“I’m incapable of doing my job without AI”: How this PM uses Claude + ChatGPT as his second brain

1. 0:00 – 3:11

   Introduction to Amir

   1. CVClaire Vo0:00

      You've actually set up these independent, fully loaded brains in Claude and ChatGPT, and then used those to hold the context for you and be more efficient with your job. So show us what one of these brains actually looks like.

   2. AKAmir Klein0:14

      Yeah, let's dive right into it. So here's an example of a project. You can see here that I have a bunch of them, and in it, we have my files and my instructions. Right now, there's 20 files. It started with two, or three, or four, and then just kept growing and growing in terms of its brain. This is what all the projects look like. They all just have a lot of files, instructions, and you can see my endless amount of threads that I have with it.

   3. CVClaire Vo0:37

      Okay, so what do you think are important files for a product manager to put in?

   4. AKAmir Klein0:42

      So I usually kick it off with a ping pong. So I give it any data that I have. If there is anything that references as to what exactly is going on to start things off, then it helps me and the thread that I'm having here in projects get an understanding as to what I'm looking to achieve and more or less, like, what we're talking about.

   5. CVClaire Vo1:00

      [upbeat music] Welcome back to How I AI. I'm Claire Vo, product leader and AI obsessive, here on a mission to help you build better with these new tools. Today, I have AI PM Amir Klein, who works at monday.com, and is in the top five AI-powered PMs on his team. He's gonna show us how he uses AI to build scrapers to get customer feedback into his custom GPTs, use those custom GPTs to not only make his product sense better, but improve feedback that he's gotten from his boss, and finally, he is the first guest that will demo voice mode live for us on air to show how he prepares for product interviews. Let's get to it. This episode is brought to you by GoFundMe Giving Funds, the zero-fee DAF. I wanna tell you about a new product GoFundMe has launched called Giving Funds, a smarter, easier way to give, especially during tax season, which is basically here. GoFundMe Giving Funds is the DAF, or donor-advised fund, from the world's number one giving platform, trusted by 200 million people. It's basically your own mini foundation without the lawyers or admin costs. You contribute money or appreciated assets, get the tax deduction right away, potentially reduce capital gains, and then decide later where to donate from 1.4 million nonprofits. There are zero admin or asset fees, and while the money sits there, you can invest and grow it tax-free, so you have more to give later, all from one simple hub with one clean tax receipt. Lock in your deduction now and decide where to give later. Perfect for tax season. Join the GoFundMe community of 200 million and start saving money on your tax bill, all while helping the causes you care about the most. Start your giving fund today in just minutes at gofundme.com/howiai. We'll even cover the DAF pay fees if you transfer your existing DAF over. That's gofundme.com/howiai

2. 3:11 – 6:24

   Using custom GPT project folders as “second brains”

   1. CVClaire Vo3:11

      to start your giving fund. Hey, Amir, happy to have you here.

   2. AKAmir Klein3:14

      Thank you, Claire. Happy to be here.

   3. CVClaire Vo3:16

      So one of the reasons why I'm excited to have you here is you don't just like using AI. We were just talking about it. You are now highly dependent, and maybe incapable, of doing [chuckles] your job without AI. So tell me how you got to this point where you're really depending on AI to do your job as an AI PM.

   4. AKAmir Klein3:37

      Yeah, I think there's, like, the whole context switching, um, is something that's really difficult for everyone, and as a PM, when you're going back-to-back meetings and jumping between different initiatives and different requirements and asks from people, you have so many things to juggle, that I discovered, with the help of everything that's out there online, that you can silo, like, brains of conversations and threads in GPT and Claude, that it made me super dependent on it in order to just do my work. [chuckles] And I'm doing it really well, I think. I'm doing it really efficiently, so that's how I got to where I am today.

   5. CVClaire Vo4:12

      Okay, so what you're saying is you, like all PMs, have to context switch and hold a million different threads in your mind at one time, and instead of burdening your poor human brain with all that context switching, you've actually set up these, like, independent, fully loaded brains in Claude and ChatGPT, and then use those to hold the context for you and be more efficient with your job.

   6. AKAmir Klein4:40

      Absolutely. Nailed it.

   7. CVClaire Vo4:41

      Okay, so show us what one of these brains actually looks like. What are some of the use cases for my collect- my file folder of, of brains?

   8. AKAmir Klein4:51

      Yeah, let's dive right into it. Um, so I'm going to, I guess, dive into what a, I guess, folder looks like, my projects look like.

   9. CVClaire Vo5:00

      Mm-hmm.

