
“I’m incapable of doing my job without AI”: How this PM uses Claude + ChatGPT as his second brain
Claire Vo (host), Amir Klein (guest)
In this episode of How I AI, featuring Claire Vo and Amir Klein, “I’m incapable of doing my job without AI”: How this PM uses Claude + ChatGPT as his second brain explores a PM’s AI second brain: scrapers, writing coach, voice interviews Amir Klein (PM at monday.com) explains how he offloads constant PM context switching into separate Claude/ChatGPT “project” workspaces, each stocked with files, instructions, and ongoing threads.
A PM’s AI second brain: scrapers, writing coach, voice interviews
Amir Klein (PM at monday.com) explains how he offloads constant PM context switching into separate Claude/ChatGPT “project” workspaces, each stocked with files, instructions, and ongoing threads.
He demos using Claude to build a Reddit scraper, generating a dataset of ~34,000 rows of conversations about AI agents and competitors, then having Claude summarize themes with frequencies and cited quotes for spot-checking.
He shows how he turns internal artifacts (kickoff decks, PRDs, website/support pages printed as PDFs) plus external research into a living knowledge base that he continuously re-uploads as his thinking evolves.
Finally, he shares “personal development” workflows: a custom writing-coach GPT trained on Wes Kao/Lenny-style guidance for concise Slack updates, and GPT Voice Mode for realistic mock product interviews with candid feedback.
Key Takeaways
Treat each initiative as its own AI “brain.”
Amir creates separate Claude/ChatGPT projects per domain, each with dedicated instructions, files, and long-running threads so he can jump between initiatives without losing context.
Get the full analysis with uListen AI
Start every project with grounding artifacts, then “ping-pong” to an outline.
He seeds the project with kickoff decks/PRDs/any available data, uses iterative back-and-forth to produce a narrative (e. ...
Get the full analysis with uListen AI
Use AI to build the data pipeline—not just analyze the result.
Instead of manually searching, he asked Claude for an automated, free approach; Claude narrowed feasible sources (Reddit vs. ...
Get the full analysis with uListen AI
Turn messy market chatter into prioritizable themes with numbers.
After scraping ~34,000 Reddit rows, he asked Claude to summarize into a frequency/percentage table so he could identify “hottest topics” and bring quantitative signals back to his team.
Get the full analysis with uListen AI
Demand citations to keep analysis honest.
To avoid blind trust, he asks the model to include 1–2 direct quotes per theme, then verifies via keyword search in the raw dataset—fast spot-checking without reading everything.
Get the full analysis with uListen AI
Everything is “text” if you capture it correctly.
He prints slides, web pages, and support/pricing content to PDFs and uploads them—creating scoped, reliable reference material for positioning, onboarding, and competitive comparisons.
Get the full analysis with uListen AI
Convert performance feedback into an always-on coach.
After receiving feedback that his writing is too long, he built a custom GPT trained on concise-writing guidance to rewrite Slack messages while preserving his voice—making day-to-day improvement practical.
Get the full analysis with uListen AI
Notable Quotes
“You can silo, like, brains of conversations and threads in GPT and Claude.”
— Amir Klein
“I had a vision… to get… thousands and thousands of conversations… with the help of Claude.”
— Amir Klein
“Now be my analyst… summarize it in a table… frequency… percentage… I need weights.”
— Amir Klein
“I hate it when the AI will be… super supportive… Push back on things.”
— Amir Klein
“Just open GPT Voice and just go for it.”
— Amir Klein
Questions Answered in This Episode
In your project “instructions,” what exact phrasing consistently gets Claude/ChatGPT to push back instead of being overly agreeable?
Amir Klein (PM at monday. ...
Get the full analysis with uListen AI
When you summarized 34,000 Reddit rows into frequency tables, how did you define categories—did the model create them bottom-up, or did you provide a taxonomy first?
He demos using Claude to build a Reddit scraper, generating a dataset of ~34,000 rows of conversations about AI agents and competitors, then having Claude summarize themes with frequencies and cited quotes for spot-checking.
Get the full analysis with uListen AI
What’s your minimum viable “starter pack” of files for a new PM initiative (kickoff deck, PRD, metrics doc, support pages, etc.) before the ping-pong begins?
He shows how he turns internal artifacts (kickoff decks, PRDs, website/support pages printed as PDFs) plus external research into a living knowledge base that he continuously re-uploads as his thinking evolves.
Get the full analysis with uListen AI
How do you prevent polluted knowledge bases—e.g., outdated PDFs, contradictory drafts, or model over-relying on old conclusions you later changed?
Finally, he shares “personal development” workflows: a custom writing-coach GPT trained on Wes Kao/Lenny-style guidance for concise Slack updates, and GPT Voice Mode for realistic mock product interviews with candid feedback.
Get the full analysis with uListen AI
What are the ethical/legal guardrails you follow when scraping Reddit and using that content internally (rate limits, PII removal, quoting, attribution)?
Get the full analysis with uListen AI
Transcript Preview
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.
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.
Okay, so what do you think are important files for a product manager to put in?
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.
[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 to start your giving fund. Hey, Amir, happy to have you here.
Install uListen to search the full transcript and get AI-powered insights
Get Full TranscriptGet more from every podcast
AI summaries, searchable transcripts, and fact-checking. Free forever.
Add to Chrome