As I look back at all the topics we've covered on this show this year, we have definitely talked a lot about AI.
So I hope you don't mind if we throw one more on the pile, because this one is probably the most actionable AI episode we've done so far.
My guest today is Tal Raviv, the genius behind the popular course entitled to build your personal PM productivity system and AI co -pilot, which is a pretty fancy name, and it's also a pretty big deal for someone who previously described himself as an AI skeptic.
But for Tal, the game changer was figuring out why so many PMs struggled to extract the full value out of their LLM tools and uncovering the tactics that actually have the power to transform your productivity.
And yes, we will be breaking those tactics down in this episode.
We'll dive into how onboarding AI is much like onboarding a new team member, why iterative prompt engineering is key, and practical advice for how to start tinkering with AI tools and build confidence in your skills.
Let's jump in. Welcome back to the Product Manager podcast.
I am here today with Tal Raviv.
And Tal, thank you so much for joining us.
This is so exciting.
Can you tell us a little bit about your background and how you got to where you are today?
Sure. I started out studying chemical engineering, which has absolutely nothing to do with where I am today.
I mean, it does have been the fact that it just like made me good at learning hard stuff and diving in.
I started a SaaS company with a few friends from college, and we did that for four years.
And then when it was time to get a real job, it was when I was basically looking around, I was like, well, I really like doing like support and marketing and coding and design, and what is the real job version of that looks like?
And then there's this new thing called product management at the time.
So I applied and I totally wouldn't be able to break in today.
But at the time I was able to break in.
Sweet. Today we're going to be talking about something else everyone else is trying to break into, which is using AI effectively.
I think all of us are kind of trying to wrap our hands around how can we use the technology more effectively, more practically.
So we're going to be focusing on how we can use AI in our daily workflows.
But before we dig in, let's take a second to talk a little bit about what's going on in your life.
So you are leading a cohort course to kind of help PMs leverage AI more effectively.
Kind of what was the journey to lead you to that path?
I started working on this course with Maven, and I started working on this one, we talked about this, but I was at one hand, working on a lot of Riverside's generative AI initiatives and leading those.
I totally wasn't using AI in my day to day.
I was like, Oh, this is great for the product.
This has nothing to do with my day to day.
This can never help me with all these things I need help with.
And so the course ended up being all about productivity and building systems for your productivity for yourself.
And then your team as a product manager really is very high leverage for your productivity in looking at your organization, managing emotions, all this low tech stuff.
That's the course I teach.
And then in parallel, I was just playing around a lot with AI.
The thing that kind of was the switch that got me to actually look at it as something that I could personally use was I was at Riverside and I just was between a rock and a hard place.
And there was just no way I could accomplish what I needed to without leveraging chat GPT.
And even then I kicked and screamed in mind.
And then my engineering team lead still was like, let me just show you.
Just like, just stop.
This is how you do it, which opened my eyes to this really simple principle that the reason I wasn't using it right was I wasn't giving it enough context, just like a person.
And that was kind of like a mental shift for me.
And then I just started pushing that further and further.
I was like, well, how much context can I give it?
And what if I gave it the same amount of context that I would give a new hire that I was helping on board?
What if I had a conversation with it like I would with a new hire?
And then what if I start to involve it in a particular initiative just like a new hire?
So I changed my mental model to really think of it as, well, what would set a person up for success?
And I'm doing that in my spare time.
And I'm teaching this course.
And then I started to realize, oh my God, this could actually be useful at work.
So it took me a long time, 2024 for that to dawn on me.
And then over time, I made that part of the course.
So now the first two weeks of the course are about getting your house in order and all the things that still really matter, even with AI.
And the final week is an intensive session of three sessions of building your PM AI co -pilot.
OK, awesome. I'm excited to break into each of these things a little bit on a more granular level shortly, especially onboard.
I think the idea of onboarding your AI is super interesting.
I'm excited to get to that.
But before we get there, you kind of described your own skepticism, working with AI initially.
And I think this is just something common that we see across the board in product management and in other fields.
Why do you think that is?