   10. AKAmir Klein5:01

       So here's an example of, um, a project. You can see here that I have a bunch of them, and in it we have, like, my files and my instructions. Um, so right now there's 20 files. It started with two, or three, or four, and then just kept growing and growing in terms of its brain. And I'll dive into, like, some of the files that are in here and kind of, like, what I'm trying to achieve to help you understand, like, how I put this into action, but just this is what all the projects look like. They all just have a lot of files, instructions, and you can see, like, my endless amount of threads that I have with it, um, like, nonstop. [chuckles]

   11. CVClaire Vo5:37

       Okay, so what do you think are important files for a product manager to put in, or what are some unique sources of, you know, context for these GPT projects that you have found particularly useful or unique?

   12. AKAmir Klein5:50

       It varies obviously on what you're working on, but I always start with some sort of, like, reference that it can understand whatever it is that I'm starting with. So I usually kick it off with a ping pong. So I give it any data that I have, any data whatsoever. So if there's a company kickoff deck, if there's a team kickoff deck, if there is another reference to a PRD, if there is-... anything that, um, references as to what exactly is going on to start things off, then it helps, like, me and the thread that I'm having here in projects get an understanding as to what I'm looking to achieve and more or less, like, what we're

3. 6:24 – 11:02

   Building a Reddit scraper with Claude’s help

   1. AKAmir Klein6:24

      talking about.

   2. CVClaire Vo6:25

      Great. And then what do you want these, you know, second brains or these GPTs to actually do for you? So what is the, you know, what does this one actually do for you?

   3. AKAmir Klein6:35

      I'll start this one kind of like where it began. Um, I started at my company, at monday.com, leading, like, the AI agents, uh, team, the whole initiative there. And I'm sure anyone who's on AI today can allude to the fact that AI agents is probably, like, one of the most fantasized thing that is in SaaS. Everybody wants it, everybody wants it to work, everybody has, like, this vision of what it can do, and it can do everything. And whatever you hear outside is also what you hear inside, and there's a thousand different voices and different opinions of where they want it to go and what they believe it will be. And I was seeking some voice of reason that was unbiased. I was looking for conversations that people were having that weren't connected to my company, that didn't have any specific narrative they wanted to go down. I just wanted to know what people are interested in, what they expect, and what they want, um, which is what led me to Reddit.

   4. CVClaire Vo7:34

      Okay, so you were a PM. You were given the great and wonderful task of building out agents. No one has agreed on what the definition of an agent is. No one has agreed what the agent should do. No one really understands if customers actually even want agents. Technically, there's a lot of complexity and, um, aspirationally, there's a lot of goals around these. So this is a very common product manager problem, where you're given a big, meaty, interesting, technically complex problem with a lot of hopes and dreams in it, and then it's your job to go make it reality. And so what I'm hearing from you is you went and sought out these different resources, um, that were external, that could give you more of an unbiased filter on which to, uh, make some product decisions, and maybe even go kind of back to your leadership partners or peoples w- with big hopes and dreams and say, "This is what people actually want," or, "This is what's actually possible." And so you mentioned Reddit. What did you actually pull from Reddit that you think was so useful here?

   5. AKAmir Klein8:35

      So I'm gonna walk you through, like, what I did here in order to scrape conversations on Reddit, uh, with the help of Claude. I had this idea, I had this vision, that I'm gonna be able to get this file or many files containing thousands and thousands of conversations that people have about anything that I wanted. In this case, I wanted about agents, AI, monday.com, all in relation with one another. Um, so I went into, I went into, uh, to Claude, and I said: "Here, help me, help me scrape conversations." I started my conversation here with just, like, "I wanna find everything that's written about monday.com, whether it be on Reddit, Twitter, LinkedIn, anything that's online. Um, and I wanted, like, a sort of an automated, free way to do this." So it went into, like, a bunch of different options. We realized that LinkedIn and Twitter have, like, more limited APIs. Some of them are, like, paid. Eventually, we got down to what I needed, how to get it, and how can do it more automated. So I asked it: "Okay, great, we know Reddit is the place I need to be. Now give me a step-by-step guide of what I need to do." So I'm not not technical. I am a little bit... I have a little bit of experience with things, um, but I wanted to make sure I'm not skipping any steps. I wanna make sure that everything I'm doing here is going to be exactly what needs to be done, 'cause I've had a vision of what I'm getting out of this here. So it literally walks me through here, like, "Open Safari, go to Brew, write a command, open your terminal," right? Like, literally step by step of what I need to do, as if I'm five years old, and I don't know how to, uh, program or code anything, which was fantastic. It tells me, "Go into Reddit, go into this, and make, make a, a developer account." I literally followed it step by step. I didn't even try changing any name. I left it as, as it asked, and then we get to, you know, doing Brew install. It's getting stuck, whatever. Eventually, we get to the good stuff, where I have everything set up. Um, it's finished. Now what do I do? I had the environment set up, and I had all the packages that I need to set up, and now it starts giving me, like, all these commands to run. So install different packages and libraries that I need in order to run the script that I write for me. "Here's a script that I'm gonna help you write, and let's get started." So basically, it writes me up this script we can see over here, where it... I put in my client ID it would ask me to take out from Reddit, my client secret, and my username. Um, and then it scrapes everything that has to do with, like, Monday, with things that I'm looking for with AI, and whatever it be. Um, and lo and behold, came out a file with 34,000 rows of conversations.

