Productside Webinar

Ask, Listen, Learn

Running Better Empathy Interviews with AI

Date:

08/18/2026

Time EST:

1:00 pm
Watch Now

Your best insights are hiding in a bad interview. 

Empathy interviews are one of the highest-value, easiest-to-waste skills in product management: hard to start, hard to conduct well, and hard to turn into insight. This workshop shows PMs how to use AI across all three: sketching a loose topic guide, rehearsing the interview through role-play so customers carry the story, and turning transcripts into product insight. You’ll leave with a repeatable way to practice the one skill AI can’t do for you: listening. 

Most PMs either wing their interviews or over-script them into a stiff interrogation. Both approaches kill the insight you came for. This workshop fixes that. You’ll walk away with a rehearsal method you can use before every interview, not just this one. 

What You’ll Learn:  

  • What is the value of empathy interviews, and how to get started.
  • Why an empathy interview needs a loose structure, not a tight script, and how over-preparing can backfire.
  • How to use AI as a practice partner: role-playing a persona and pressure-testing your ability to stay open-ended and follow the “why.”
  • How to use AI to turn a mess of interview notes and transcripts into insight your product team can act on.

Welcome and Attendee Roll Call

Kenny Kranseler & Ryan Cantwell | 00:00:00 – 00:02:33

Happy Tuesday everybody and welcome to the Productside webinar: Ask, Listen, and Learn. How to run better empathy interviews with AI as your helper. We’ll get started in just a couple of minutes to give everybody a chance to find their way into our Zoom room.

In the meantime, if you can drop in the chat where you’re dialing in from, we’d love to see who’s here and where they’re coming from.

Ryan’s from the happy city of Pittsburgh. That’s right. This is a nice time of year in Pittsburgh. Don’t come in January, Kenny. I won’t.

We’ve got two kinds of Londons going on. Sheffield. Hey, welcome. Kat, I recognize you. I remember we had class together a while ago. And we have Canada represented, and someone from Egypt. Welcome, Sah. Someone from Cape Town. Welcome, Nicole.

Speaker Introductions

Kenny Kranseler & Ryan Cantwell | 00:02:33 – 00:03:56

All right, let’s kick things off. I want to be the first to welcome you to our Productside webinar on how to run better interviews with your customers using AI. And I’m not going to tell you to go use AI to do interviews with your customers. If you’re dialing in for that, you’re not going to get that answer. But I will help you leverage AI to do better discovery interviews, to truly understand your customers. It’s all about asking, listening, and learning from them.

I am Kenny Kranseler, a principal consultant and trainer at Productside for about the last seven years. Before that, I spent time doing product management at places like Kentucky Fried Chicken, Microsoft, Amazon, and a whole host of startups in the Seattle area.

And I am Ryan Cantwell. I have been with Productside about five years. Before that, I spent time doing B2B portfolios across hardware, software, and services. I have probably seen something that’s complex that combines all three. Excited to share those experiences with you today.

About Productside and Logistics

Ryan Cantwell | 00:03:56 – 00:05:15

If you are familiar with Productside, welcome back. If you’re not, a quick word on who we are. Productside is a group of very passionate product people who love helping product teams all over the world build products that customers want to buy and use. The way we do that has grown a lot:

  • Assessments: where we see where you really stand
  • Training: what builds the muscle
  • Advisory: where we sit beside you and your team to help make the hard calls

All three areas have one goal: achieving the business outcomes you can point to that move your organization and yourselves forward. Most of all, know that Productside is your partner in product. We are here to be that friendly companion you can always trust and find answers from.

Webinar Engagement and Resources

Ryan Cantwell & Kenny Kranseler | 00:05:15 – 00:07:36

A few logistics before we dive in:

  • You will get the recording link in your inbox after we wrap up today, so feel free to share it
  • Drop questions in the Q&A as they come to you. Don’t sit on them until the end. We like addressing things as we go
  • Kenny will be demoing live throughout the session, so you’ll see hands-on what good looks like
  • Use the chat to interact with each other as well. Compare thoughts, swap stories, share something you’ve tried recently
  • We’d love for you to connect with us on LinkedIn. Scan the QR code on screen or use the link in chat. It’s where we put leadership thinking worth stealing, AI tips, and the ongoing arguments about what makes product management work

The best sessions are the ones where you push back on us in real time and we get to respond. So don’t hold back. This works best when we have a conversation.