Why are people skeptical or hesitant to bridge that gap between awareness and implementation?
I think for most of the last two years since strategy BT came out, 20 and a half, it hasn't been that great at a lot of stuff.
Or it would require so much editing and thinking that it was just like what the person you hire or somebody really junior or why we don't hire an intern for this particular role is just going to create more work than it's worth.
And I think first of all, what changed is that AI got way better.
Context windows got bigger, personalities got better.
And just like the models, they just got more intelligent.
And I think at this point, there was a famous talk recently with head of product of Anthropic and head of product of OpenAI both on stage and they said something to the extent of intelligence is no longer the bottleneck.
There's ways to make it more intelligent, but it's not the bottleneck.
The bottleneck is either pre -training or context or just like what does it know?
What is it exposed to?
I think also if we show up to an LLM and we just have a one sentence prompt, even if it's like the biggest prompt expert in the world swears by it and fine -tuned it and crafted and polished that prompt, it's still one sentence, right?
It's still going to basically give you the average of the internet.
And if you really wanted to do the kind of work that you do, any one of us as individuals is not the average of the internet.
There's also knowledge that we bring, we combine and we synthesize.
So starting point, let's give it that knowledge, see what it can do.
Yeah, I think this is a really good point.
I think sometimes we step into a chat and kind of expect, well, it knows everything, so it should have the perfect answer.
But it's kind of like, if you stop someone on the street while you're just walking to the grocery store and you're like, hey, can you spin it on a PRD?
They're going to give you their best attempt without all the context and like owned knowledge that you have.
So this is, yeah, it's kind of one of those like dumb moments that I don't think we all really have had yet.
I'll give another example that I think makes this like really vivid.
This is when I talk to people that don't happen to be product manager and how I explain.
I also use, I've been experimenting with using AI as my nutritionist and keeping me on track.
And the way I explain it is, well, if I was to work with a real nutritionist, I wouldn't show up and be like, cool, tell me what to eat and how much and when and all that stuff.
Right. Like the nutritionist would be like, no, how about you sit down and ask you some questions?
Like, no, no, no, I just want you to tell me you suck as a nutritionist.
I'm out of here. Right.
Like that wouldn't happen.
That's not a conversation that would happen.
But somehow that's the conversation we have with AI.
Yeah. That's spot on.
Hope it's working out for you.
I want to talk about this AI Copilot.
And this is all really, really interesting.
There's so many things, so many angles to approach when we're talking about using AI effectively.
So let's start with the tool.
So tell me about this tool, AI Copilot.
What's the game changer here?
I'll say the components.
The components of the AI Copilot is pick chat GPT, cloud, something else.
I personally really like cloud for this.
Practically speaking, the project's feature, it's a paid feature.
It's totally worth it.
It's just really good for a lot I'm about to describe.
It just makes it really easy.
These things just take one click.
This can be accomplished with any LLM with a little bit of duct tape and a little bit of manual work.
So first of all, you use custom instruction and system prompt, whatever you want to call it to tell it how to behave.
So you want it to be someone that is challenging you, that asks you lots of questions, that doesn't accept your assumptions necessarily.
Like who would be this ideal coworker that's sitting next to you in kind of like pair PMs?
And it would be somebody who pushes you on certain behaviors or values.
So the behaviors I put for myself is biased towards action, try to get value to customers soon.
I try to make decisions without all the data, but that could be totally opposite for other industries or companies.
And from there, so that's like the personality that you hire, kind of like thinking of an interview process.
That's the kind of person you want to hire.
The next step is you hire them and you want to onboard them.
So if a new PM joined the team, we'd probably give them, here's the doc with like the mission vision strategy, here's the deck with the customer persona.
And if it was a coach for me, I would probably say, here's my performance review, her last several performance reviews.
Here's what I'm struggling with.
Here's what my manager gives me feedback on.
Here, again, going back to the new PM, here's some gossip.
Hey, you know, sit down.
Let's have a cup of coffee.
Let me tell you about some of these stakeholders.
Let me tell you about some of the people who are going to be on your team.
Here's what they're particularly good at.
