For a lean fund, the first call is the expensive step. Each one takes thirty minutes or more, plus scheduling across time zones, and most of it is spent on questions the team asks every founder. Meanwhile good applications wait in the queue, and founders wait to hear anything at all.
An asynchronous step moves the routine part of that call out of the team's calendar. The founder takes the interview when it suits them; the team reviews it when it suits them; live partner time goes to the founders worth a deeper conversation. Done badly, it feels like a hoop and puts off the founders you most want. What follows: the formats an asynchronous interview for startups can take, when each fits, how to design one, how to assess it fairly and how to respect founders' time and data.
What an asynchronous founder interview is
In an asynchronous interview, the fund sets the questions in advance and the founder takes the interview within a window, usually a few days. Nobody from the fund is present. The team reviews the result later, alongside the deck and application.
It sits between the written application and the first meeting. In the Gompers, Gornall, Kaplan and Strebulaev survey of 885 venture investors, firms met the management of about 28 companies for every deal they closed, out of roughly 101 considered. An asynchronous step gives a fund a way to hear from more founders than it can meet, and to make the first meeting count for those it does.
"Asynchronous" describes the fund's side. For the founder, the interview can still be live: a newer format puts an AI interviewer on the call, so the founder pitches and answers out loud in real time, while the fund's team watches the recording later.
Four common formats
- Recorded one-way answers: each question appears on screen and the founder records a short video or voice answer, one clip per question.
- Written questions: the founder types answers within a word limit. Nothing is recorded, and it suits founders who are more precise in writing.
- A live AI interview: one recorded call where the founder pitches, often sharing their screen to show the deck, and an AI interviewer then asks the fund's questions and follow-ups out loud within a fixed time.
- A text interview: the same questions as the AI interview, in a chat, with the same rubric. It is the fallback for anyone who declines recording or lacks a quiet room or a good connection.
When an asynchronous interview for startups works
It is a tool for a specific bottleneck, not a replacement for every conversation. Use it where the main problem is volume and consistency, and switch to a live call with a partner where the relationship matters more than the comparison.
| Situation | Recorded or written answers | Live AI interview | Live call with a partner |
|---|---|---|---|
| Many inbound applications, small team | Good fit: no calls to book | Good fit: no calls to book, and the founder gets a real back-and-forth | Hard to scale; the queue grows |
| Founders across many time zones | Good fit | Good fit: the founder starts whenever suits them | Scheduling can take longer than the call |
| Hearing the pitch itself | Weak: answers are cut into separate clips | Strong: the founder pitches with their own deck on screen | Strong, but only for the founders who get a slot |
| Follow-up questions on this founder's claims | Limited to what was written in advance | Yes: follow-ups drawn from the pitch, the deck and open claims | Yes, if the partner has read the deck |
| A warm referral or a competitive round | Poor fit: can feel like a hoop | Poor fit for the same reason | Better: the founder is choosing you too |
| Founder on a phone or a weak connection | Works for written answers | Needs a desktop browser for screen sharing; offer the text version | Works by phone |
A sensible default for many funds: an asynchronous step for inbound applications that pass a first thesis check, and a direct call for warm introductions and deals where speed matters. The use case on first-round founder screening walks through that flow.
How to design the interview
Design it as if you were the founder receiving it. Every question should be one you would otherwise ask on a call, and the whole thing should take less time than the call it replaces.
- 1State the purpose, the total length and its parts, who will see the recording and how long it is kept, before anything starts.
- 2Ask for consent to each recording separately, for example camera, microphone and screen, and offer a text interview to anyone who declines or prefers it.
- 3If founders share their screen, warn them that notifications and other windows can show, and suggest sharing one window or tab.
- 4Choose four to six fixed questions covering the customer, the team, traction, the main risk and what the round is for. Our questions to ask founders guide explains what each one tests.
- 5Add a small, capped number of follow-ups from the deck or the pitch: an undefined metric, a claim that needs context or a missing fact your thesis needs.
- 6Fix the length and keep it the same for every founder on the template. For a live AI interview, a pitch of about 5 minutes and a total cap of 15 to 20 minutes leaves room for the questions.
- 7Let founders rejoin after a dropped connection, set an expiry with a reminder, and offer a manual alternative, such as a short call, for anyone who cannot use the format.
