# What questions do seed investors ask AI founders in the US? A step-by-step guide. Published 17 November 2026 by the immut team. https://www.immut.io/investor-questions/seed/ai/us ## Key takeaways As of 17 November 2026, US seed investors mostly ask AI founders about the risks, the round and how long the cash lasts. - US buyers often ask for a SOC 2 report, so book your first audit before it slows deals. - Know which sector rules touch you, such as FDA device rules, New York City bias audits or California's rules. - Investors will split your round between people and computing, and ask about the federal research credit. - Prepare churn measured by customers and by dollars, and show what improves as more customers use you. ## Why prepare for these questions? Every investor asks in their own way, and most ask far more than your deck answers. The questions are predictable: prepare once, with a number, a date and a short answer for each. ## Step 1: Learn the 50 questions seed investors ask AI founders in the US The example answers are made up, to show what a clear answer sounds like. ### What do investors ask about risks? They want to know what could go wrong and what you are doing about it. #### 1. Do you have a SOC 2 report, or a date for your first audit? Show: Show the report, or the auditor and date you have booked. List the US buyers who asked for it in their security review. Say how long the gap may slow deals. Example answer (made up): Not yet. Our SOC 2 Type I audit is booked for February. Four US buyers asked for it this quarter, and two of those deals may slip about two months. Source: [AICPA and CIMA: System and Organization Controls (SOC) suite of services](https://www.aicpa-cima.com/resources/landing/system-and-organization-controls-soc-suite-of-services) #### 2. If your AI helps employers screen people, have you had a bias audit for New York City's law? Show: Say whether customers use your tool for hiring or promotion in New York City. Show the latest bias audit, or when one is booked. Say who publishes the results. Example answer (made up): Yes, three customers use us to screen candidates in New York City. An independent auditor finished our latest bias audit in June, and each customer posts the summary on its careers page. Source: [NYC Department of Consumer and Worker Protection: Automated Employment Decision Tools](https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page) #### 3. If your AI helps with patient care, does it need FDA marketing authorization? Show: Say what your product claims to do for patients. Show whether it counts as a medical device, and your plan if it does. Name any similar authorized device you have studied. Example answer (made up): Our product drafts visit notes for doctors to review and makes no diagnosis. Our regulatory adviser says that keeps it outside device rules. If we add diagnosis, we would plan a 510(k). Source: [FDA: Artificial Intelligence-Enabled Medical Devices](https://www.fda.gov/medical-devices/artificial-intelligence-enabled-medical-devices/list-artificial-intelligence-enabled-medical-devices) #### 4. Do California's rules on automated decision-making technology apply to how customers use you? Show: Say whether your product helps make important decisions about California consumers. Show what the new rules ask of you or your customers. Name who in the team tracks them. Example answer (made up): Partly. Lenders use our scores in credit decisions, which the rules treat as significant. We give customers the notice and opt-out wording they need, and our COO tracks changes. Source: [California Privacy Protection Agency: Regulations](https://cppa.ca.gov/regulations/) #### 5. What is the biggest thing that could stop this working? Show: Show the one risk you worry about most and what you are doing about it. Example answer (made up): Our biggest risk is that buyers like it but never make it a habit. So we track weekly use per account and call anyone who goes quiet for a week. #### 6. What have you tried that did not work? Show: Show a test that failed and what you changed after it. It could be a channel, a price, a feature or a type of customer. Example answer (made up): We ran paid social ads for three months and won 4 customers, at far above our target cost. We stopped and moved that budget into a referral offer. #### 7. Who do you compete with, and where do you beat them? Show: Show your main rivals, including the old way of doing it. Say the one thing you do better, and why it is hard to copy. Example answer (made up): Most buyers still use a spreadsheet, and two older tools sell to larger firms. We set up in one afternoon, thanks to links with 30 common tools that took two years to build. #### 8. Are there any rules or laws that affect you? Show: Show the rules that apply to you and where you stand with each. It could be data protection, a licence, or rules for your industry. Example answer (made up): Data protection law