ChatGPT Prompts for Founders (2026): 36 Prompts With Outputs

Key Takeaways

• A founder prompt that works has four parts most prompts skip: your real numbers, a hard constraint, an output format and a destination. "Put it on my Launch board in Storyflow" turns an answer that scrolls away into a plan the team opens tomorrow.
• The 36 prompts are grouped by the work a founder does in order: set up context, research customers, decide strategy, plan the launch, raise, hire, and run the week.
• Each prompt lists what comes back in chat and, where it fits, the board it builds through the Storyflow plugin. Plugin prompts that read other tools (Stripe, Granola) need those plugins installed too.
• Put your company context in a ChatGPT Project's instructions once. OpenAI says project instructions override your global custom instructions inside that project, so every prompt starts informed.
• Weak output is almost always a missing input. The fix table below maps eight symptoms (invented numbers, generic advice, flattery, a messy board) to the edit that fixes each.
• Storyflow's plugin runs on ChatGPT's model, so it spends no Storyflow AI credits; Storyflow itself is paid-only during early access, from $7.99 a month billed annually.

Why Most Founder Prompts Get Generic Answers

You ask ChatGPT for a go-to-market plan and get one that would fit any startup ever founded: "identify your target audience, leverage social media, measure results". It is not wrong. It is the average of everything ChatGPT has read about go-to-market plans, because the prompt gave it nothing else to work with.

The model is not the bottleneck. The prompt is.

ChatGPT does not know your numbers

It does not know you have 18 paying pilots, nine months of runway and a launch on March 3. Without those, it writes for a hypothetical company, and the advice is hypothetical too. The single biggest upgrade to any founder prompt is a line of real numbers.

It will invent what you do not give it

Ask for a market size without inputs and you get a confident figure with no source. Ask for an investor update without metrics and you get placeholder growth. OpenAI's own prompt engineering guide says GPT models do best with precise instructions that spell out the logic and data needed. For a founder, that means: use only the numbers I give you, and write "unknown" for the rest.

The answer has nowhere to go

Even a sharp answer dies in the chat. Your cofounder reads it only if you share the chat or its Project, and even then the plan sits inside a thread, so it gets copied into a doc, the doc drifts from the chat, and ChatGPT can no longer see the version the team works from. A destination in the prompt fixes that: with the Storyflow plugin installed, ChatGPT writes the plan onto a board and can read it back next week.

Set Up Once: A Project and the Storyflow Plugin

One setup pass makes all 36 prompts shorter, because the context they share is written once.

Put your company context in a ChatGPT Project

OpenAI's Projects article says Projects keep related chats, files and instructions together, and that project instructions override global custom instructions inside the project. Make a project per company or per raise, paste Prompt 1 below into its instructions, and add your deck and metrics as files. Connected apps work in project chats: select + or refer to the app by name.

Install the Storyflow plugin

Open Storyflow's listing in ChatGPT's plugin directory ("Plan projects on visual boards", by Storyflow LTD), select Install plugin, sign in to Storyflow, and choose which projects ChatGPT may use: all, or the ones you pick. Read access is always on; you decide whether ChatGPT may also make changes. In a chat, type @ and pick Storyflow, or choose it from the + menu. Availability depends on your ChatGPT plan, workspace and region. Our plugin install guide covers every step, and OpenAI's Plugins in ChatGPT article is the source.

Decide when ChatGPT asks first

OpenAI's permissions article offers Always ask, Allow read actions, Allow low-risk actions and, for some apps, Allow all actions. Start with Allow read actions so ChatGPT reads boards freely and asks before it builds.

Know what your prompts share

A plugin sees only what its connected account can access, and ChatGPT may share relevant context from the conversation with the app to answer you. OpenAI's connected apps article says information accessed from apps is not used to train models by default on Business, Enterprise and Edu; on Free, Plus, Go and Pro it may be if Improve the model for everyone is on. Decide before you paste cap tables or salaries.

The 36 Prompts at a Glance

Replace anything in square brackets. Where a prompt ends with a board instruction, drop that clause to keep the output in chat.

Stage 1: Set Up the Context

These three prompts give every later prompt something to stand on. Run them once per company, then again when the numbers move.

Prompt 1: Write the project instructions

"You are helping me run [company], a [stage] startup selling [product] to [customer]. We are [number] people with [monthly revenue, or no revenue yet] and [months] of runway. Our next milestone is [milestone] by [date]. Use only numbers I give you and label everything else as an assumption. Prefer tables to paragraphs. When I say board, build it in my [company] project in Storyflow."

In chat: nothing to read; paste it into your ChatGPT Project's instructions. On the board: every later prompt that says "board" lands in the right project.