Workshop Agenda Overview and the Value of Empathy Interviews

Kenny Kranseler | 00:07:36 – 00:09:22

Here’s what we’re going to walk through today. The question I’m going to start with is: in a world where you can get information from customers in so many different ways, why do empathy interviews still matter? And once I convince you that they do, we’ll get into the live demo.

Today’s agenda:

  • Why empathy interviews still matter in today’s world
  • A live demo: I’ll actually conduct an empathy interview using AI, in real time, so you can see the process unfold
  • How to get started: using AI to build a loose topic guide
  • How AI helps you conduct and improve interviews
  • How to analyze interviews and extract insight using AI
  • Building high-quality AI habits in general
  • Q&A

Before we dive in, we’d love to get some input from you. How many empathy interviews have you run personally, or close to personally, in the last quarter? Drop your answer in the chat. We’ll give you a few seconds.

Poll: Current State of User Discovery

Kenny Kranseler | 00:09:22 – 00:10:57

All right. Looks like a lot of people are not doing a lot of interviews yet. About 40% of you have done some, which is good. A couple of you have done extensive interviews in the last three months. But a bunch of you have not. If you haven’t talked to a customer in the last quarter, you should change that behavior. We’ll give you some ideas on how to do them better when you do get started.

Poll results:

  • About 40% have conducted at least one interview in the last quarter
  • A small group are doing extensive interviews regularly
  • The majority have not conducted any customer interviews in the last three months

Moving Beyond Feature Requests

Kenny Kranseler | 00:10:57 – 00:13:28

Why do interviews still matter? A lot of it comes back to a quote Henry Ford made about 120 years ago: “If I had asked people what they wanted, they would have asked for a faster horse.” We wouldn’t have cars if people just defaulted to asking what customers want.

The core problem: most product managers default to asking “what do you want?” All standard input methods, when they ask that question, tend to surface feature requests. They do not help you understand needs.

Getting your information secondhand puts an automatic filter on the feedback:

  • Customer success and support teams compress and interpret what they hear before it reaches you
  • You miss the functional, emotional, and social jobs your customers are really trying to get done
  • Functional: what task are they trying to accomplish?
  • Emotional: how do they want to feel as a result, confident, in control, seen as capable?
  • Social: how do they want to be perceived by others when they do this job well?

None of this surfaces if you don’t know how to actually draw those elements out in a conversation. And you won’t get there by asking for a feature list.

The Pitfalls of Synthetic and Scripted Interviews

Kenny Kranseler | 00:13:28 – 00:15:07

There are two traps that product managers fall into, and both of them kill the insight you came for.

Trap 1: Handing everything over to AI. Run synthetic interviews. Have AI play the customer. Use AI-generated personas and never talk to a real person. This is fast, cheap, and always available. It is also seductive. But you do not get the stories. You do not get the understanding. You cannot be confident that the people you’re effectively talking to are the right people, or that their responses reflect real lived experience.

Trap 2: Over-scripting the interview. Writing a detailed, fixed question list and asking the same questions to everyone in the same order. This approach shuts down the customer’s story. You do not get the emotion. You do not get the unexpected detail that changes everything. You miss the examples that only come up when someone is actually talking, not answering a questionnaire.

What you actually want is a topic guide, not a script. Enable customers to tell you the stories around how they use your product, where they use it, when they use it, what challenges they face, and what they’re really trying to accomplish. Let the conversation go where it needs to go.

Setting the Stage: Topic Guides and Persona Setup

Kenny Kranseler | 00:15:07 – 00:18:21

For the rest of this session, don’t think of me as a Productside consultant. Think of me as a product manager at a company I’ll call Meridian CRM. I need to conduct interviews with my customers.

I’ve created a persona named Priya. Priya is a sales operations manager who owns the weekly pipeline forecast across 40 sales representatives.

  • Functionally: she needs to compile an accurate forecast from all the input she gets from those reps
  • Emotionally: she wants to walk into the VP’s office feeling confident, not hedging
  • Socially: she wants to be seen as the source of truth on revenue data

Now, how do you use AI to build a loose topic guide before the interview? Here’s the prompt I’m using in Claude:

I gave it the persona description and asked it to suggest four to five loose topic areas to cover in an empathy interview. Not specific questions. Areas to explore.