4. 11:02 – 14:06

   Analyzing 34,000 rows of Reddit conversations

   1. AKAmir Klein11:02

      [chuckles]

   2. CVClaire Vo11:03

      Okay, so you have this, like, very, again, a very classic product manager problem of, "I know people are talking about this product." Like, "I, I know. I know there's information out there, and I can manually cull through and, you know, do Reddit searches and search through Twitter, but I wanna do this in a programmatic way. I'm technical, but not that technical, and dear God, I don't wanna write a bunch of Python." And so what you did is you went to Claude and you said: "I have this general problem. I just wanna know what's out there." Uh, you actually said... The prompt is really interesting. It says, "I wanna know everything out there, everything everyone-

   3. AKAmir Klein11:34

      I'm ambitious

   4. CVClaire Vo11:34

      ... is saying." That's a very broad-

   5. AKAmir Klein11:36

      I'm, I'm super ambitious.

   6. CVClaire Vo11:36

      Yeah, very ambitious flow. And then, you know, Claude narrowed you into, "Well, here are the actual accessible sources of data for you," and then you wrote a script to go pull all that Reddit data. And then, what did you do with that, you know, you said 34,000 or something, 3,400, um, line spreadsheet? What did, what did you do with that data?

   7. AKAmir Klein12:01

      Um, so I'll show you. Basically, first of all, I wanna show you what the file looks like-

   8. CVClaire Vo12:05

      Mm-hmm

   9. AKAmir Klein12:05

      ... just so you get a glimpse as to what this, like, monstrous file is. So this is what 34,000-... rows of threads looks like, um, in relation to the conversations that I wanted, the theme that I was looking for. So this is great, um, and it really helps me understand, okay, there's a lot of conversations. I also went crazy with it. I said, like, "Give me different files. Show me how it looks like in compet- in, in terms of our competitors. Show me conversations that are happening, like, in relation to our competitors, like comparing us to any other competitor and their AI, and give me some sort of idea." When I had all that information, I went back into Claude, and I said, "Okay, now be my analyst-

   10. CVClaire Vo12:42

       Mm

   11. AKAmir Klein12:42

       ... um, and start breaking this down." I'm, like, really into data. Um, I have, like, a little bit of a, a data background. So I came into it, and I said, "Here, I'm giving you this, this file." This is one of the files that was pulled out with 30,000 rows of conversations. And I said, "I want you to summarize it in a table where you put, like, the frequency, so how often it comes up, at what percentage. I need weights because I need numbers to go back to my team and for myself to know what to prioritize, see what the biggest, hottest topics are that people are, are discussing and some of the key discussion points." And out comes this table, which is really nice-

   12. CVClaire Vo13:17

       Mm-hmm

   13. AKAmir Klein13:17

       ... and really helpful because I have, "Okay, here's, here's a s- here's some sort of idea." I know, like, what are the top five areas that people are wishing and hoping and having conversations about around this, their expectation and their dreams around what AI or an agent would do.

   14. CVClaire Vo13:32

       And did you find that this... Did you spot-check this at all? Did you just trust it? Did you feel like this was relatively accurate based on what you saw in that, in that spreadsheet?

   15. AKAmir Klein13:40

       Yeah, it's a really good question. Um, I did check it. Obviously, I'm not checking 34,000 rows of conversations, but, um, I did do, like, keyword searches, and I looked at different threads and conversations that I pulled up. Also, what I typically do with whenever I ask GPT or Claude to analyze something, I say, "Reference. Give me at least one or two quotes as to where this came up from." And I have tons of conversations where it does that, so then I fact-check it. I copy that. I go back into-

   16. CVClaire Vo14:04

       Yep

   17. AKAmir Klein14:04

       ... the document and make sure

5. 14:06 – 18:04

   How to build effective custom GPT knowledge bases

   1. AKAmir Klein14:06

      that it was actually there.

   2. CVClaire Vo14:07

      Awesome. And so you built this scraper, you took the data, you analyzed it. Did you then feed this analysis back into that GPT project we saw before? Kinda how does this p- become part of your corpus of one of your secondary brains?

   3. AKAmir Klein14:23

      Yeah, fantastic. So I'll show you, um, how that looks like back in the project itself. So here I am in the project.

   4. CVClaire Vo14:30

      Mm.

   5. AKAmir Klein14:31

      Um, you can see that I have, like, the Reddit, like, the AI threads on Reddit and in general. Like, these are different, like, Reddit thr- files that I had, and this is how I started things off. I had my kickoff that I had, um, that my manager did, and I had a couple of pages that, like, I took from our site of, like, how Monday works, and I just command P'd it, so it's a PDF, and put in. Another classic example of everything is text.

