they want to invest.
2. **The argument in 6 moves** — the six claims the deck must prove, in the order you'd prove them. For each:
   - The claim (one sentence)
   - What evidence or section supports it (one line)
   - The objection a skeptical investor will raise (one line)
   - The slide likely needed for this claim (one line)
3. **The riskiest single claim** — where the deck will live or die, with a one-line note on what kind of evidence would shore it up.
4. **Recommended ordering** vs. chronological storytelling: do you recommend leading with the team, the market, the problem, or the product — and why?
5. **Two alternative narrative shapes** that I might be missing — e.g., "the contrarian wave", "the regulatory unlock", "the workflow you already own" — and for each, whether my evidence supports trying it.

Don't recommend slide templates. Don't write copy. Help me make the argument sharper.

My raw thinking:
[paste]

Run this before you sit in Figma. It usually surfaces the one claim you'd buried in a "Vision" slide that should be your entire opening.

Prompt: Individual slide narrative draft

I am drafting the [SLIDE NAME] slide of my pitch deck. The slide's job is: [the claim this slide must make, in one sentence].

Available evidence: [paste concrete metrics, customer testimonies, market data, financials].

Draft 3 alternative versions of the slide's headline-and-subhead pair. For each:

- **Headline** (max 9 words, declarative, no jargon)
- **Sub-headline** (max 20 words, supports the headline with a specific fact)
- **3 supporting bullets** (under 12 words each, specific and verifiable)
- **Visual recommendation** (1 sentence: what chart or diagram would best support this slide)
- **Risks**: what a skeptical investor's first objection to THIS slide would be (1 line)

After the three versions, recommend which version to use given my stage and audience, and explain in 2 sentences.

My company context:
[paste 2-3 sentences about your company, stage, and raise]

Founders who run this prompt per slide end up with decks that read as one coherent argument, not a sequence of slides. The "risks" line per version is where you usually find that you've been choosing your comfortable version rather than your strongest.

Investor updates and fundraising communications

Investor updates are the highest-leverage writing a founder does. They shape whether your existing investors help in the next round, refer customers, or quietly write you off. They are also the writing most founders procrastinate.

Prompt: Monthly investor update draft

Act as a founder writing a monthly investor update. Below are my raw notes — metrics, wins, slips, asks, and reflections from the month. Draft an update using this structure:

**1. Headline** (one short sentence capturing the most important thing this month)

**2. Highlights** (3-5 bullets, each starting with a strong verb, specific metrics or named outcomes — no "we made great progress" vague language)

**3. Lowlights / what didn't work** (2-4 bullets, honest but framed constructively — investors can smell selective reporting, so honest slips build credibility)

**4. Key metrics table**:
| Metric | Last month | This month | Target | Notes |

Leave rows blank where I haven't given you the numbers and ask me to fill them.

**5. Asks** (2-4 bullets — concrete asks for investors: intros to specific customers, candidates, advisors — name the role or company if you know it, otherwise describe the profile)

**6. What I'm thinking about** (1 short paragraph, 3-4 sentences: a strategic question or tension I'm wrestling with. Shows investors you're thinking ahead, not just reporting.)

**7. Thanks** (1-2 bullets: thank specific investors for help this month by name)

Constraints:
- Total length under 500 words.
- No corporate filler, no emojis, no "I'm incredibly excited to share".
- Specific numbers always beat adjectives.
- If a metric went down, say so plainly without panic narration.

My raw notes for the month:
[paste]

The "what I'm thinking about" section is the one most founders skip and the one investors most value. It signals whether you're running the company or being run by it.

Prompt: Investor diligence response (email)

I received the following diligence question from an investor: [paste their question].

Help me reply in a way that:
1. Answers the question directly, without pivoting.
2. Is specific and quantitative where I can be (real numbers, timeframes, named customers where permissible).
3. Acknowledges what I don't know with precision rather than bluffing.
4. Closes with the next step I'm proposing, not leaving the investor to drive.

Draft 2 versions of the reply:
- **Version A** (direct and concise, under 150 words)
- **Version B** (slightly more contextual, under 250 words, for investors who like narrative)

After both, list 2-3 follow-up questions they might send based on this reply, so I can prep answers in advance.

Company context to keep in mind:
[paste 2-3 sentences]

Founder diligence replies often drift into sales-pitch mode and lose credibility. This prompt actively counteracts that.