Claude came back with:

  • How does Priya get the forecast built today?
  • What does trustworthy data mean to her?
  • Moments where the numbers go wrong
  • What happens when leadership pushes back?
  • What does she wish she didn’t have to double-check?

By throwing in the persona and a simple prompt, I’ve got a solid topic guide. Now I’ve got my guide, my persona, my interviewee lined up. Let’s talk about how I actually prepare to conduct the interview.

Don’t forget: all the prompts we’re using today will be shared with everyone who attended. You do not need to write them down. Be present. You’ll have them in your inbox after this session.

Interviewing Techniques and Ethics

Kenny Kranseler & Ryan Cantwell | 00:18:21 – 00:22:21

Before jumping into the live interview, a few core principles:

Ask open-ended questions. Closed questions get yes or no. You won’t get the color, the understanding, or the context of what happens when a customer faces a challenge.

  • “Why” is your best friend in empathy interviews
  • “Tell me more” is useful when you want to fill a silence without leading
  • Never propose solutions during an interview. You’re here to understand problems, not to test your ideas

Embrace silence. If there’s a pause, let it happen. Customers will fill it, and what they say next is often the most valuable thing in the conversation. My favorite interview trick: say nothing, and they will fill the void. People have a natural propensity to fill silence in discussions. Let it happen.

On interview format: one-to-one vs. group interviews. Both can work, but with caveats:

  • Multiple customers together: you get different perspectives and they can build on each other’s thoughts, but watch for groupthink. One dominant voice can consolidate everyone else’s opinion
  • User and buyer together: be careful if the buyer is the user’s boss. The user may not tell you everything in front of their manager. Ask yourself what you need to learn and structure accordingly
  • Multiple people from your company: this is often a good idea. You get multiple perspectives from your own team, and different team members will hear different things based on their own biases. That’s valuable. And with AI transcription, you no longer need a second person just to take notes

On AI in note-taking: AI can now handle transcription and synthesis, so your second attendee from your company can focus entirely on listening and observation rather than keeping a written record. That’s a meaningful shift.

All right. Let’s do the live demo. I’m going to conduct a real empathy interview with Claude.

Live Demo: Persona Roleplay with Claude AI (Priya Interview)

Kenny Kranseler | 00:22:21 – 00:31:27

Here’s what I’ve set up. I took the Priya persona, put her core needs and context into a markdown file, and attached it to my Claude window. Then I used a roleplay prompt that tells Claude to act as Priya: be fully in character, answer only what I ask, do not volunteer extra information, stay in character throughout, and only break character when I say stop.

Here is the interview, live:

Kenny: Hi, Priya. I’m trying to get user feedback on my CRM product.

Priya (Claude): Oh, hey. Yeah, sure. Meridian, right? We’ve been on it, coming up on three years now, I think. What do you want to know?

Kenny: Tell me what a typical Monday looks like for you as you’re using our CRM application.

Priya: Mondays. Okay, so I’m usually in by around 7:30 because the forecast call with Greg, that’s our VP of sales, is at 10:00 and I need to have the roll-up done before that. So I pull the pipeline report, and that’s when the fun starts. Half my reps update their deals Friday afternoon in this big batch. So I’m never totally sure if what I’m looking at Monday morning is actually current or just Friday’s version of current, if that makes sense. I usually end up pinging a couple people directly: hey, is this still right? And half the time it’s not.

Kenny: That sounds tough. What do you do after your meeting with Greg to make the forecast more accurate over time?

Priya: Honestly, not as much as I probably should. In theory, I’d love to be systematically going back through and tightening up data quality across the board. But in practice, Monday afternoon rolls around and it’s just whichever three deals blew up in the meeting. Like, why did this one move again? And I’m chasing those specific fires instead.

Kenny: So how do you actually do some of that firefighting?

Priya: Say Greg flags the Marquez deal in the meeting. I’ll go straight to the rep, usually Slack: what’s the real status here? Sometimes I’ll pull up the account in Meridian and just start reading through whatever notes exist. Or I’ll dig through my own inbox if I remember being copied on anything. If it’s a big enough deal, I might even sit in on the next call with the rep just to hear where things actually stand.

She mentioned offhand that she keeps a personal spreadsheet to track the deals she doesn’t fully trust.

Kenny: Tell me about that spreadsheet.