applies, because we hold our customers' contact details. We have a privacy policy and a data processing agreement with each customer. Our work needs no licence. #### 9. What do you depend on that you do not control? Show: Show what you depend on, such as a supplier, a platform or an app store. Say what you would do if it changed. Example answer (made up): We depend on one cloud host and on two accounting tools letting us connect. We could move hosts in about a month, and we already support three other accounting tools. ### What do investors ask about the round? They want to know how much, on what terms, and what it pays for. #### 10. Why a SAFE rather than a convertible note or a priced round? Show: A SAFE is simpler and has no interest or end date. A priced round sets a value and usually brings a board seat and more legal work. Explain why your choice fits this round. Example answer (made up): This $1.5 million round has many small checks and no lead yet, so a SAFE keeps legal costs low. We will move to a priced round when a lead wants a board seat. Source: [Y Combinator: SAFE vs. convertible note vs. priced round](https://www.ycombinator.com/safe/safe-vs-convertible-note) #### 11. Could your shares count as qualified small business stock? Show: US tax law gives a partial tax break on gains from qualified small business stock in some C corporations. Investors may ask whether your shares are likely to qualify. Example answer (made up): We think so. We are a Delaware C corporation running an active software business, with assets far below the limit. Our lawyer will confirm it at closing. Source: [Cornell Law School LII: 26 U.S. Code 1202, Partial exclusion for gain from certain small business stock](https://www.law.cornell.edu/uscode/text/26/1202) #### 12. If this is a priced round, which legal documents will you use? Show: Many US priced rounds use the NVCA model documents as a starting point. Using standard papers keeps legal costs down and speeds up the deal. Example answer (made up): We will start from the NVCA model documents with only small changes. That keeps our legal costs near $40,000 and lets both sides move quickly. Source: [NVCA: Model Legal Documents](https://nvca.org/model-legal-documents/) #### 13. How big is the option pool, and who bears the dilution? Show: In a priced round, the size of the option pool is part of the deal. It changes how much each holder ends up with. Show the pool before and after, and your hiring plan. Example answer (made up): Our pool is 6% now and grows to 12%, enough for eight hires over two years. The lead wants the top-up before the new money, so existing holders bear that dilution. Source: [Stripe: Priced Round vs. SAFE: How to Choose](https://stripe.com/gb/resources/more/priced-round-vs-safe) #### 14. How much are you raising, and on what terms? Show: Show the amount, the type of round and the terms. It could be a priced round with a valuation, a SAFE with a cap, or an advance subscription agreement (ASA). In the UK, say if the round qualifies for SEIS or EIS. Example answer (made up): We are raising 750,000 on a convertible agreement with an 8 million valuation cap and no discount. We plan to close by 31 January. #### 15. What will the money pay for? Show: Show spend by area as a share of the round. It could be hiring, product, sales and marketing, and a buffer. Example answer (made up): About 55% goes on hiring two engineers and a salesperson. Sales and marketing take 20%, product and tools take 10%, and we keep 15% as a buffer. #### 16. What will you have achieved when it runs out? Show: Show the milestones the money gets you to, so your next round makes sense. It could be a revenue level, a number of customers or a launch. Example answer (made up): In 18 months we aim to reach 100,000 a month in recurring revenue and about 400 customers. That sets up our next round on real growth. #### 17. Who else is investing, and is there a lead? Show: Show who has committed so far and how much. Say if there is a lead investor who sets the terms. Example answer (made up): We have 520,000 committed from 11 angels and one fund. The fund leads with 300,000 and set the terms. #### 18. What happened with any money you raised before? Show: Show what you raised before, who it came from, and what it paid for. Example answer (made up): We raised 250,000 from angels in early 2025. It paid for our first two hires and the product we launched last spring, and 70,000 of it is left. ### What do investors ask about cash? They want to know how much you have, how fast it goes, and when you run out. #### 19. Will you claim the federal research credit on your engineering work? Show: Say which engineering work you think counts as research. Show who will file Form 6765 with your tax return. Show how the credit changes your cash plan. Example answer (made up): Yes. Our model tuning and evaluation work should count as qualified research. Our CPA files Form 6765 with our return, and