Here's where they need you to lean in more.
Here's the stakeholder and what makes them feel good and what stage of the process, but don't step on this landline.
And this just goes on and on.
Here's the org chart.
Here's my team. This can be endless.
This would be what a new person would be absorbing from different sources joining a new company.
So you hire them, you've onboarded them, and then you got to put them to work.
Well, you have to tell them, hey, I need you to work on this initiative, or I want you to guide me on this initiative.
So you start a thread and you say, here, I barely know anything about this initiative, I just had this dumped on me.
All the conversation with the founder, we have to move fast on this.
Here's what I know.
Here's what I think.
And you'd have a conversation with it.
And then during that entire thread about that particular initiative, you can start asking it for outputs.
Like it can be as simple as the classic example of writing documents, fine.
But it can go way beyond that.
It can be a thought partner.
It can simulate a hard conversation.
What's the most important thing you should do next?
The answers to that are really amazing.
This is where the magic happens, because it will constantly reference actual names of people in the organization.
It'll reference, hey, this is an opportunity to really act on that feedback that's constantly in your performance reviews.
Hey, this is a really good time to loop in this stakeholder.
And I've been blown away by the timing and what I suggest those things.
It's very astute. And during the initiative, I suggest gossiping even more, all the more context.
So think of you're working on something and two weeks into it, you have this hallway conversation with head of sales and they just drop this constraint on you, or your manager says, well, if we had this new information and if it can't achieve this, if that's another scope, this isn't worth shipping
at all. And you sit back at your desk and you're like, oh my God, you won't believe the conversation I just had.
Like it could be whoever's sitting next to you.
Do that with your co -pilot, right?
They need to know that too.
And you can get all the...
I can go on and on about all the things you can do, you can have it work with you on and help you with.
But at the end of the day, it's also worth closing the loop and telling you what happened, sharing any retro results.
The magical moment is where you can ask it, okay, listen, this is the end of the story.
It's great. We shipped it, it worked, it didn't work.
Can you please tell me, kind of like, what's all the new information that you learned in this thread?
What should another new product manager know and learn from this initiative?
And then you can, with Claude, one click with chedgipd copy paste, add that back to the brain, the project knowledge, or use it in your next thread.
That becomes kind of all that context that you added this time.
It just got a little bit bigger.
For me, the reason I like Claude, that's one click.
It's really optimized for that.
It immediately applies to all your threads, but that's the high level.
So that's how you get this copilot that gets smarter over time, gets wiser, starts to apply lessons from other initiatives right away, and so on.
I'm a little stunned right now, I'm gonna be honest with you.
This is a lot more comprehensive and impressive than I expected.
So first of all, was this the conversation that you had with your colleague before who said, no, let me show you?
Or how did you put together that the tool was capable of that kind of retention and decision making capabilities and emotional intelligence on top of all of the tactical?
I'm overwhelmed. I didn't know.
The way this connects for me was like this lightbulb moment with my colleague or Leon where he's like, listen, that example that started with, hey, I really need to write a lot of user stories really fast.
These drain my brain cells and it's a ton of time and they have to be really precise.
And how the heck is chatgbt gonna help me with that?
I tried screenshotting Figma, I tried Figma plugins.
I was like, this is more work than it's worth.
And then his suggestion was, how about you sit down in chatgbt, give it a template and then tell it everything that it needs to know about this user story.
I was like, OK, that's a lot of typing is like, no, don't type, dictate, just talk.
And another really important development in the last two years, year is speech to text got really good.
So OpenEye released this whisper model.
There's a lot of app developers that implemented it on Mac OS, on phone apps and so on.
And if you compare that with like the native Apple Windows dictation, it's just so much more accurate.
It makes me way more excited to use it.
It's way more useful.
I basically can just hold down a button.
And just talk for a long time and it's super accurate.
So now the process starts to look like I'm basically talking the way I would when I have a new engineer joining an initiative or I'm kicking off an initiative and chatgbt is very good at taking context formatting in a certain way and inferring in between.
And so when I did that, I was blown away because first of all, I'd never saw the designs that did things in the user stories that I never thought about.