Priya: Yeah, I know. It’s embarrassing. It’s just a Google Sheet. I’ve got maybe 20 to 30 deals in there at any given time, the ones where I don’t fully trust the data in Meridian. I check them manually before every forecast call. I’ve tried a couple of times to explain to the reps why it matters to update their notes in real time, but you know how that goes. Reminders don’t really change the culture.

She paused, then added: Honestly, just automatic pings to reps if they haven’t touched a deal in a while would probably fix most of it. Though, actually saying that out loud, I don’t know if that’s really it. We already nag them plenty and it doesn’t change what happens Friday afternoon.

Kenny: Thanks, Priya. I appreciate your time.

Priya: No problem. Good luck with Greg’s meeting. Or mine, I guess. Take care.

Out of character: that’s a wrap on the roleplay. Ready to debrief whenever you are.

Human Critique of the Interview

Ryan Cantwell & Kenny Kranseler | 00:31:27 – 00:33:35

All right. The audience thought Kenny did a good job. Ryan, what did Kenny do wrong?

Ryan Cantwell: A few things I noticed as I was watching.

First: early on, Kenny asked “What are you not getting with our CRM system?” A stronger follow-up to that would have been: what’s the consequence of not getting that? Push deeper on impact, not just the gap.

Second: the spreadsheet moment was excellent. Kenny spotted the signal, named it, and dug in. Tell me about the spreadsheet. That’s exactly right. Manual processes came up several times, and Kenny homed in on that anchor. Strong technique.

Third: Priya mentioned “manual detective work” at one point. I’d have pushed harder there. When you’re doing that detective work, where do you go? How many steps does it take? What time is invested? Get the specifics.

Fourth: at the very end, Priya said something along the lines of “if you could fix this by Monday, it’d be a lot less painful.” I would have asked one of my favorite interview questions right there: on a scale of 1 to 10, how painful is it today? Quantify it. Get a number. That number becomes evidence you can carry into a business case.

Using AI for Self-Critique

Kenny Kranseler | 00:33:35 – 00:36:33

Let’s also see what Claude has to say. Here’s my self-critique prompt: I’ve got a transcript. Be direct. Give me one thing I did well, not generic. Give me what I should improve and where I fell short.

Claude’s feedback:

What Kenny did well: after Priya mentioned the spreadsheet and showed visible discomfort about admitting to it, the natural wrong move is to reassure her, which would have killed the thread. Instead, Kenny went straight into a specific, concrete, non-judgmental follow-up. Strong technique.

What Kenny could have improved:

  • Used some leading questions, which can pull customers toward the interviewer’s assumptions instead of their own experience
  • Cut off some stories before they were finished. Let the customer complete their thought before moving on
  • After Priya volunteered the spreadsheet and was in a detail-rich, specific moment, Kenny pivoted to asking what features could go into Meridian. That’s a move into solution space at exactly the wrong time. The better move: go deeper into what she’s experiencing, not toward what she wants built

So according to both Ryan and Claude: the spreadsheet anchor was the highlight. The leading questions and solution-space pivot were the things to fix.

Body Language and Active Listening

Kenny Kranseler | 00:36:33 – 00:37:50

A few things that stay 100% human, regardless of how good AI gets:

Active listening. You are talking to another person. Summarize what you heard back to them. Make them feel that you actually caught what they said. That allows them to go deeper.

Reading body language. The virtual roleplay with Priya cannot show you what a real interviewee looks like when they’re uncomfortable. In a real interview, seeing that physical discomfort is data. What someone does can be completely different from what they say, and you will only catch that if you are watching.

Sitting in silence. In the virtual demo, silence does not work the same way. In a real conversation, silence is powerful. If someone is sitting quietly after a question, that pause is information. Do not rush to fill it.

The one skill AI cannot do for you is listening. You must be present. You must be in the room, paying attention, watching, and following what is happening in front of you.

Handling Difficult or Complaining Customers

Kenny Kranseler & Ryan Cantwell | 00:37:50 – 00:39:37

A question came in from someone who went into an empathy interview knowing the customer would be challenging, and ended up locked in a room with someone who did nothing but complain for an hour. How do you handle that?

Kenny Kranseler: Sometimes, letting someone vent is actually valuable. Do not shut it down immediately. Let the moaning happen, and use it as an opportunity. Ask them to tell a story about where that challenge actually occurred. Let them walk you through a specific moment. You may be hearing a gripe on the surface, but inside that gripe is real insight about where your product is failing them.