   6. CVClaire Vo14:56

      Yep.

   7. AKAmir Klein14:56

      And then it... So it has an idea. It knows what Monday is. It knows what in, more or less, like, what our goal is as a team, and here are some conversations that are happening, so you have some sort of idea as to what people's dreams and hopes are when it comes to AI.

   8. CVClaire Vo15:09

      Yeah, I just wanna call out that little trick for folks that may- might have missed it, 'cause I do the exact same thing, which is I actually just use our marketing website or our support website, and I just print things as PDFs and download them and then upload them to GPT projects or whatever kind of store. Because then I know it's scoped to that specific piece of content. I've done it a lot with, like, support pages. I've done it with the homepage, obviously, 'cause that's usually, like, the source of truth of your, your positioning, and then I do it a lot with, like, pricing pages. So I'll go, like, print-

   9. AKAmir Klein15:36

      Mm

   10. CVClaire Vo15:36

       ... a bunch of competitor pricing pages and drop them in a GPT. So I think that's a really great source of information for a GPT.

   11. AKAmir Klein15:44

       No, for sure. I think, like, once you st- like, when I started with this and I had the CSV file from, from, um, Reddit, my mind was like, "Okay, literally everything is text." So, like, even you may not necessarily think of, like, a slide. It was a slide from, like, Google Slides as, like, text, but it is, so you just download it as a PDF, and you have just, like, whatever it is, however many slides, here as a reference. Um-

   12. CVClaire Vo16:11

       Mm-hmm

   13. AKAmir Klein16:12

       ... and like the example that you said as well, like, the list is endless of what is text. So I have it also, like, instructions more or less of, like, what I want it to do. Um, make sure that it is a professional in product management and expert in, in all this stuff. It has a good product strategy mindset and product sense mindset. Um, I also gave it, like, a lot of, like, instructions on, you know, "You know how to give feedback to me. You know how to give candid feedback, and you're not gonna be, like, too nice to me. You know how to challenge- "

   14. CVClaire Vo16:40

       Yep.

   15. AKAmir Klein16:41

       [chuckles] ... 'cause I hate it when the AI will be, like, super supportive of every idea that I come up with. Like, "Push back on things. You have the knowledge, so feel free to challenge me on things that I come up with." And then the ping-pong begins, and then we start going back and forth, and I have thousands of conversations here.

   16. CVClaire Vo16:55

       And so are you using this as a thought partner? Are you using this to prep for meetings with your boss? Like, what are the most common use cases for this GPT for you?

   17. AKAmir Klein17:04

       All my projects, I start with a scope. I wanna get to some sort of, like, outline, overview that'll help me with a PRD, with a product review, with some- something that I can deliver to my team, so a narrative. So it always starts off with a ping pong, always. And then once I've gone to some sort of document, and I have, like, this overview doc of what it is, I've played around with it, I've downloaded it, and I put it back into its knowledge, and now it has that-

   18. CVClaire Vo17:28

       Nice

   19. AKAmir Klein17:28

       ... as a basis, and then it goes from there. The more work I do with it, the more knowledge I add into it, uh, into its files, and then everything that comes after that is just super intuitive and fluid. I'll give you examples. Like, if I need to write something, how often I get asked questions by, like, marketing and with the communications team of like, "Okay, you're releasing something. Can you give me, like, a two-line sentence of what it is?" Those are classic examples of, like, catch me in the middle of the day on Slack and asking me then. I'm, I, I'm- I don't know. I don't have the capacity to think about that, but here is something that has a brain for this. I ask it that question, and it usually does a pretty good job, and then, lo and behold, here

6. 18:04 – 21:53

   Creating a custom writing coach from Lenny’s Newsletter

   1. AKAmir Klein18:04

      we go.

   2. CVClaire Vo18:05

      Great. And then, you know, speaking of documents, I think one of the common complaints that, you know, people have about AI-generated documents is the writing maybe isn't great, and at the very least, maybe it's good thinking, but it's quite long. And I know you've also built a custom GPT to [chuckles] make sure that your writing is concise. So can you show us a little bit of what that looks like?

   3. AKAmir Klein18:28

      ... Yeah, there's al- also a background to that. Um, so if any of my past managers and colleagues see this, um, they will be able to confirm this, but I've been given feedback multiple times, and my writing is good, but it can be really long.

   4. CVClaire Vo18:43

      Mm-hmm.

   5. AKAmir Klein18:43

      So my Slack messages on updates and channels are really long, and it's hard for people to get through. And Lenny released... I think there was a newsletter where he had Wes Kao, like, say more or less, like, her recommendations on how to write concisely, how to deliver a message well. And again, example here of everything is written. Um, I took these newsletters, I Command+V'd it, downloaded them, and I put it into a custom GPT as, like, my guidelines to how to write. Um, and I'll show you what this custom GPT looks like.