So like fill in the gaps and I almost had to do no editing.
I might have like, if I had to delete anything, it was basically things that were just redundant.
It was just being overachiever.
That was like my label moment of like, Oh, context, and then you add in the average of the internet because, you know, products are not that unique.
Like it can figure out that, Oh, like the hallucinate part is a good thing there.
It hallucinates good stuff, connects the dots.
And once I was thinking in that mode, I was just test, I guess, kept pushing the boundary of like, how much context can you get it?
It's incredible. Okay.
So let's go into kind of like tactical instructional mode here.
I'm sure that there's tons of PMs listening who are like salivating at the idea of using this tool.
So we've, we've got our, you know, LLM of choice.
If you're just kind of wanting to get up and start, what are like the critical steps that you would say for onboarding the L .M?
I would start simple.
If you're not paying, you know, pro for any of these services, you should for other reasons, but you don't have to do it for this reason.
Start one thread and tell it how you want to behave.
I'd even say step zero for all this is download something like better dictation for your computer.
You know, the super whisper, better dictation whisper flow.
There's a ton of these and practice with really good dictation.
Everything else I say, this is going to be way easier.
Open a new thread. Tell how you want to behave.
This is kind of a system problem.
Putting it at the top of the thread is like 90 % of the value and then say, if you have a doc, that's like how the, you know, you can copy paste the landing page of your, you know, you can talk about how you would describe your product at a cocktail party or something, even if you don't do all the org
chart stuff and you don't talk about the stakeholders and you don't give the performance reviews, even just a little bit.
It's way better than what you had before.
And you can pause there.
Just do that. Just do like a little bit of context.
And then give a context on an initiative that you want to work on.
And that could be another 60 seconds of here's what I know.
So far, we know this signal.
We know we hear these support tickets.
That's that. And that's it.
Hit enter. See what happens.
Have you in your course had other folks follow this process?
Like, have you been able to see other people generate value out of it?
I would love to hear like anecdotes of how you've seen this in action.
We did a few beta groups and then started to create a community.
And the community is kind of like the people who are really pushing it forward, like really experimenting like way beyond just one person like myself.
And I want to share, I just pulled it up right now and I just share like everybody's their aha moments.
So one person created really, really impressive prototypes, not the kind of prototypes that like we've seen on, you know, it's winter demos, like things that are like, wow, that's really precise because I had so much context.
Some product managers just use it to learn about an industry that they were like, dropped into.
This is a not PM example.
And I know somebody who used it for quarterly planning.
They're like a product leader, which is like, that's even further than I've ever pushed in terms of context that it would need.
A lot of people use it as a thought partner.
I think that's a really good mental model for this.
It's like just for conversation and understanding and it helps you think of better stuff.
And I think for writing PRDs, one person reported here that what made this particularly a better way of writing PRDs is it will ask you a lot of questions first.
It was enough to know what it doesn't know and it will prompt you back to have a conversation with you and say, before I write this parody, I would really want to know these things like, oh, of course, you know, ask me anything, just write it for me.
It's amazing. Yeah, it's a really fascinating example of people just like pushing it way beyond what I was doing.
So incredible. My head is already I can feel the gears turning in my mind how I can leverage this because it sounds like a lot of this, the general principles here are cross applicable to a lot of different capacities.
Super cool. Let's dive a little bit into prompt engineering because we're talking about giving context and like how important it is to give good quality information to the LLM and to design your prompts well.
So what are some of the best practices that you've leveraged, like strategies that people should adopt to create better prompts and just be able to provide better quality context?
I think there was like a phase in LinkedIn where like the entire feed was prompt engineering slide decks and like people changing their titles to prompt engineer.
And I remember just having like this like weird feeling about that.
I'm like this can't be this way for long.
This whole way of treating prompts like magic spells or magic potions or Pokemon that you collect like, it just doesn't feel right and lo and behold over time those became less and less important.
So when it comes to prompt engineering, there's two things.
One, I've already talked about context, treat it like a person, use dictation, makes it a lot easier.