Ryan Cantwell: Be empathetic. You do not want to damage the relationship. Let them vent. But if the conversation drifts entirely off topic and you genuinely cannot get it back, it is okay to cut bait. I once interviewed someone who was more interested in telling me about their weekend with their grandkids than answering my questions. In that case:

  • Acknowledge what they shared warmly and genuinely
  • Then redirect: “Thank you for that. Let’s get back to what I wanted to ask you about.”
  • If they keep going, you have permission to end the interview. Not every conversation will be salvageable, and wasting an hour hurts everyone

Poll: Common Struggles in Empathy Interviews

Kenny Kranseler & Ryan Cantwell | 00:39:37 – 00:42:13

Time for our second poll. Given what you just saw in the live demo, what is your biggest struggle in empathy interviews?

  • Staying open-ended
  • Turning interviews into insight
  • Following the “why”
  • Sitting in silence
  • Avoiding leading questions

Results: avoiding leading questions came out on top as the biggest challenge.

That is consistent with what we see. We all have a perspective. Sometimes it is really hard not to share it. The risk: instead of going deeper on what someone is saying, you start asking “what if we tried this?” or “would this solve it?” Do not do solutioning during empathy interviews. Use what you hear as fodder for brainstorming later, not in the moment.

A trick that can help: when you introduce yourself at the start of an interview, tell the customer upfront: “I’m going to ask you some questions that you might feel I should already know the answer to. I want you to know I’m here to learn from you.” That gives you the card to play dumb and removes the pressure to perform expertise during the conversation.

Synthesizing Jobs to Be Done with AI

Kenny Kranseler | 00:42:13 – 00:45:35

Once you have done your interviews, and I would encourage you to do more than one, AI can help you make sense of what you collected. Here is what you want to pull out of a good analysis:

  • Persona validation: is the person you interviewed actually the right persona for the role you are building for?
  • Jobs to be done: what are they trying to accomplish when they use your product?
  • Desired outcomes: what is the end state they are trying to reach?
  • Pains: what blocks them from reaching that outcome?
  • Gains: what value do they get when they succeed?
  • Says vs. does: where do what they say and what they actually do differ?

Here is the single-interview transcript analysis prompt I’m running in Claude now. I’m asking it to analyze the Priya interview and return: a summary of her needs, her functional, emotional, and social jobs to be done, her key pains and gains, and quotes that illustrate what she says versus what she actually does.

Claude’s output from the Priya interview gave us:

  • Needs summary: accurate, real-time pipeline data she can trust before a high-stakes leadership call
  • Functional job: compile a reliable forecast roll-up without manual verification across reps and channels
  • Emotional job: walk into Greg’s meeting feeling confident, not hedging on numbers she isn’t sure about
  • Social job: be seen as the source of truth on revenue data across the organization
  • Pain: batch updates on Friday afternoon mean Monday’s data is unreliable, forcing her to manually chase accuracy through Slack, email, and a personal spreadsheet
  • Gain: if the data were trustworthy in real time, she could eliminate the spreadsheet and the manual detective work entirely
  • Says vs. does: she says automated reminders would probably fix the problem; she then immediately contradicts herself, noting they already send plenty of reminders and it doesn’t change behavior. What she does is manually verify deal by deal, suggesting the real problem is data trust, not notification frequency

You can run this same prompt across multiple interview transcripts to find patterns across customers, not just one. That is where the signal gets strong.

Building High-Quality AI Habits

Kenny Kranseler | 00:45:35 – 00:47:20

A few habits that will make AI work better for you in this process and in your product management work broadly:

Give AI context to be successful. AI is a great assistant, but if it does not know who your persona is, what you’re trying to accomplish, or where you’re coming from, it will not do a good job. In this session, I made sure Claude understood who Priya was, what I was trying to learn, and what the interview was for.

Treat AI as a teammate, not an oracle. It does not have all the information. You have to provide it. Do not assume it knows your product, your customers, or your business.

Give it structure. Tell it how you want feedback to come back. Specify the format, the depth, and what you are looking for. Consistent structure produces consistent, usable output.

Tell it what you’re trying to achieve. Do not just query it for random ideas. Give it goals. Where are you trying to go? What outcome does the output need to support?

Iterate. This is not a one-and-done relationship. Give AI feedback on its outputs. Let it adjust. The more you work with it on a specific task, the better it gets at understanding what you actually need.