Hiring and recruiting prompts

Hiring is where founders make the most expensive, most reversible-mistakes. Bad JDs repel the candidates you want. Bad rubrics let you hire the wrong person. AI helps on both — but only with prompts that preserve your judgment.

Prompt: Job description pressure-test

Act as a senior operator who has hired 50+ people in this function. Below is a draft JD I've written for a [ROLE] at [STAGE] [COMPANY TYPE]. Review it for:

1. **Clarity gaps**: 2-4 places where a strong candidate wouldn't actually understand what success looks like. Quote each gap and restate in one sharper sentence.
2. **Scares-off-the-good-ones**: requirements that are overstated, listed without justification, or unintentionally filter out candidates we want. List each, with a one-line suggestion.
3. **Sells-too-hard**: places where the JD sounds corporate-recruiter rather than genuine. List each, with a one-line note on what to cut or rewrite.
4. **Missing essentials**: things a candidate judging this JD would want to know that I haven't told them (team size, reporting line, comp philosophy, growth path, hybrid policy, equity stance). List anything missing.
5. **Real-candidate voice**: 3 sentences drafted in a more founder-direct voice that I might use to replace chunks of the JD's "About You" section.

After the critique, sketch a tightened, founder-voiced intro paragraph (4-5 sentences) that starts with what the role will actually do in their first 90 days.

Draft JD:
[paste]

My context:
[2-3 sentences on company stage, team, and what you've tried so far]

Most JDs get measurably better after one pass with this. The "what you'll do in the first 90 days" frame is the single highest-payoff rewrite — it filters for candidates who imagine doing the job, vs. candidates who imagine having the title.

Prompt: Structured interview rubric

Help me build a structured interview rubric for a [ROLE] hire. Below is what we're trying to test and what I think "great" looks like.

Produce a scoring rubric:

1. **Competencies to assess** (4-6): one line each. Name the competency, not the trait — e.g., "navigates ambiguity" not "scrappy".
2. For each competency, 3 observable behavioral signals:
   - A **green-signal** (excellent) — what the candidate says or does that proves it
   - A **yellow-signal** (mixed) — partial or one-off demonstration
   - A **red-signal** (concerning) — what we'd hear that would worry us about this competency
3. **One suggested assignment or sample work** that would let me assess as many competencies as possible with a single artifact — with 1 line on what to look for in the output and 1 line on what a strong submission looks like.
4. **Two questions I should NOT ask** that are common but actually filter the wrong way for this role — with a one-line note on what to ask instead.
5. **Recommended scoring scheme**: how to turn the rubric into a single 1-5 score, with thresholds for advancing.

What we're testing for:
[paste, even messy notes are fine]

What "great" looks like at our stage:
[paste]

Hiring rubrics are one of the highest-leverage documents in a startup — they get reused silently on every candidate. Getting one good rubric per role saves a fortune in mis-hires.

Prompt: Reference call prep

I'm doing a reference call tomorrow with [references for a candidate for ROLE]. I have a strong lean but want to specifically check for [2-3 specific risks I'd want to verify]. Below is a short note on the candidate's claimed background.

Produce a reference call guide:

1. **Opening framing** (1-2 sentences) to put the reference at ease in a way that gets them to be candid, not cheerlead.
2. **5-7 specific open-ended questions** sequenced from least to most evaluative. Open-ended — no yes/no. For each:
   - The question (1-2 sentences)
   - What a "good answer" would sound like in 1 line
   - What an "evasive or concerning answer" would sound like in 1 line
3. **Questions targeted at the specific risks I named**, in a way that doesn't lead the reference but does probe honestly. 1-2 questions per risk.
4. **The one question to always close with**: the version of "is there anything I didn't ask that I should have?" most likely to surface something useful.
5. **A short "translation" cheat-sheet**: 3 phrases references commonly use that sound positive but often mean something more complicated (e.g., "she's a real driver" → can be abrasive with peers), with the honest read.

Candidate background note:
[paste]

Specific risks I want to probe:
[list]

Reference calls are almost universally underprepared-for. Ten minutes of prep with a structured prompt dramatically improves the signal you actually pull from them.

Go-to-market and growth prompts

Prompt: GTM plan draft from ICP and constraints

Act as a GTM operator who has launched 20+ products. I want a draft GTM plan based on the inputs below. Do NOT write marketing copy. Plan only.