   6. CVClaire Vo19:14

      You've seen the doom and gloom headlines: AI is coming for your job. But the reality is a little bit brighter. In Miro's latest survey, 76% of people say AI can boost their work. It's just that 54% still don't know when to use it. As a product leader and a solo founder, I live or die by how fast I can turn fuzzy ideas into crisp value propositions, roadmaps, and launch plans. That's why I love Miro's innovation workspace. It drops an AI copilot inside the campus, so stickies, screenshots, and brainstorm bullets can become usable diagrams, product briefs, and even prototypes in minutes. Your team can dive in, riff, and iterate, and because the board feels like a digital playground, everyone has fun while you cut cycle time by a third. Miro lets humans and AI play to their strengths so that great ideas ship faster and happier. Help your teams get great done with Miro. Check out miro.com to find out how. That's M-I-R-O dot com. So while you're pulling that up, I have to ask, did you try Lenny Bot yet? Because [chuckles] I, I know he has his own bot with all of these newsletters and some content loaded in, but you wanted something really precise, right? So you created your own.

   7. AKAmir Klein20:37

      I had something really spec- specific-

   8. CVClaire Vo20:39

      Mm

   9. AKAmir Klein20:39

      ... because I know what it is that I needed. Um, so I took in her newsletters, 'cause he hosted her-

   10. CVClaire Vo20:47

       Mm

   11. AKAmir Klein20:47

       ... on, on his podcast. She also referenced a bunch of books.

   12. CVClaire Vo20:51

       Mm.

   13. AKAmir Klein20:51

       I looked for those books online and put them in here. I never read them, but I just put them in here. [chuckles] And I was like, "Okay, you must know then how writing and delivering is." And then you have this here, this, this instructions that, like, this GPT rewrites Slack messages to be more concise, to be more clear, to be readable. Maintains the natural voice of the user, 'cause I want anyone in my company, anyone in general, to use this GPT, so it doesn't, like, make it super AI-y. Like I said here as well, like, you know, avoid too much dashes. Don't give so much bullet points, all that stuff like that. And then I come into here so often, I'll write a message out on Slack to, like, myself. I'll copy it, paste it in here, and it does a really good job of making it super concise, sticking to these guidelines of, like, delivering to the point, having really good follow-up and lead up as to what the next section will be. It is a lifesaver. People respond to my Slack messages more now.

   14. CVClaire Vo21:42

       So I know that you didn't read the books, but I do have to recommend, as somebody with a liberal arts degree, those are exceptional books on writing, so I highly recommend reading them. We'll link to them, and the newsletter, and

7. 21:53 – 24:08

   Using AI for professional development and feedback

   1. CVClaire Vo21:53

      Wes's website in the show notes. But what I love about the approach here is a lot of times people get sort of professional development feedback like you got. "Your writing's great, it just needs to be shorter. Like, you just write too much." And a lot of people don't really build for themselves a way to coach into improving those skills, right? They're like, "Thanks for the feedback, but I'm not gonna sit with my boss for an hour and have them, like, edit my Slack messages and edit my PRDs. Uh, maybe I'll read a book, or maybe I can do some research," but it's, it's a hard thing to put into practice into your day-to-day. And so what I like about this is you took some coaching feedback that you got, and you're like, "Cool, I'm gonna create myself a coach for this very specific development piece that I need to work on, and I'm gonna rely on experts that I think are great, and then I'm gonna use it [chuckles] day to day to actually make things better." And then what I liked, uh, again, about what you said is, "And then I'm gonna share it with, with my colleagues, 'cause I'm sure I'm not the only one that has, has this problem."

   2. AKAmir Klein22:57

      Yeah, exactly.

   3. CVClaire Vo22:57

      So I think it's a really... yeah, a really nice process.

   4. AKAmir Klein23:00

      And I think it's really cool that you mentioned, like, all that. Like, I'm not gonna now send this and have, like, somebody rewrite it. Like, go pre-GPT era and my dependency on AI, I did that. I got this feedback. I'm all for feedback. I love feedback-

   5. CVClaire Vo23:13

      Yeah

   6. AKAmir Klein23:13

      ... and I love improving anything on stuff like that. I would literally do that. I would write out a message to myself, send it to my manager, and say, "Hey, tell me how this can be improved." And he wouldn't respond. He came back a couple days later. By then, like, I lost momentum as what I wanted to say, and then this is a game-changer for me. And yeah, I definitely think other people probably feel it as well.

   7. CVClaire Vo23:32

      Well, and it makes me think of another idea that maybe folks would find useful, which is, you know, people get performance reviews on maybe an annual or twice annual basis. Like, what do you do with that? And, you know, maybe what you do is you load up your performance review, and you load up your peer feedback into a GPT, and you reference it. And you say, "Hey, I just made this doc, but based on the feedback that I've gotten in the past, is there any, like, blind spots that I might have?" I think it's a really great method for self-reflection and self-improvement if you're... you know, if you're open to feedback like it seems like you are.