Key Pillars Summary

Kenny Kranseler | 00:47:20 – 00:48:55

Three pillars for using AI to do better empathy interviews:

Getting started. Use AI to build a loose topic guide from a persona. Ask for areas to probe, not specific questions. Give the conversation room to go where the customer takes it.

Conducting the interview. Rehearse with AI before you go live with a real customer. Use roleplay to practice staying open-ended, following the “why,” avoiding leading questions, and sitting in silence. Do this more than once. And when you are in the real interview, stay human. You are talking to a person. Listen actively. Watch their body language. Catch what they do, not just what they say.

Analyzing the output. Use AI to take messy transcripts and extract the real insight: jobs to be done, pains, gains, says versus does, persona validation. Do it across multiple interviews to find patterns you can actually act on.

The one thing AI cannot do for you is listening. You want to be present. You want to be in that conversation. Use AI to coach you beforehand and synthesize afterward. The conversation itself is still yours.

Resources and Upcoming Courses

Ryan Cantwell | 00:48:55 – 00:51:59

A few things to be aware of as we wrap up:

  • State of AI for Product Management report: scan the QR code on screen or download it from chat. Free insights to help you and your organization mature in AI adoption
  • Certification outcomes study: free to download. See the impact that product management certification has had on others in the field, and what it has done for their career trajectory
  • In-person training in Austin: we’re delivering Optimal Product Management live in Austin in early December. Great place to be that time of year
  • Next webinar: Beyond the Backlog on Wednesday, September 2nd, with Kenny and Dean Peters. Dean’s going to walk through how GitHub can help product managers, not just developers, use AI more effectively as part of their product work
  • Live online courses: Optimal Product Management and AI Product Management available in September, October, and November. Wherever you are in your product management journey, there’s a path for you

Q&A: Navigating Stakeholder Conflict and CS Alignment

Kenny Kranseler & Ryan Cantwell | 00:51:59 – 00:57:30

Question:
A product manager conducted a customer interview facilitated by a colleague on the customer success team. Afterward, the PM shared insights with the VP. The CS colleague disagreed with the PM’s interpretation. The VP sided with the CS colleague who had been there 20 years. How do you handle that?

Kenny Kranseler:
A few things. First, transcribe every customer interview. Full stop. Not just to help in a situation like this one, where there’s a he-said-she-said dynamic, but because it gives you documented, quotable evidence of exactly what was said. When someone challenges your interpretation, you can go back to the record. “Here is what the customer said. Here is the quote.” That is a different conversation than two people’s memories competing against each other.

Ryan Cantwell:
Schedule a debrief immediately after the interview, with everyone who was in the room, and mean immediately. Before anyone’s interpretation has had time to solidify. Collect your thoughts together, agree on what the key insights were, and get alignment in the moment. Once people leave and an hour goes by, everyone’s version of events starts to drift.

More broadly, this is also a stakeholder management situation. Customer-facing roles like CS and sales often have long-established relationships with customers, and those colleagues can sometimes act as gatekeepers or interpreters. Be sensitive to that:

  • When someone from CS or sales brings you to a customer, they may answer on the customer’s behalf. Watch for that and redirect
  • Bringing a neutral third party to the interview as a witness can help: “What did you hear between the two of us?”
  • Sometimes you need the CS or sales person out of the room entirely. Their presence changes what the customer is willing to say, especially when the buyer and user have a hierarchical relationship

I have literally kicked salespeople out of customer interviews. I told them upfront: I’ll come on the road with you, and I want a half hour with these customers where I can do an empathy interview. But you’re leaving the room. They were not happy. But I was not going to get real feedback with them present. It takes that kind of directness sometimes.

Q&A: Interview Scaling and Pattern Recognition

Kenny Kranseler & Ryan Cantwell | 00:57:30 – 01:00:45

Question:
How many interviews before you can trust that a pattern you’re hearing is real, versus just listening to the loudest voice?

Kenny Kranseler:
The rule of thumb I use: stop doing interviews after you feel like you can forecast what the customer is going to say. When you start to hear the same themes coming up before they’re said, when you can predict the story before the customer tells it, that’s when you have enough signal. That number will vary depending on the complexity of the problem and the diversity of your customer base. For most situations, you’ll reach that point somewhere between five and ten interviews. But don’t stop too early just because you feel like you got your answer in the first two.