Produce:

1. **ICP sharpening** — restate my Ideal Customer Profile in one tight paragraph, flag any internal contradictions, and propose 2 sharpening questions I should answer before locking it.
2. **Three candidate GTM motions** for our stage and budget, ranked by likely fit. For each:
   - The motion (1 phrase, e.g., "outbound to mid-market ops leaders")
   - Why it fits our stage and ICP (2 sentences)
   - What success would look like in a 90-day test (specific funnel metrics)
   - What it would cost in a 90-day test (rough FTE or dollars)
   - The single biggest risk with this motion
3. **Recommended 90-day pilot plan**: pick the highest-fit motion and lay out a month-by-month plan with milestones, a budget placeholder, and a kill criterion (what signal would make me stop this motion and try the next).
4. **Sequencing vs. your other two**: how does this motion interact with the other two I should NOT pursue yet — which of them become follow-ons conditional on pilot success?
5. **Metrics to instrument from day 1** — a list of 5-7 specific numbers to track, with the question each answers.

Inputs:
- Stage: [paste]
- ICP: [paste]
- Budget / team constraints: [paste]
- Existing channels and what's working or not: [paste]
- Anything I've already tried: [paste]

A good GTM plan earns its keep by forcing one motion, one test, one kill criterion. AI is good at producing those constraints explicitly if you ask.

Prompt: Pricing experiments for a new pricing page

Below is my current pricing and packaging, and what I'm trying to learn. Draft 3 pricing experiment designs I could run in the next 30-60 days:

For each experiment:
- **Hypothesis** in one sentence (specific: what changes, what behavior shift, by how much)
- **Mechanism** — the specific change to the pricing page or sales motion (1-2 sentences)
- **Required sample size** — what kind of traffic or customer interactions would I need to see a real effect (rough estimate)
- **Risks** — what could go wrong, including customer-experience risks, support load, or signaling damage to existing customers
- **What I'd measure** — 3-5 specific metrics with the question each answers
- **Stop rule**: what signal would tell me to kill the experiment and not roll it out

Then give me a single recommendation of which experiment to start with, given my stage and constraints, and one sentence on what to do if it succeeds.

Context:
- Current pricing: [paste]
- What I'm trying to learn: [paste]
- Stage, traffic, customer segment: [paste]
- Constraints: [paste]

Pricing experiments have asymmetric upside — the right one can shift revenue per customer dramatically. They also have asymmetric downside if done ham-fistedly. This prompt keeps you honest about both.

Operational writing prompts

Prompt: Board memo before the meeting

Help me draft a board memo that pre-reads a tough decision I want the board to discuss, not just inform them of. Structure:

1. **Context** (1 short paragraph: what situation forces this decision by this date)
2. **The decision, framed**: 2-3 specific options I'm presenting, one sentence each — including the "do nothing" option.
3. **My recommendation** (1 sentence) and **my confidence** (Low/Med/High, + 1-line rationale).
4. **What I want from the board**: advice on the decision, OR approval of the recommendation, OR connections/inputs to inform the decision. Be specific. If I only want approval if they have serious objections, say that.
5. **Key supporting evidence**: 3-5 bullets max — specific metrics, customer testimonies, financial projections.
6. **Risks and mitigations**: 2-3 bullets — for each risk, one-line mitigation.
7. **Pre-read questions**: 2-3 questions I'd like each board member to think about before the meeting so discussion is sharp.
8. **Appendix if needed**: flag any long appendix I should consider attaching (financial model, customer letters), so I can decide rather than dump it all into the memo.

Tone: factual, low-drama, in the voice of a competent CEO writing to a competent board. No posturing, no soft-pedaling.

Topic, situation, and decision options:
[paste rough notes]

Board memos that ask for discussion rather than approval get better board discussion. The "what I want from the board" line alone reshapes most board meetings.

Prompt: Town-hall / all-hands message

I am writing a brief all-hands message (under 250 words) about [SUBJECT — could be a reorg, a strategy shift, a customer escalation, a fundraise closing, a layoff]. Below are my raw notes on what I want to say.

Draft:

1. **One opening line** that names the subject plainly — no throat-clearing, no "I want to share".
2. **The substance** — 4-6 bullets, each one short direct sentence. Group by: what happened, why it matters, what changes for the team, what does NOT change, what I'm asking of the team.
3. **One honest line about what I don't yet know** — questions on which I'll follow up as I have answers, rather than pretending I have them all.
4. **One line on what this means emotionally or for the team's morale** — only if the subject warrants it; if not, skip.
5. **Closing line**: a forward-looking, specific, not-cheerleader statement about what we do next.