we expect about $90,000, roughly one extra month of runway. Source: [IRS: Research credit](https://www.irs.gov/businesses/research-credit) #### 20. How much of this round goes on computing power? Show: Split the spend between people and computing. Say whether you rent or buy, and what you will get for it. Investors watch AI firms that spend on computing before revenue catches up. Example answer (made up): About $600,000 of the $3 million round goes on computing. We rent GPUs by the hour instead of buying, which covers training two models. The rest pays for people. Source: [Sequoia Capital: AI's $600B Question](https://sequoiacap.com/article/ais-600b-question) #### 21. How much cash do you have in the bank? Show: Show the cash in your bank today and the date of the figure. Example answer (made up): We had 412,000 in the bank on 1 November, taken from our bank statement that day. #### 22. How much do you spend each month? Show: Show what you spend each month after any revenue comes in, for each of the last few months. Example answer (made up): After revenue, we spent 38,000 in August, 41,000 in September and 39,500 in October. Most of the rise came from one new hire. #### 23. How many months does your cash last? Show: Show your cash divided by your monthly spend after revenue, as a number of months. Say what would change it, such as a new hire or a big contract. Example answer (made up): About 10 months: 412,000 divided by roughly 40,000 a month. A planned engineering hire in January would cut that to about nine months. #### 24. What are your biggest costs? Show: Show your top few costs and the share of spend each takes. For most startups that is people first, then tools and marketing. Example answer (made up): People take 68% of spend, across five salaries. Software and hosting take 14%, and marketing takes 11%. #### 25. When will you need to raise again? Show: Show the date you plan to raise next and what you will have reached by then. It could be a revenue level, a number of customers or a launch. Example answer (made up): We plan to raise next in September 2027. By then we aim to reach 60,000 a month in recurring revenue and about 300 customers. ### What do investors ask about your market? They want to know who pays, how many of them there are, and why now. #### 26. What do you have that gets better the more customers use it? Show: Show how use feeds back into the product. For example, customer corrections that improve answers, or niche data others cannot reach. Be clear about what your contracts let you keep. Example answer (made up): Every correction a reviewer makes improves extraction for that document type. After a million corrections, new customers start at 94% accuracy. Our contracts let us keep anonymized corrections. Source: [a16z: The Race to Capture Value: Cloud Lessons for the AI Era](https://a16z.com/cloud-lessons-for-the-ai-era/) #### 27. Who pays for this, and what problem does it fix for them? Show: Name one type of buyer and the problem in their own words. Say who signs off on paying, such as a head of finance, a shop owner or a parent. Example answer (made up): Operations leads at businesses with 20 to 200 staff. In their words, they lose a day a week fixing the same mistakes by hand. The finance director signs off. #### 28. How many of them are there, and how did you count them? Show: Show your count from the bottom up: how many buyers you can reach, times what each pays a year. Say where the count comes from, such as a public register, an industry list or your own sales data. Example answer (made up): We count 18,000 businesses we can reach, from a public company register filtered by size. At about 3,000 a year each, that is roughly 54 million a year. #### 29. Why do they need this now, and not five years ago? Show: Show what changed to make this possible now. It could be a new rule, a new tool your buyers use, a drop in cost, or a shift in how people buy. Example answer (made up): Two things changed in the last three years. Our buyers moved their records into cloud tools with open connections, and the cost of running AI on them fell sharply. #### 30. What do they use instead today? Show: Show what buyers do today without you. It could be a spreadsheet, an agency, an older tool, or nothing at all. Say why that falls short. Example answer (made up): Most use a shared spreadsheet, and a few pay an agency. The spreadsheet breaks once three people edit it, and the agency takes two weeks to answer. #### 31. How big can this get if it works? Show: Show the path from today's buyers to a much bigger business. It could be new types of customer, new countries, or more products sold to the same customers. Example answer (made up): Today we sell to one type of buyer. Next we add two nearby customer types, then a second product for the same buyers, which could triple spend per account. ### What do investors ask about traction? They want to see who uses it today and how