Prompt 2: Build the company fact sheet

"Interview me one question at a time until you can fill a one-page fact sheet: customer, problem, product, pricing, traction with dates, team, competitors and our three biggest risks. Then show the sheet as a table and list every field you had to guess. Put it on a Company board in Storyflow as one wall with a heading per field."

In chat: eight to ten questions, then a two-column table and a list of guessed fields. On the board: a Company wall the rest of the prompts read from.

Prompt 3: Find what you cannot yet prove

"Read my Company board in Storyflow and list what is missing for a seed pitch: claims without evidence, numbers without a date, and questions an investor would ask that the board cannot answer. Add a Missing wall with one to-do per gap."

In chat: a gap list grouped by claim, number and question. On the board: a Missing wall of to-dos, which becomes the research plan.

Stage 2: Research Customers and the Market

Research prompts go wrong when they ask ChatGPT for the answer instead of the method. These ask it to structure your evidence, not replace it.

Prompt 4: Write a customer interview guide

"Write a 30-minute interview guide for [customer role] about how they handle [problem] today. Ask about what they did last time, not what they think of my idea; no question may mention our product. Group the questions as opening, current workflow, the last time it went wrong, what they spend, and close. Put the guide on my Research board in Storyflow as a document."

In chat: 12 to 15 questions in five groups. On the board: one shared guide, so every founder runs the same interview. The rule against pitching follows Rob Fitzpatrick's The Mom Test, a book on learning from customer conversations that accelerators such as Seedcamp teach from.

Prompt 5: Turn a month of calls into an objections board

"Using @Granola, read my customer calls from the last 30 days. List each objection, how many calls raised it and one direct quote. Then put them on my Research board in Storyflow: one wall per objection with the quotes as notes, and a table of objection, count and our current answer."

In chat: a ranked objection list with counts and quotes. On the board: walls by objection plus a summary table. Granola's ChatGPT integration limits its free Basic plan to your own notes from the last 30 days, which is why the prompt says 30.

Prompt 6: Map three competitors

"Compare [competitor A], [competitor B] and [competitor C] for [customer]: pricing as published on their own sites, target customer, the job each one wins and where each is weak. Mark anything you could not confirm. Then put it on my Research board in Storyflow as a comparison table and a positioning map with price on one axis and [capability] on the other."

In chat: a comparison table with unconfirmed cells marked. On the board: the table plus a two-axis map of notes. Our ChatGPT competitor analysis page shows the full board.

Prompt 7: Size the market from the bottom up

"Build a bottom-up market estimate for [product] in [region]: the number of target customers, the share we could reach in three years, and the price per year. Put each assumption on its own row with its source or the words my guess, show the arithmetic, and do not give a top-down figure. Add it to my Fundraising board in Storyflow as a table."

In chat: an assumptions table and the multiplication behind the result. On the board: the same table, so you change one assumption and ask ChatGPT to redo the arithmetic.

Prompt 8: Write the ideal customer profile

"From my Research board, write the profile of our first 50 customers: company type, size, the trigger that makes them look for a tool, who signs, who uses it daily, and who we should turn away. Put it on my Research board as a persona wall."

In chat: a profile in six labelled parts. On the board: a persona wall beside the evidence it came from. Our ChatGPT buyer persona page goes deeper on the format.

Prompt 9: Draft a validation survey

"Draft a seven-question survey to test the top objection on my Research board with [customer]. Multiple choice where possible, one open question last, no leading wording, and one question that would prove the objection wrong."

In chat: seven questions with answer options and a note on what each tests. On the board: usually none; send the survey from your form tool.

A persona board with research quotes beside each wall, the shape prompts 5 and 8 build from call notes.

Stage 3: Decide the Strategy

Strategy prompts fail by agreeing with you. Each of these builds in the counter-argument.

Prompt 10: Write three positioning statements and attack them

"Write three positioning statements for [product], each for a different segment on my Research board, in the form: for [customer] who [trigger], [product] is the [category] that [benefit], unlike [alternative]. Then argue against each in two sentences. Put them on my Strategy board in Storyflow as a mind map with one branch per segment."

In chat: three statements, each followed by its strongest objection. On the board: a three-branch mind map you can extend with evidence.

Prompt 11: Compare pricing models

"Propose three pricing models for [product]: per seat, per location and usage-based. For each give the price points, what a typical [customer] pays a year, how revenue changes from 5 to 50 customers, and what would prove the model wrong. Use only the numbers I gave you. Put the comparison on my Strategy board as a table."

In chat: a three-column comparison with a row per question. On the board: a pricing table next to the objections it has to answer.

