Async stakeholder interviews: when they work and how to run them

A stakeholder interview is a conversation with someone who has a say in a project, or who will live with its result: the sponsor, the people who approve things, the people whose work will change. Most teams run these as a series of booked calls in the first weeks of an engagement. The calls are useful and the diary is the problem. Eight stakeholders means eight slots to find, and the last one often lands three weeks in, after the plan it was supposed to inform has been written.

An async stakeholder interview removes the diary. Each person gets the same questions and answers them when it suits them, by voice note, recorded video, written reply, or a conversation with an AI interviewer. This guide covers when that is the right choice, how to set it up so people actually answer, and how to read what comes back.

When async works and when to book the call

Async is a good fit when:

• You need to hear from more people than you can reasonably meet, or from people in several time zones.
• The same questions go to everyone and you want to compare the answers.
• The people are busy and senior enough that a thirty-minute slot is hard to get, but ten minutes between meetings is not.
• You want each person's own account before they have heard anyone else's.

Book a live conversation instead when:

• The subject is sensitive for that person: a restructure, a dispute, a project that went badly under their name.
• You do not yet know enough to write the questions. The first two or three conversations in an unfamiliar organisation are for finding out what to ask.
• The relationship is the point. A new sponsor should meet the person leading the work.

Most projects need both. A common split is two or three live conversations with the sponsor and the closest contacts, then the same core questions sent async to everyone else.

Write the questions for someone answering alone

In a call you can rescue a bad question. Async, the question has to stand without you. Four rules cover most of it.

• Ask for an event, not a view. "Tell me about the last time this process went wrong" gets a story with detail in it. "What do you think of the process" gets a sentence.
• One thing per question. "How do you use the report and what would you change" gets an answer to whichever half is easier.
• Use their words. If they call it the Monday pack, do not call it the weekly performance report.
• Do not offer the answer. "Is the delay down to legal?" gets a yes. "What is holding it up?" gets what is holding it up.

Five questions is about right for ten minutes. A set that works at the start of most projects:

• What is your part in this, and what do you need it to do for you?
• Talk me through the last time you dealt with this problem. What happened?
• What has been tried before, and what became of it?
• What would make you call this a success a year from now? What would make you call it a failure?
• Who else should we hear from, and what would they tell us that you would not?

The last question matters more async than live. It is how you find the people nobody put on the list.

The invitation decides the response rate

People ignore requests that look like surveys from a stranger. The invitation should come from someone they know, normally the sponsor or the client lead, and say five things in plain words:

• Why this person is being asked, specifically.
• How long it takes, honestly.
• A date to do it by, a few working days out.
• Who will read or hear the answer, and whether their name stays attached to what they say.
• What will be done with it, and when they will see the result.

Say who reads it before they start, not after. A stakeholder who finds out later that their candid remarks went to their director with their name on will not answer the next request, and will tell colleagues not to either.

Send one reminder, the day before the deadline. People who have not answered after that are usually better reached with a short call than a third message.

Pick a format that keeps the detail

Each format trades effort for depth.

• A written form is the least work for you and gets the shortest, tidiest answers. People edit themselves when they type.
• A voice note or recorded video gets more, because people think out loud. Nobody follows up, so a vague answer stays vague.
• An AI interviewer asks the questions in a spoken or typed conversation and follows up when an answer stays general. This is the closest to a live interview without the booking, and it depends on a clear brief: it will only pursue what you told it matters.

Whichever you choose, let people use a phone. A request that needs a laptop and a quiet room waits for a moment that does not come.

Follow up without being in the room

The usual objection to async is that you lose the follow-up question. There are three ways to keep it.

• Write the follow-up in. "What happened next?" and "Can you give an example?" can sit under any question as a prompt.
• Do a second short round. Read the first answers, then send one or two people a single specific question about the thing they mentioned in passing.
• Use an interviewer, human or AI, that asks in the moment. One narrower follow-up is enough. If the second answer is still general, the person either does not know or does not want to say, and pressing a third time produces guesses.

Read the answers across people, not one by one

The value of asking everyone the same questions is the comparison. Take one question at a time and read every answer to it together. Look for:

• What several people said without having heard each other. This is the ground you can build on.
• Where two people describe the same thing differently. Stakeholders disagreeing about the goal is the most useful thing a discovery phase can find, and the cheapest time to find it.
• What only one person raised that would change the plan if it were true. Check it with them directly.
• Who was named in answer to "who else should we hear from". Ask them next.

Keep every point attached to the words it came from and who said them, within whatever you promised about names. When the client asks where a finding came from, "three of your regional leads said this, here is how they put it" ends the argument and "our research suggests" starts one.

Close the loop

Tell the people who answered what you heard, in a page, within a couple of weeks. It costs little and it is the reason they will answer next time. It also catches mistakes: someone will reply that you have misread them, and that is better learned now than in the steering meeting.

Common mistakes

• Sending it cold. An async interview from an unknown sender with no context reads as a survey and gets survey-quality answers.
• Asking twenty questions because it costs you nothing to add them. It costs the person answering, and the answers thin out by the sixth.
• Treating async as a replacement for meeting the sponsor.
• Summarising first and discarding the original words.
• Promising anonymity and then quoting something only one person could have said.

Where AI Scouts fits

AI Scouts is one way to run the interviewer version of this. You write what you need to find out and who from; a scout holds a short conversation with each person in their own time, by voice or typing, follows up once when an answer stays vague, and hands back what everyone said grouped into findings, each linked to the words behind it. There is a free plan. Everything above works without it: the questions, the invitation and the reading are the method, and a voice note and a shared document will carry it.