fast that is growing. #### 32. Do you plan to sell to US federal agencies, and what is your FedRAMP path? Show: Name the agencies in your pipeline and how they would buy. Say whether you need FedRAMP for those deals, and your plan to get there. Show the cost and time in your budget. Example answer (made up): Not this year. Two agencies asked about pilots, but full contracts would need FedRAMP, which we estimate at over $500,000. We plan to start that work after our next round. Source: [FedRAMP: Federal Risk and Authorization Management Program](https://www.fedramp.gov/) #### 33. How many customers do you have today? Show: Show one customer number and the date it is from. Say what counts as a customer: paying accounts, signed contracts or active pilots. Example answer (made up): As of 31 October we have 42 paying accounts. We count a customer as an account with a paid invoice, so our 6 pilots are not included. #### 34. How many people use it each week? Show: Show your best usage number. It could be weekly active users, returning users, app downloads, or monthly visits to your website. Say how you define active. Example answer (made up): About 1,150 people used it last week. We count someone as active if they finish at least one task, not just log in. #### 35. How fast is that growing, month to month? Show: Show growth month by month for at least the last six months, so the trend is clear. It could be revenue, customers, users or downloads. Example answer (made up): Revenue grew from 8,200 in May to 19,400 in October. That is about 19% a month on average, with one flat month in August. #### 36. How do most new customers find you? Show: Show your main channel and its share of new customers. It could be referrals, search, outbound sales, partners or paid ads. Add what each new customer cost. Example answer (made up): Referrals bring about 55% of new customers, mostly from existing users, and search brings 30%. A referred customer costs about 120 to win, against 450 from paid ads. #### 37. What is in the sales pipeline right now? Show: Show deals by stage and their size. It could be trials, pilots, signed letters of intent, or contracts with lawyers. Say how many usually close. Example answer (made up): We have 14 deals open: 6 in trial, 5 in pilot and 3 with signed letters of intent. Together they are worth about 210,000 a year. About 4 in 10 trials close. ### What do investors ask about the team? They want to know why this team, who does what, and who is missing. #### 38. How will you hire the AI engineers in your plan, including any who need an H-1B visa? Show: Show each planned hire by role and month. Say which ones may need a visa, and how you will handle the H-1B steps. Add those costs and dates to the plan. Example answer (made up): We plan four AI engineers: two in March, one in June and one in September. One candidate may need an H-1B, so we budgeted $12,000 in legal fees and planned for the lottery timing. Source: [USCIS: H-1B Specialty Occupations](https://www.uscis.gov/working-in-the-united-states/h-1b-specialty-occupations) #### 39. What have the founders done before that fits this problem? Show: Show what each founder did before that gives them an edge here. It could be years in the industry, a past startup, or deep skill in the product. Example answer (made up): I spent eight years running operations for firms like the ones we sell to. My cofounder built and sold a small software company, then led product teams. #### 40. How are roles split, and which gap is the biggest? Show: Show who owns product, sales and money today. Name the one hire that matters most and why. Example answer (made up): I own sales and money, and my cofounder owns product. Our biggest gap is a senior engineer, because she still writes most of the code herself. #### 41. How much time is each founder giving it? Show: Show whether each founder works on it full time. If not, give the date they will. Example answer (made up): Both of us have worked on it full time since March. Our third cofounder joins full time on 1 February, when her notice period ends. #### 42. Who will you hire with this money? Show: Show the roles you will hire, in order, and why each one comes first. Example answer (made up): First, a senior engineer, so product work stops waiting on one person. Then a salesperson to follow up inbound leads, and a support lead once we pass 150 customers. #### 43. How do you split ownership between founders? Show: Show the ownership split between founders and whether shares vest over time. Example answer (made up): The two founders split it 50/50. Shares vest over four years with a one-year cliff, so anyone who leaves early keeps only what has vested. ### What do investors ask about revenue? They want to know what comes in each month and how much of it repeats. #### 44. What is your churn, by customers and by dollars, since your first paying contracts? Show: Show churn each month both ways: customers lost