Question:
What’s the difference between a leading question and just being friendly? How do you draw the line?

Ryan Cantwell:
Avoiding leading questions came up as the biggest struggle in our poll for good reason. We all have perspectives, and it’s genuinely hard not to share them.

The practical difference:

  • A friendly question opens space: “What does a typical Monday look like for you?”
  • A leading question closes it by embedding an assumption: “Is it frustrating when the data isn’t current on Monday morning?”

The test: does your question give the customer room to answer in any direction, or does it funnel them toward the answer you’re already imagining? If you can hear yourself suggesting an answer in the question, stop and rephrase it. Replace the suggestion with curiosity.

Going deeper on what someone said is always safe. Suggesting a solution, a feeling, or an outcome is almost always a lead.

Thank you, everyone, for joining today. Keep the prompts, use the process, and get out there and talk to your customers. See you at the next one.

Webinar Panelists

Kenny Kranseler

Kenny has led product teams from KFC to Amazon to Seattle-based startups. He has the range and strategic insight to differentiate what works from what doesn’t.
LinkedIn

Ryan Cantwell

Ryan Cantwell helps B2B teams align strategy and execution. With energy, clarity, and storytelling, he makes product thinking contagious at Productside.
LinkedIn

Webinar Q&A

Empathy interviews are structured one-on-one conversations with customers designed to uncover the real motivations, behaviors, pain points, and unmet needs behind how they use a product; not just what features they want next. They matter now more than ever because most standard input methods, from support tickets to customer success summaries, surface feature requests rather than genuine needs. As Productside’s Kenny Kranseler explains, customers will ask for a faster horse when what they actually need is a car. Empathy interviews uncover the functional job (what someone is trying to accomplish), the emotional job (how they want to feel doing it), and the social job (how they want to be perceived); none of which appear in a feature request list or a secondhand CS summary.
AI empathy interview preparation works best when used before and after the conversation; not during it. Before an interview, product managers can give AI a detailed customer persona and ask it to generate four to five loose topic areas to explore, creating a flexible guide rather than a rigid script. AI can also run realistic roleplay simulations, letting PMs practice staying open-ended, following the “why,” and avoiding leading questions before speaking with a real customer. Productside’s live demo used Claude to roleplay a sales operations manager persona named Priya, generating authentic responses about forecast challenges, manual spreadsheet workarounds, and data trust issues; all before a real customer conversation took place.
A topic guide is a short list of areas to explore in a customer interview; a script is a fixed set of questions asked in a predetermined order. Product managers should always use a topic guide, never a script. Scripts shut down the customer’s story by forcing them into a question-and-answer format that produces surface-level responses. Topic guides give customers room to take the conversation where their real experience lives; surfacing unexpected details, emotional context, and manual workarounds that no pre-written question would have found. In Productside’s live demo, the most valuable insight of the entire interview, Priya’s hidden Google Sheet tracking 20 to 30 deals she didn’t trust in the CRM, came up organically mid-conversation, not in response to any planned question.
After conducting empathy interviews, product managers can use AI to transform messy transcripts into structured insight by running a targeted analysis prompt. The output should include: a needs summary, functional and emotional and social jobs to be done, key customer pains and gains, persona validation, and a “says versus does” comparison that flags where customer behavior contradicts what they claim. This last element is especially powerful: in Productside’s Priya interview, Priya said automated reminders would fix her data accuracy problem, then immediately contradicted herself by noting that reminders were already being sent with no behavior change; revealing that the real problem was data trust, not notification frequency. Running the same prompt across multiple interview transcripts surfaces patterns that no single conversation can confirm.
The four most common empathy interview mistakes are: asking leading questions that funnel customers toward pre-assumed answers, pivoting into solution space before the customer’s story is fully told, cutting off a customer mid-thought to move to the next topic, and failing to quantify pain by asking customers to rate the severity of a problem on a concrete scale. AI helps fix all four through pre-interview roleplay practice: PMs can simulate full conversations, then ask AI to review the transcript and flag exactly where they led, solutioned, or rushed. Productside’s live debrief showed Claude identifying that Kenny moved into solution space at exactly the wrong moment; right after Priya had opened up about her spreadsheet and was in her most detail-rich, insight-rich state. That self-critique loop, practiced repeatedly before a real customer call, is how PMs build the muscle to stay curious instead of pitching.