   8. AKAmir Klein24:04

      Yeah, absolutely. The, the- it's limitless what you can

8. 24:08 – 31:49

   Preparing for product interviews with GPT voice mode

   1. AKAmir Klein24:08

      do.

   2. CVClaire Vo24:09

      Okay, so for our last use case, you know, you seem like a person that likes a lot of feedback and likes to prep. You actually use GPT voice for mock interviews, and we are going to push the bounds of the technical capabilities-

   3. AKAmir Klein24:24

      Sure

   4. CVClaire Vo24:24

      ... of this podcast and see if we can demo this. So while we're getting set up, can you tell us a little bit about the problem you were trying to solve and why, why you kind of went down this path?

   5. AKAmir Klein24:34

      ... Yeah, okay, so basically, anyone who's prepping for interviews, for, for product interviews, knows that there is a lot of content out there. Uh, YouTube is loaded with mock interviews. Exponent is a fantastic channel. Um, there's a lot of different people I could shout out, but they're not gonna come to the top of my head, haven't been interviewed for, in a, for a job in a while. But basically, interview prep, what I would do, and I think a bunch of people would do leading up to product interviews, was watch these videos, um, over and over again. I'm obsessive. I would watch the same three videos over and over and over again, remember any nuance and stuff like that, that when I walked into an interview, it was like copy/paste of what I just listened to. And I would feel super awkward to be like: "Let me practice on something." I tried in front of a mirror, it didn't work. My mirror's not talking back to me. I tried with my wife, and I was like: "Hey, these are the type of thing, things to ask me." She didn't know what I wanted from her. So when GPT Voice came out, I'm like, "I'm going to do this. I will be that lunatic in my living room talking to AI as if we're in the movie Her," and it was, like, the most exciting thing ever. Like, I forgot that I'm talking to AI. I feel like I have this, like, this interview coach who's with me at all times, who's constantly, like, challenging me and pushing me and, like, finding the right things, coming up with fantastic questions, and I'm like, "This is what I needed." And I walked into every interview after that super prepared. It was the best feeling ever, and literally can't recommend it more. Anytime anybody asks me, like, how to prepare for a product, uh, interview, I said, "Just open GPT Voice and just go for it."

   6. CVClaire Vo26:06

      Okay, so I have to ask, because you have this very structured setup from, you know, a writing GPT perspective and a product GPT perspective, but it sounds like you're just opening up GPT Voice and, and going for it. Are there any tricks to getting GPT Voice into the right mindset? And then once you tell us, let's try it. Let's try it live. Let's see if we can make it happen. I'm gonna make you-

   7. AKAmir Klein26:29

      Okay

   8. CVClaire Vo26:29

      ... feel so awkward. [chuckles]

   9. AKAmir Klein26:31

      Let's... We, we could, we could definitely do it. Um, I'm gonna be super rusty. Um, there's-

   10. CVClaire Vo26:37

       That's fine.

   11. AKAmir Klein26:37

       First of all, there's two different approaches. Yes, I am organized and structured-

   12. CVClaire Vo26:40

       Mm

   13. AKAmir Klein26:40

       ... but I do love to freestyle it. I did make a custom GPT for something like this. Um, I can show it to you so you have an idea of what it is, but basically, like, so similar to the other one, it has, like, instructions on, like, what product sense questions, like, you need to look out for and what product execution questions you need to look out for. And like, for instance, the, the interviewer, the interviewer needs to clarify the company's mission and the product goals and identify specific target users and pinpoint their real pain points, and then design a solution that solve those pain points, and then they all have a sequence to it. So I wanna make sure that it understood, like, what exactly you need to look out for and what type-

   14. CVClaire Vo27:21

       Mm

   15. AKAmir Klein27:21

       ... of things you need to ask. And then obviously, I put in some knowledge base to it, too. I have my own product manager interview prep document. I took Ben Harris's, uh, posts. He's fantastic on product coaching, um, on product sense and product execution, and now it has this whole background. You can pop into this GPT and have a conversation with it. Um, and it's super guided on what a product sense interview looks like, what a product execution interview looks like, so that's, like, more structured. You can use-

   16. CVClaire Vo27:49

       Mm

   17. AKAmir Klein27:49

       ... the voice mode and go for it. Um, freestyling, we can try to freestyle. Let's do that.

   18. CVClaire Vo27:54

       Yeah, let's, let's try it.