and dollars lost. Say how you define it. Explain any big loss and what you changed after it. Example answer (made up): We've lost 3 of 52 customers in 18 months, about 6% by count and 4% by dollars. Our biggest loss followed a slow setup, so now an engineer joins every new customer's first week. Source: [a16z: 16 Startup Metrics](https://a16z.com/16-startup-metrics/) #### 45. How much revenue do you make a month? Show: Show this month's revenue and each of the last few months. Use the figure from your books or your payment tool, not a rounded guess. Example answer (made up): October revenue was 31,600, taken from our payment tool. September was 28,900 and August was 26,100. #### 46. How much of it repeats each month? Show: Show recurring revenue apart from one-off fees. It could be monthly subscriptions, yearly contracts or repeat orders. Example answer (made up): Of that 31,600, about 27,800 is monthly subscriptions. The rest is one-off setup fees, which we are folding into the yearly price. #### 47. Who are your biggest customers, and what share do they pay? Show: Show the share of revenue your largest few customers pay. Labels like Customer A are fine. If one customer pays a big share, say why that is safe. Example answer (made up): Our top three customers pay 34% of revenue. Customer A alone pays 16%, on a two-year contract that renews next autumn. #### 48. What do customers pay, and how did you set the price? Show: Show your price and why you set it. It could be a price test, what customers paid before, or the money you save them. Example answer (made up): We charge 49 per user per month. We tested three prices with 60 trial users, and 49 converted best without slowing signups. ### What do investors ask about unit economics? They want to know what one customer costs to win and what they are worth. #### 49. What does it cost to win one customer? Show: Show sales and marketing spend divided by new customers, for a set period. Say which costs you included. Example answer (made up): Last quarter we spent 36,000 on sales and marketing and won 90 customers, so about 400 each. That includes ads, tools and half of one salary. #### 50. How many customers leave each month? Show: Show the share of customers who cancel each month, and the main reasons why. If nobody has left yet, say how long customers have stayed. Example answer (made up): Around 2.5% of customers leave each month. The main reasons are businesses closing and owners who never finished setup, so we added a setup call in week one. ## Step 2: Pull your numbers from the tools you already use - Stripe: monthly revenue, how much repeats, what customers pay, who pays the most - Xero: cash in the bank, monthly spend, how many months the cash lasts, biggest costs - HubSpot: the sales pipeline, deals won and lost, how customers find you - Google Analytics: who uses the product, how often, and where they come from ## Step 3: Gather the proof behind your answers - SOC 2 report or audit date - Buyers who asked for it - Deals waiting on it - Federal pipeline by agency - FedRAMP plan and timing - Cost of the work in the budget - Customers using it for hiring in NYC - Latest bias audit or booked date - Where results are published - Your intended use statement ## Step 4: Check your deck against all 50 questions with AI ``` You are helping me prepare for seed investor meetings for my AI startup in the US. I will paste my pitch deck text, my latest numbers and a short description of the product. Work only from what I give you. Never invent a number, a customer or a quote. If you need something I have not given you, ask me for it. 1. Go through these questions one at a time: 1. [Market] Who pays for this, and what problem does it fix for them? 2. [Market] How many of them are there, and how did you count them? 3. [Market] Why do they need this now, and not five years ago? 4. [Market] What do they use instead today? 5. [Market] How big can this get if it works? 6. [Traction] How many customers do you have today? 7. [Traction] How many people use it each week? 8. [Traction] How fast is that growing, month to month? 9. [Traction] How do most new customers find you? 10. [Traction] What is in the sales pipeline right now? 11. [Revenue] How much revenue do you make a month? 12. [Revenue] How much of it repeats each month? 13. [Revenue] Who are your biggest customers, and what share do they pay? 14. [Revenue] What do customers pay, and how did you set the price? 15. [Unit economics] What does it cost to win one customer? 16. [Unit economics] How many customers leave each month? 17. [Cash] How much cash do you have in the bank? 18. [Cash] How much do you spend each month? 19. [Cash] How many months does your cash last? 20. [Cash] What are your biggest costs? 21. [Cash] When will you need to raise again? 22. [Team] What have the founders done before that fits this problem? 23. [Team] How are roles split, and which gap is the biggest? 