Prompt 12: Run a pre-mortem

"It is [date six months from now] and our launch failed. Write the five most likely reasons, ranked. For each, give the early warning sign we would see in the first four weeks and one thing we can do this month to lower the risk. Put them on my Launch board as a Risks wall with one to-do per action."

In chat: five ranked failure modes with signals and actions. On the board: a Risks wall whose to-dos you assign on the board.

Prompt 13: Write a decision memo

"I have to decide [decision]. Write a decision memo under 300 words: the options, what each costs in money and weeks, what we learn from each, your recommendation, and what would change it. Put it on my Strategy board as a document."

In chat: a memo in five labelled parts. On the board: the memo stays beside the evidence, so the next person asking why can read it.

A mind map on a Storyflow board, the shape prompt 10 builds with one branch per segment.

Stage 4: Plan the Launch

The launch is where a plan in the chat costs the most, because four people need the same version on the same day.

Prompt 14: Build the launch board

"Build a launch board in my [company] project in Storyflow for a [launch type] on [launch date]: walls for Positioning, Audience, Channels, Assets, Launch week and Risks, and a task table with task, owner, due date and status, rows kept in date order."

In chat: a short summary of what it built. On the board: six walls and a task table. The owner column links to teammates only when everyone named is already on the board; if one name matches nobody, the whole column stays plain text, so invite the team first. Our ChatGPT product launch plan page shows the layout.

Prompt 15: Work back from the launch date

"Working back from [launch date], list every task for a [launch type] by a team of [number], with the latest safe start date for each and which task has to finish before another can start. Flag anything that needs more than one person. Add the tasks to the table on my Launch board."

In chat: a dated task list with the blocking order written out. On the board: new rows in the task table. Storyflow has no dependency links or Gantt chart, so the blocking order lives in the task text.

Prompt 16: Outline the landing page

"Outline the launch landing page: five headline options in our customers' own words from my Research board, three proof points we can actually support, a section answering the top objection, and one call to action. No invented testimonials or numbers. Put the outline on my Launch board as a wall."

In chat: headline options, proof points and section order. On the board: an Assets wall your designer works from.

Prompt 17: Rank the launch channels

"For [customer], rank five launch channels by how directly each reaches the person who signs. For each give what we post or send, the hours a week it needs from us, and the one number that tells us it is working. Put the ranking on my Launch board as a table."

In chat: a ranked table with effort and a signal metric. On the board: the Channels wall gets the table.

Prompt 18: Write the launch-day run sheet

"Write a launch-day run sheet from 7:00 to 19:00 in 30-minute rows: who does what, which links go out where, who answers support, and when we check signups. Put it on my Launch board as a table."

In chat: a 24-row schedule. On the board: the run sheet the whole team opens on the day.

Prompt 19: Review the launch two weeks later

"Read my Launch board in Storyflow and tell me which tasks never got done, which still have no owner, and what we should cut from the plan next time. Then add a Lessons wall with one note per lesson."

In chat: three lists. On the board: a Lessons wall, which becomes the starting point for the next launch board.

Stage 5: Raise the Round

Fundraising prompts carry the highest cost of an invented number, so every one of these tells ChatGPT where its facts come from.

Prompt 20: Draft the monthly investor update from Stripe

"Using Stripe, get last month's new revenue, revenue from active subscriptions and the number of cancelled subscriptions, each with the month before for comparison. Then draft our investor update: the metrics first, then up to three asks, then a five-line recap of wins and misses. Under 250 words. Mark any figure you could not pull."

In chat: a draft update with sourced metrics and gaps marked. On the board: optional; ask for the metrics table on your Fundraising board so each month sits beside the last. The order follows advice from Y Combinator partner Aaron Harris, summarized by Visible.vc: metrics first, asks as close to the top as possible, the recap at the bottom. Stripe's MCP docs describe what its plugin can read; check every figure against the Stripe Dashboard before you send.

Prompt 21: Write the story in ten sentences

"Write our fundraising story in ten sentences, one per slide: problem, why now, product, traction, market, business model, competition, team, the raise, and what it buys. Use only facts from my Company and Research boards, and mark any sentence that needs evidence we do not have yet."

In chat: ten sentences with evidence gaps marked. On the board: none yet; this is the outline for prompt 22.

Prompt 22: Build the deck on the board

"Build a 10-slide deck on my Fundraising board in Storyflow from the story above, one slide per sentence, with the metrics table on the traction slide. Leave a note beside any slide where the evidence is thin."

In chat: a slide list. On the board: a deck next to the research it cites; ChatGPT can add, remove, reorder and update slides later. Storyflow decks present fullscreen and export as PDF, with no PPTX export or speaker notes. More in our ChatGPT presentation guide.