And the second thing is just tell it what you want, just like a person, like be clear about what you want and then hit enter.
If it's not what you wanted, think why didn't do what you wanted and go back to the previous message and clarify it.
There's always like a, you know, whatever tool you're using.
There's a little pencil on that bubble in the chat.
And I think this is one of the most underrated features of cloud or chat GPT.
You can go back up and you can edit and add another sentence or tell it, but I don't want you to approach it this way or be really clear about this or make sure that this is covered and it will just redo that.
It'll read you everything that's below that.
And that pencil, I think is the biggest tool for prompt engineering, in quotes, because at the end of the day, you're just trying something.
It's a code you try something.
Did it work? Didn't work.
Let me change it. Try it again.
And you don't have to worry about, you know, using this special phrase or asking it to think a certain way, like all these things are fading away.
They're quickly evaporating.
So it's not the right skill.
I don't think it's not the right skill.
I think the right skill is tinkering, diving in, and just like iterating, just like not being afraid of it.
It's all just a mindset change.
Web designers, this one's for you.
I've got 30 seconds to tell you about Wix Studio, the web platform for agencies and enterprises.
So here are four things you can do in 30 seconds or less on studio.
Adapt your designs for every device with responsive AI, reuse assets like templates, widgets, sections, and design libraries across sites and share them with your team.
Add no code animations and gradient backgrounds right in the editor and export your designs from Sigma to Wix Studio in just a click.
Time's up, but the list keeps going.
Step into Wix Studio and see for yourself.
Oh, okay. I like that.
It's practical, but it's not intuitive to just think about, you know, how can I think differently about how I'm designing my prompts?
Okay. So I want to talk a little bit about how you have sort of approached this because I think maybe something we haven't been clear about throughout this episode is like this is all sounding like it's a tool that you're selling.
This is more of a skillset than anything.
So why have you decided to focus on this as a workshop or offering it as a workshop rather than building a scalable AI that already does all of these things?
Why did you feel that that was the best approach?
So as a career product person, coder, I was very tempted.
Like when my first impulse was, okay, that's an insight.
Like how do I build a product here?
And when I just dug into like what's missing, what's stopping very smart people from doing this very smart thing?
What's in between? It's not functionality.
It's not some UI or, you know, some like network effect or anything that products can provide.
It's simply permission or guidance or a mindset shift.
So I thought, well, I could build a product and then I could scratch my head for the following year, trying to figure out onboarding because everything I just described is a hell of an onboarding for somebody to do on their own in their free time.
Nobody's going to do that.
Or I can be more valuable by being with people live, answering questions, saying them, Hey, you cleared out your schedule and now we're here on the zoom together, so I'm going to give you five minutes to do this thing.
There's nothing, you know, you've cleared out your schedule and I'm here to answer your questions and basically build it out together.
For me, that's like a 90 something percent activation rate.
If you want talking product terms, right.
And I think that value retention and all that just is a way better approach to it.
So I think another reason is also that a lot of the things that are missing or that could be better, I would just be so shocked if that wasn't on open AI or Clods, there's so many things that you can tell would make this process even better, like how they manage memory and changes over time and do threads
know about each other and that's like you were mentioning before, so clearly applicable to all industries and all roles that I'd rather just help people use this correctly than the barriers mindset, behaviors, permission right now.
That totally makes sense.
And it's just, yeah, I think building an AI product when you know who you're up against is that's a, it's a big bet.
So I think that just leveraging the tools that are available more effectively that's an area I think a lot of people really need to grow in.
And speaking of which, I think this is just kind of a common perception across the board that we're all sort of behind.
I think a lot of people know that they could be doing more with AI and there's some, maybe an intimidation factor about getting started and really digging in.
So what would you say to people to help them overcome the perception and like really start building these skills in a way that feels accessible?
You know, I was talking to somebody today, a director of product, and she said this really well, she was like, well, our entire organization just missed the bow on AI and productivity and applying in our roles.
And I told her, no you didn't, there is no bow.
Everybody feels that way.
I've had that conversation with so many people.
These are people at these cutting edge companies that are in every other way.