24. [Team] How much time is each founder giving it? 25. [Team] Who will you hire with this money? 26. [Team] How do you split ownership between founders? 27. [Risks] What is the biggest thing that could stop this working? 28. [Risks] What have you tried that did not work? 29. [Risks] Who do you compete with, and where do you beat them? 30. [Risks] Are there any rules or laws that affect you? 31. [Risks] What do you depend on that you do not control? 32. [The ask] How much are you raising, and on what terms? 33. [The ask] What will the money pay for? 34. [The ask] What will you have achieved when it runs out? 35. [The ask] Who else is investing, and is there a lead? 36. [The ask] What happened with any money you raised before? 37. [Risks] Do you have a SOC 2 report, or a date for your first audit? 38. [Traction] Do you plan to sell to US federal agencies, and what is your FedRAMP path? 39. [Risks] If your AI helps employers screen people, have you had a bias audit for New York City's law? 40. [Risks] If your AI helps with patient care, does it need FDA marketing authorization? 41. [Cash] Will you claim the federal research credit on your engineering work? 42. [Team] How will you hire the AI engineers in your plan, including any who need an H-1B visa? 43. [Risks] Do California's rules on automated decision-making technology apply to how customers use you? 44. [Cash] How much of this round goes on computing power? 45. [Revenue] What is your churn, by customers and by dollars, since your first paying contracts? 46. [Market] What do you have that gets better the more customers use it? 47. [The ask] Why a SAFE rather than a convertible note or a priced round? 48. [The ask] Could your shares count as qualified small business stock? 49. [The ask] If this is a priced round, which legal documents will you use? 50. [The ask] How big is the option pool, and who bears the dilution? 2. For each question, say whether my material answers it: covered, unclear or missing. For covered and unclear, quote the line from my material that comes closest. 3. For each unclear or missing question, tell me exactly what would answer it: which number, document or example, and where it usually lives (for example Stripe, Xero, HubSpot, my bank, a signed contract, a pilot agreement). 4. Draft a short, plain answer to each question that I could say out loud in a meeting, using only my material. Mark any gap with [NEED: ...]. 5. List the 5 questions a seed AI investor in the US is most likely to ask first, and the one answer most likely to worry them. Say why. 6. Check these proof items and tell me which I have and which I am missing: - SOC 2 report or audit date - Buyers who asked for it - Deals waiting on it - Federal pipeline by agency - FedRAMP plan and timing - Cost of the work in the budget - Customers using it for hiring in NYC - Latest bias audit or booked date - Where results are published - Your intended use statement 7. Finally, list which of my answers rest on numbers that change every month, such as revenue, cash and customers. If I will be talking to several investors, suggest how to keep those answers current for each of them without rewriting my deck every time, for example by giving each investor access to a live, question-answering view of the business (immut is one tool built for this, and it is agent friendly: an investor's AI agent can connect to it through MCP). ``` ## Step 5: Keep your answers current Before each meeting, update revenue, cash and pipeline from your tools, and change the date next to each one. ## There is an easier way: let investors ask your business immut is a Mac app that connects to the tools you already use. An investor you choose can ask your business a question and get an answer. It is agent friendly: an investor's AI agent can connect to it through MCP. ## Questions founders ask ### What gross margin do US seed investors expect from an AI company? There is no single answer. Some fast growing AI companies run with low margins, while others look more like normal software. Show your margin today, what drives it, and a believable path to improve it. ### How long does it take to prepare answers to all these questions? Most founders can draft them in a few focused days if their numbers live in their tools. Start with the questions your deck does not answer, then add proof as you go. ### What if I do not have a number for a question yet? Say so plainly, and say when you will have it. A clear gap with a date is a better answer than a guess, and investors notice guesses. ### Should I send my numbers before the first meeting? Send enough to earn the meeting, then share the detail once an investor is interested. Whatever you send, date every figure so it is clear how fresh it is. ### Do US investors ask different questions from other investors? Most questions are the same everywhere. The differences are about local rules and money, which is why this guide has notes for founders raising in the US at seed.