   19. AKAmir Klein27:54

       I'll open my freestyle. One other thing that I would add, you can also plop in your CV and the link to the, uh, role that you're applying for, so there's more context as to who you are and what they're looking for in the role itself, and then there's more of, like, guidance towards things to pick on. But you can freestyle it, and also, like, if you have, like, inter... back-to-back interv- [chuckles] back-to-back inter- if you have back-to-back interviews with, like, GPT Voice, it can learn from, like, what you've made mistakes on beforehand. Like, I was really bad at user segments. I used to always struggle with it, and then, like, I'll be like: "Okay, now let's, let's talk through this and figure it out." And then the next two interviews, they go, "Your, your, um, user segment actually has improved. You really figured out how to, like, narrow down different user segments that are super different." So anyways, uh-

   20. CVClaire Vo28:35

       Great

   21. AKAmir Klein28:36

       ... let's get started.

   22. CVClaire Vo28:37

       Okay, let's try. For those that are listening, not watching, I've made Amir very uncomfortable by making him talk to AI.

   23. AKAmir Klein28:44

       How do we open here?

   24. CVClaire Vo28:45

       Let's tell it this.

   25. AKAmir Klein28:46

       Let's, let's start. Can you, c- can you hear me?

   26. SPSpeaker28:48

       Like, everything's working fine on my end. How can I help you today?

   27. AKAmir Klein28:52

       So I want you, um, to interview me, um, for a product sense interview. I'm interviewing for a role at monday.com as an AI product manager, um, and I need you to give me a product sense question. It's really important that you don't guide me in anything, that you are super candid, that you let me take the reins on where I take this interview, and in terms of, like, the direction. Um, don't lead me to anything, don't be, um, specific or give me any hints. Um, and at the end of it, give me some candid feedback. Let me know if that makes sense to you.

   28. SPSpeaker29:27

       That makes perfect sense. I'm happy to help you prep for your interview, so let's dive right in. Uh, here's a product sense question for you: Imagine you are tasked with improving the user experience for one of monday.com's AI-powered features. How would you go about identifying areas of improvement, and what steps would you take to implement those improvements? Uh, feel free to take it in any direction you like.

   29. AKAmir Klein29:52

       Claire, I'm gonna stop there.

   30. CVClaire Vo29:53

       Oh, that's great.
9. 31:49 – 33:04

   Additional use cases for voice mode

   1. CVClaire Vo31:49

      And then I'm gonna call out a couple other, you know, maybe adjacent use cases that people can be inspired by. My, my darling husband is one of the people that can prep for public speaking by, like, going to the bathroom and s- speaking in the, in the mirror, and I can't, I can't do it. I, I hate, I hate looking at myself [chuckles] too much, which is odd to say for, for a podcaster. But, you know, just thinking about maybe doing kind of like speech and public speaking prep via voice mode, like, "Where did I do well? Where did I not do well?" Um, and then I'm even thinking for my kids, we're doing a really, um... We're trying to do a consistent job of getting them to read out loud at night and think about their punctuation and emphasis and all that. And I'm like, "Oh, my gosh, my kids should do" [chuckles] "oral dictation to GPT voice mode and get a little, get a little feedback." Although I love, I love our bedtime story time-

   2. AKAmir Klein32:43

      Yes. [chuckles]

   3. CVClaire Vo32:43

      -to do that.

   4. AKAmir Klein32:44

      No, totally. The world of AI with kids is unreal. It's unreal.

   5. CVClaire Vo32:49

      Yeah.

   6. AKAmir Klein32:49

      Uh, it's just, you know, we can go that for hours as well.

   7. CVClaire Vo32:52

      I know.

   8. AKAmir Klein32:52

      Uh, the stuff that I do with my daughter is so much fun.

   9. CVClaire Vo32:55

      I think at some point we're gonna do a roundtable parents in AI episode, where we get, like, half a dozen of us on to talk about all of our tips and tricks.

10. 33:04 – 35:43

    Recap of Amir’s AI workflows

    1. AKAmir Klein33:04

       Yeah, totally.

    2. CVClaire Vo33:04

       Okay, Amir, this was so great. So just going back to the beginning, what did we cover? Product GPTs, using Claude to help you discover and technically implement ways to access customer and market feedback data, do data analysis, and then, as you say, create these GPTs as a secondary brain that you can ping pong with, that you can go back and forth with and have a thought partner on. Um, you taught us how to take feedback well, which is, AI aside, [chuckles] a really good trait, but take feedback and then put that feedback from your, um, your boss or your peers into a custom GPT that then can coach you and improve something you need to improve on. For you, it was using, um, concise writing tactics. And then the third piece, again, a very feedback-oriented person, you prepare for interviews. You wanna do a great job. There's a lot of resources out there to prepare for interviews well, and so you used GPT voice to do mock interviews, and not only do mock interviews, but get ongoing feedback about how you're improving on product sense and product execution interviews. So you really are dependent, or at least a power user of AI, and you have been-

    3. AKAmir Klein34:16

       I, I love the power user.

    4. CVClaire Vo34:17

       Mm-hmm.