Prompt 23: Rehearse the hardest questions

"Act as a seed investor who passes on most deals. Read my Fundraising board and ask the 15 hardest questions you would ask in a first meeting, ranked by how likely each is to end the meeting. Then write a two-sentence answer to each using only what is on the board, and say which answers are weak."

In chat: 15 ranked questions with draft answers. On the board: ask for a Q&A wall if you want the team to rehearse from one list.

Prompt 24: Build the investor list

"Make a table for our [round] raise with columns for firm, partner, stage they lead, check size if published, warm intro path, status, next step and date. I will paste the names; leave unknown cells blank rather than guessing. Put it on my Fundraising board."

In chat: the empty or partly filled table. On the board: a pipeline table; for more than 50 firms, a CRM such as HubSpot holds it better.

Prompt 25: Write the investor one-pager

"Write a one-page summary for investors who will not open a deck: a one-line pitch, the problem, traction with dates, the team and the raise. Under 250 words. Put it on my Fundraising board as a document."

In chat: the one-pager. On the board: a document you export as PDF. See our ChatGPT one-pager page.

Prompt 26: Prepare the diligence checklist

"List what a seed investor typically asks for in diligence from a [entity type] startup, grouped as company, finance, legal, product and team. Mark which items already exist on my boards. Add the rest to my Fundraising board as to-dos."

In chat: a grouped checklist. On the board: to-dos for the gaps. It is a starting list, not legal advice; your lawyer's list wins.

Stage 6: Hire the First People

Hiring prompts go generic fastest. Tie every one to a milestone and a budget.

Prompt 27: Plan the next 12 months of hires

"Given [months] of runway and our milestone of [milestone] by [date], propose a hiring plan for the next 12 months: role, why now, the milestone it unblocks, start month and the salary range you assume. Label every cost as an assumption. Put it on my Team board in Storyflow as a table."

In chat: a hiring table with reasons and assumed costs. On the board: the plan, where it sits beside the runway numbers.

Prompt 28: Write the scorecard before the job post

"Write a scorecard for our first [role]: the outcomes we expect at 6 and 12 months, five competencies with what good looks like for each, and three red flags. Nothing that would fit any company."

In chat: outcomes, competencies and red flags. On the board: add it to the Team board so interviewers score against the same sheet.

Prompt 29: Write the job post from the scorecard

"Write the job post for [role] from the scorecard: what they will ship in their first 90 days, who they work with, the salary range [range], and what we do not offer yet. Under 400 words, no buzzwords."

In chat: a post ready to edit. On the board: usually none; it goes to your careers page.

Prompt 30: Design the interview loop

"Design a four-stage interview loop for [role] that tests each scorecard competency exactly once, including one work sample we can review in under an hour. Put the loop on my Team board as a flowchart with each stage connected to the next."

In chat: four stages with what each tests. On the board: a flowchart with labelled connections.

Prompt 31: Write the first 90 days

"Write a 30-60-90 day plan for the new [role]: what they learn, what they ship and who they meet in each period. Put it on my Team board as three walls with one to-do per item."

In chat: three lists. On the board: three walls of to-dos; set the assignee and due dates on the board yourself.

Stage 7: Run the Week

These five replace the parts of a founder's week spent assembling status. They work best once a board exists.

Prompt 32: Start the weekly review from what changed

"What changed on my Launch board in Storyflow in the last week, and who changed it? Group the changes by wall, then list tasks past their due date and anything nobody touched."

In chat: a change log by wall plus overdue and untouched items. On the board: nothing new; this one reads. The Storyflow plugin reports changes over the last day or week.

Prompt 33: Pick the week's priorities

"From my Launch board and last week's changes, pick the three things that most move [milestone] this week, one per founder, and the one thing we should stop doing. Explain each in one sentence."

In chat: three priorities and one stop. On the board: ask for a This week wall if the team wants it pinned.

Prompt 34: Turn the team meeting into to-dos

"Using @Granola, read today's team meeting and list every action item with the person who said they would do it. Then add each one to my Launch board in Storyflow as a to-do, with that person's name and the date they gave written in the text."

In chat: the action list with names. On the board: one to-do per action; set the assignee and due date on the board, since the text only names them.

Prompt 35: Clean up a board that grew messy

"Read my Research board and reorganize it: group loose notes into walls by theme, move everything about [old plan or segment] into an Archive section, and list any notes that repeat each other."

In chat: a summary of what moved and a list of likely duplicates. On the board: walls by theme and an archive section. ChatGPT has no tool for deleting board elements through the plugin, so you delete the duplicates yourself. Allow changes for this one, and check the board after.