Write the questions plainly and avoid jargon. Founders may be answering in their second or third language, and a clear question is fairer to everyone. An AI interviewer should follow fixed rules: one question at a time, a neutral tone, no advice, no comments on delivery and no questions about protected characteristics.
Assess answers on substance, not delivery
Recorded interviews make it tempting to judge performance: confidence, fluency, slick slides, a polished background. None of that is evidence about the business. Assess what was said, and the evidence behind it, against criteria you wrote before you watched anything.
Hiring research supports the structure. Google's re:Work guide describes structured interviewing as uniform methods for assessing candidates for the same role, with vetted questions, detailed notes, standardised rubrics and interviewer calibration, and reports increased predictive validity and decreased differences between demographic groups. The type of question matters too: Kanze, Huang, Conley and Higgins found that at TechCrunch Disrupt New York City from 2010 to 2016 investors tended to ask men about gains and women about risks, and those asked gain-focused questions raised significantly more. Fixed questions with the same balance for everyone address that directly.
- Include: understanding of the customer, quality of reasoning, evidence for claims, what the team has learned, clarity about risk.
- Exclude: accent, fluency, presentation polish, appearance and background, including the room and the equipment.
- If AI drafts an assessment, give it the transcript only, never the video or the screen images, and make every rating cite the moment in the transcript it relies on.
- Have a person decide who advances, and record a reason when they depart from the rubric.
Check that format doesn't change the result: the same content given live, by voice or in text should land within a small tolerance of each other. For founders in the EU or UK, Article 22 of the GDPR gives people the right not to be subject to decisions based solely on automated processing that significantly affect them, so a person should make every advance or pass decision. Our guide on bias in startup screening covers the wider process.
Respect founders' time, consent and data
Founders talk to each other. A fair, short interview with a quick decision builds your reputation; a long one followed by silence damages it. Treat the interview as part of what you offer founders, not only a filter.
- Tell founders when they will hear back, and keep to it.
- Let founders read their transcript and flag transcription errors, and keep the recording as the record.
- Set a retention period for each recording, separately from transcripts, and delete them when it ends. Withdrawing consent should delete the recordings.
- Never share recordings outside the team that needs them.
- If you pass, say so clearly. A short reason is a kindness.
Common mistakes
- Sending the interview to warm referrals who expected a conversation.
- Asking twelve questions when five would do, or letting the call run past the length you promised.
- Making camera or screen sharing mandatory with no text option, which excludes founders with poor bandwidth, a phone only or a wish not to be recorded.
- Rating answers after discussing them with colleagues, so the first opinion sets the rest.
- Using the step to delay a decision the team could already make.
Where Delsivo fits
Delsivo is an AI investment workspace for venture funds, at the design partner stage; nothing is generally available yet. Interviews are planned for the first release, not the Pilot.
In the first release, Delsivo plans a live AI founder interview over WebRTC that the founder starts from your portal whenever suits them. First the founder shares their screen with the camera on and pitches for about 5 minutes while the AI listens. Then an AI interviewer asks your core questions out loud, followed by follow-ups drawn from the pitch, the deck and open claims. The call is capped at a length your fund sets, 15 to 20 minutes, and every founder on a template gets the same core questions and time plan.
Camera, microphone and screen are recorded, each with its own consent given before the call, and a text interview is always available with the same questions and rubric. Screen sharing needs a desktop browser; on a phone the founder can take the text interview. Founders can rejoin after a dropped connection and flag transcription errors.
The AI assesses only the transcript, on substance and evidence, with timestamped citations. Accent, fluency, presentation polish, appearance, background and the video and screen images are excluded; your team watches the recording. Analysts decide who advances and can override any assessment with a reason. The founder interviews feature page has more detail.
Sources
- Google re:Work: A guide to structured interviewing for better hiring practices (updated March 2026) (checked 6 November 2026)
- Kanze, Huang, Conley and Higgins: We Ask Men to Win and Women Not to Lose (Academy of Management Journal, 2018), London Business School research record (checked 6 November 2026)
- Gompers, Gornall, Kaplan and Strebulaev: How Do Venture Capitalists Make Decisions? (NBER Working Paper 22587, September 2016) (checked 6 November 2026)
- GDPR, Article 22: Automated individual decision-making (checked 6 November 2026)
Delsivo publishes this guide and makes software for venture funds, so we have a stake in the topic. Facts about other companies, research and regulations come from the sources listed, on the dates shown, and can change. Corrections are welcome through our contact page.