Constraints:
- No emojis.
- No "I'm excited to share".
- No "we're incredibly focused on".
- If layoffs or a reorg, no euphemisms; use the words.
- If good news, no chest-thumping; assume the team wants to know what we do next, not how great we are.

My notes on the subject:
[paste]

The "what I don't yet know" line is the one that earns trust in difficult messages. Teams stop reading corporate all-hands for the same reason investors stop reading selective updates — they assume the unsaid parts are the important ones.

Prompt: Strategic tension memo — naming the thing I can't decide

I'm wrestling with a strategic tension that I haven't been able to resolve. Don't pretend it has an obvious answer — it doesn't. Help me name and frame it.

Produce:

1. **The tension, stated as a real question** — sharp enough that someone could disagree with either answer. (1-2 sentences.)
2. **The two sides**, each in 3-4 bullets:
   - Side A: what this position argues, and what evidence or instinct supports it
   - Side B: what this position argues, and what evidence or instinct supports it
   - Where the two sides actually agree (often more than either admits)
3. **What changes if I commit to Side A**: 3-month, 12-month, and 24-month consequences — what I gain, what I foreclose, what I can't undo.
4. **What changes if I commit to Side B**: same structure.
5. **What I would need to learn, in the next 60 days, to make this decision with more confidence** — 3 experiments or inputs.
6. **A decision I could make today that's reversible**: a path that commits to neither side fully but lets me learn — if such a path exists. If not, say so clearly.

My rough thinking on the tension:
[paste, messy is fine]

Most strategic paralysis comes from framing a tension as a "decision" when it's actually about which information to gather next. This prompt surfaces that distinction.

Common mistakes founders make with AI prompts

Treating AI drafts as finished work. The prompts here produce strong first drafts. The first draft is a starting point — your judgment, your tone, your specific numbers and named customer anecdotes are what make founder writing land. Don't skip the editing pass.
Asking for "what should I do?" Founder prompts should help you see the decision more clearly, not outsource it. The prompts above explicitly instruct the model not to recommend — that instruction is doing real work.
Pasting sensitive customer or financial data into public AI tools without thought. Pasture-ize names, dollar amounts, and customer identifiers when using a tool whose data handling you don't control. For internal model deployments or paid tiers with clear data policies, this is less of a concern.
Not curating a prompt library over time. Founder writing is repetitive by nature — investor updates, JDs, interview summaries recur monthly or per hire. A versioned library means each recurring task starts from your best prior draft, not a blank page. Tools like Promptwright are built exactly for this — keep your templates versioned, shareable with co-founders and heads of team, and run them against your preferred models.
Using AI for energy / conviction moments. AI is great at structuring an argument. It is bad at your conviction in your argument. Don't let a model talk you out of the read you have on your company — use it to pressure-test the framing, not to replace your gut.

How to install these as habits

The prompts above only compound if they become reflexive. Three concrete moves to make that happen:

Pair each recurring founder-writing task with one prompt. Investor update → Investor Update Prompt. Customer interview → Synthesis Prompt. JD → JD Pressure-Test Prompt. Within a month the friction is gone.
Save your best-edited versions of each prompt. After you use one, notice what didn't land. Update the constraint that was missing, save the new version. A prompt you've edited three times is a competitive advantage — it does the structured thinking you'd otherwise redo every week.
Share the library with co-founders and key hires. A head of GTM using your GTM prompt and a head of People using your JD prompt both saves time and improves quality. A shared library — like what Promptwright provides — is the right vehicle for this.

Get started

If you only adopt one prompt from this guide, adopt the monthly investor update prompt. Run it once and you'll see why: it does the boring structural work of an update in seconds, leaving you to spend the saved time on the parts that actually require your judgment — what to be honest about, what to ask for, what to think out loud about.

If you only adopt two, add the customer interview synthesis prompt. Customer insight decays faster than any other kind of learning in a startup; structured synthesis is what stops it from evaporating.

For keeping your evolving founder prompt library in one place — versioned, shareable with co-founders, runnable against leading models without the back-and-forth — sign up at Promptwright. The first prompt you save and the one you've refined the most, six months from now, are the difference between a founder who's writing their company forward and one who's writing the same thing every week.

AI Prompts for Founders and Startups: Templates for Strategy, Fundraising, Hiring, and Growth