The most modern ways of running organizations.
They're all expressing some kind of FOMO.
And like I feel like, I hear the sentence, I feel like other people are way ahead here.
I feel like somebody else is doing this better.
So first of all, I can assure everybody that's not true.
This is something that everybody right now is in the same place.
It feels that way because everybody's talking about it.
Though I think if you feel like that, just start tinkering, try it for something really small and specific and keep those principles in line.
Context and iteration.
Just set aside 10 minutes to do that.
Go that tinkering muscle and you'll see that that's snowballs.
It'll pull you in. It'll just make you want to iterate more.
And if that's the right thing to choose, like very quickly, you'll find yourself being the one teaching others.
So I think a lot of it's like it's intimidation.
It's feeling like you're really behind.
It's not even worth me trying.
If I try, I'm not going to do it the right way.
News flash like everybody feels that way.
I'm glad you said that because I think that's true.
I think that there is this tendency of thinking like, Oh, well, you know, the moments past, it became big.
Everyone's ahead of us.
Why even try? But when you think about it that way, it's not a boat.
It's a taxi. It's a rideshare.
It's there as a service for you.
It's at your disposal.
You tell it exactly where to go.
Yeah. I like that reframing as well.
I'm really appreciating the mindset work that we're talking about today, Tom.
Well, anyway, so you'd mentioned a few different success stories or kind of breakthroughs that people had had either yourself or someone else.
What's the one that you think is like the most woe that you've seen working with AI yourself or working with other folks who are learning how to use it?
For me personally, my, Oh my God moment was having like pricing conversations and it didn't have to say got it, right.
It didn't just like spit out the perfect answer, but it was just a really good conversation and even when it got it wrong, it made me wonder why did that feel wrong and say that the right word for this is thought partner.
It was just like having a smart person sitting next to me, helping me think through something and bouncing off ideas.
And then it helped me get to a better answer, how to bundle something, how to price something and it's kind of like a really smart rubber duck.
It's like the next level of rubber duck.
You know, it's probably not what we imagined.
Like this would be the aha moment for AI.
We probably imagined something very sci -fi and like the super computer from Hitchhiker's guide to the galaxy.
But I think people ask me, you know, do you feel that it's wrong to outsource your thinking to AI?
Do you feel that it's impacting?
I think it's making me smarter.
Just like if I had more smart people more available to talk to as much as possible, that would make me feel smarter, right?
That's what we all seek.
And I heard when we look for, you know, where we work and who we surround ourselves with.
So Hey, for me, pricing was a really cool moment that I didn't think, you know, do an amazing job, but it's also like glow, but like flow.
You feel afterwards of like, I just had a really good conversation.
That's a really great example because it's such a nuanced conversation.
We just had a really great chat with Jim Consu from Duolingo.
Is that a product to Duolingo?
It's a very complex conversation that you have to have when you're thinking about pricing and it's such a, you know, you really have to be so careful about where exactly you place those markers.
So yeah, I can only imagine it's to your benefit to have an incredible tool that's able to help you workshop your decision -making process and like apply that in future situations as well.
This is so fascinating to all.
Thank you so much for joining us.
I have really learned a lot today.
Where can people catch up with you online if they are curious about taking your course or just want to hear more about your thoughts?
Sure. Uh, LinkedIn is, is the first answer.
There's the course it's on Maven for people who are listening to this and they're just like, I just want to run ahead with this and just tinker away on Maven.
I also am putting this like the demo I did.
You can just see everything I did.
I use this notion playbook where it has like all the prompts and the steps and it's much more structured.
Basically everything that's in this conversation.
Just like laid out.
That's something you can just purchase without having to take the whole course.
And the course is just really good for people who want to have access, open Q and a office hours, reach me and like just do much more together.
Awesome. Well, thank you so much for joining us.
It has been an absolute pleasure.
Thank you. Thanks for listening in for more great insights, how to guides and tool reviews, subscribe to our newsletter at the product manager .com slash subscribe.
You can hear more conversations like this by subscribing to the product manager, wherever you get your podcasts.