    5. AKAmir Klein34:18

       I love the power user. I think there's, there was, um... In, like, my company, they really try to reinforce AI to everybody, and they did this, like, top 10 AI power use U- like, PMs using AI, um, since January 2025. My son was born in February, so I was on pat leave for two months. So I was active at the company for four months of this six-month range. There is 90 PMs there. I was number four of power users, and I have a two-month delay on people, so I'm... Consider myself number one or number two. [chuckles] Um, and I also wanna add, like, a bonus thing, Claire, if that's all right-

    6. CVClaire Vo34:57

       Mm-hmm

    7. AKAmir Klein34:57

       ... that your episodes inspired me a lot, that I vibe quoted, um, a scraper that can do what I did for anyone. It's in the works. I wish that I'd finish it up until this point, but basically, like, the concept is what I wanted, find any discussion anywhere, um, where users can kind of just write in their own natural language what they want to scrape from any platform. So let's say, for instance, I have, like, I want every discussion that relates to AI with monday.com. I have hard mentions that, like, with all the keywords, it'll always reference that-

    8. CVClaire Vo35:26

       Mm

    9. AKAmir Klein35:26

       ... towards whatever it is I'm searching, the platform that I want, the, the output, and then I want it to search for it and give me the file. So it's in the works. It's not done yet. I have, like, the results are okay, but when it's out there, um, I'll share it.

    10. CVClaire Vo35:42

        Well, if it's live

11. 35:43 – 38:50

    Lightning round and final thoughts

    1. CVClaire Vo35:43

       by the time we, uh, put this episode out, we will link to that in the show notes, or we'll do a follow-up, and I'm happy to be a beta user, 'cause that looks super useful. All right, we're gonna wrap up with two lightning round questions and get you back to spending time with voice mode. My first question is, for PMs that are not on the top of that leaderboard, what do you think the number one thing they're missing out on is?

    2. AKAmir Klein36:11

       Honestly, it's the ability to be at so many places simultaneously. That's really what it is. I can be asked on the spot from pretty much anyone in the company about something, and I can give you an answer really quickly. I couldn't have done that pre-AI.

    3. CVClaire Vo36:29

       Amazing. Uh, I, I think it's such a great... I mean, it's, you know, as a, as a leader-... it's our favorite trait of our favorite PMs, is know everything, are everywhere at all times. It's really tough as a PM [chuckles] to do that, though, 'cause of, you know, just physical limitations, space, time, mental capacity, all that kind of stuff. And so, you know, I do recommend these, these tools because they just give you a lot of horsepower in terms of answering questions, getting stuff done right away, and holding all that context without totally, totally breaking your brain. Okay, last and final question I ask everybody: When AI is not listening, and it's giving you bad results or analyzing your data incorrectly, what is your tactic for getting it back on track? Do you bribe it with money?

    4. AKAmir Klein37:16

       I just yell.

    5. CVClaire Vo37:18

       [laughs]

    6. AKAmir Klein37:18

       I just yell. [chuckles] I'm in voice mode anyways, right? Can you not hear my tone? Um... [chuckles]

    7. CVClaire Vo37:24

       You're the first person that's admitted to getting, [chuckles] getting angry with AI on this podcast.

    8. AKAmir Klein37:30

       So pre- pre-voice, it was caps. I'd be like-

    9. CVClaire Vo37:32

       Okay

    10. AKAmir Klein37:32

        ... "This is not what I asked for."

    11. CVClaire Vo37:34

        Mm-hmm.

    12. AKAmir Klein37:34

        Um, and now voice, you can, you can get upset. Uh, no, but in, in mostly what I do is I'll copy the response, and I'll say, "This isn't what I was looking for. This is an example of what I was looking for."

    13. CVClaire Vo37:46

        Okay.

    14. AKAmir Klein37:46

        Um, that's usually what I do, um, to just have some sort of benchmark reflection on it.

    15. CVClaire Vo37:52

        Amazing. All right. Well, thank you so much for sharing your workflows with us. Where can we find you, and how can we be helpful?

    16. AKAmir Klein37:58

        I'm, like, not too active on social media. I, like, pop into LinkedIn from time to time, so that's always good. And honestly, anyone who's posting anything cool on what they do with AI is super, super helpful. The list of things that I do here, a majority of it is just inspiration from what I've seen online, so keep it up.

    17. CVClaire Vo38:17

        Great. Well, you keep it up as well. Thanks for being here.

    18. AKAmir Klein38:20

        Of course, Claire, thank you for having me.

    19. CVClaire Vo38:21

        [upbeat music] Thanks so much for watching. If you enjoyed the show, please like and subscribe here on YouTube, or even better, leave us a comment with your thoughts. You can also find this podcast on Apple Podcasts, Spotify, or your favorite podcast app. Please consider leaving us a rating and review, which will help others find the show. You can see all our episodes and learn more about the show at howiaipod.com. See you next time!

Episode duration: 38:50