Prompt 36: Check the plan against the week

"Read my Company board and this week's priorities. In five bullets, tell me where our stated plan and what we actually worked on disagree."

In chat: five bullets. On the board: none; this is the question a cofounder should ask on Fridays.

A project tracker with each job's stage, next step and due date, the board prompts 32 to 34 read from and write to.

Worked Example: An Illustrative Prompt Chain From Calls to a Deck

Tarnwell is a fictional five-person startup selling inventory software to independent bike shops. It has 22 shops on paid pilots and plans to raise a seed round in eight weeks. Here is the chain its founder, Priya, could run in one ChatGPT Project, with the output shape each step should produce. The figures are part of the illustration, not results.

What the chain is doing

Every step reads the board the step before it wrote. The objections inform the positioning; the positioning sets the story; the story becomes the deck; the hard questions point straight back at the two guessed numbers from step 1. A chain beats a single prompt because each answer is checked against something that already exists.

Where Priya intervenes

She corrects the fact sheet before anything else reads it, deletes one objection that came from a single talkative shop owner, and rewrites the traction slide herself. ChatGPT drafts; she decides what is true.

What she does not ask ChatGPT for

Valuation, legal terms, and anything the board cannot support. Prompt 23 exists so the weak spots surface in rehearsal rather than in the meeting.

The Prompt Pattern Behind All 36

Every prompt above follows the same structure. Use it to write your own.

Order matters less than presence

OpenAI's guide suggests identity, instructions, examples and context in roughly that order for developer prompts. In a chat, the bigger win is that each element is there at all. A prompt with all seven in plain sentences beats a beautifully formatted prompt missing two.

Examples are the shortcut

If you have an investor update you liked, paste it and say "match this format". One real example does more than three adjectives.

How to Edit a Prompt When the Output Is Weak

Weak output is diagnostic. Read what went wrong, then change the input that caused it rather than asking ChatGPT to "try again".

When it invents numbers

Do not correct the number in chat and move on; the next prompt will invent another. Add the rule to your Project instructions (Prompt 1) so it applies everywhere.

When the board comes out messy

Ask ChatGPT to read the board back and reorganize it (Prompt 35), or rerun the build prompt with the walls named. Naming the table columns in the prompt is the most reliable way to get a table you can use.

When the plugin does not respond

OpenAI's troubleshooting article suggests starting a new chat and asking for a specific supported task, such as finding a document by its title, before reconnecting. If a connected app still is not used, do not keep reconnecting; check the account and workspace first.

When you attach a photo

A photo attached in the ChatGPT chat does not reach Storyflow. Ask ChatGPT for the board's upload link, add the photo there, then ask it to place the photo.

What the Storyflow Plugin Can and Cannot Do With These Prompts

Through the plugin, ChatGPT can read your projects, boards and documents, report what changed in the last day or week, and build notes, to-dos, tables, documents, walls, mind maps, flowcharts, charts, dashboards and slide decks. It can move and group elements and add images, and create a new project if you granted all projects.

It cannot generate AI images, read board comments or run the Storyflow app's own menu workflows; those belong to the in-app Storyflow AI, which is separate and uses Storyflow AI credits. Plugin actions run on ChatGPT's model and your ChatGPT plan. An invited free Storyflow account can make 25 plugin actions a week (resetting Monday, UTC); Plus, Pro and Max have no weekly limit.

When These Prompts Are the Wrong Tool

You need a number you can defend

ChatGPT can structure a market model; it cannot source the inputs. Every figure in an investor update or a deck needs a source you can name.

You need legal or tax answers

Prompt 26 and the hiring prompts produce starting lists. Equity, contracts, employment terms and diligence documents go to a lawyer.

Your team already works from a wiki

If decisions live in Notion pages and the team reads them there, change the destination clause to Notion and keep the rest of each prompt. OpenAI's Notion app article says ChatGPT may be able to create or update Notion pages, depending on the connection and your permissions. Our Notion vs Storyflow in ChatGPT comparison covers which jobs fit which.

The Bottom Line

Good founder prompts are not clever. They carry your numbers, a rule against guessing, a format and a place for the answer to live. Start with Prompts 1 and 2 today, so ChatGPT knows the company. Then run the chain that matches your next deadline: 4 to 8 if you are still learning the customer, 14 to 19 if a launch date is set, 20 to 23 if you raise this quarter. With the Storyflow plugin installed, each answer lands on a board your team opens tomorrow, and ChatGPT can read it back next week. For the rest of the stack, see our best ChatGPT plugins for startups, and for the research board itself, our ChatGPT research board page.