Rami (00:00)
And I think it's really easy to be a bit lazy and use a bazooka to destroy a paper and use an LLM for flows that just don't require

Ilan (00:08)
Hi everyone, welcome to Prompt and Circumstance I'm Ilan David is not here today, but we have a very exciting interview to share with you. We are talking to Rami from one of our favorite AI enabled companies, Querio.

Rami (00:09)
you

Ilan (00:35)
Rami, how's it going?

Rami (00:35)
Good

to be here, Ilan. I think ⁓ I'll keep up with David today as your sparring partner. So I'm very excited. Very, very excited.

Ilan (00:43)
So Rami, today we are talking about product management and AI, your adventure coming into product management, starting your own company, how

What is your company? What is that product? How does it help product managers? And then we want to talk about some AI workflows that product people can take advantage of. So let's start from the beginning. tell me about yourself.

Rami (01:13)
like straight up anything about myself. Yeah, so...

Ilan (01:15)
Yes, anything.

Tell me a funny story about yourself,

Rami (01:18)
Whenever I start a new team, I can be a bit hard to swallow initially because I just rush to do things before I really understand anything. So good segue. So yeah, I'm Rami. I'm the co-founder of Querio We're an AI enabled startup, which is also building the best data product in the market, a new AI BI tool, which we'll get into later. But before that, I was working at Amazon for about half a decade. And I started off in operations, very relevant to my funny story, but ⁓ in operations, you're managing.

these 100, $100 million warehouses with a ton of people you have to track your time up, whatever it is. And there was a major like technical outage at one of the warehouses. And then you have a bunch of employees there who are usually supposed to be like doing work, know, packing things, whatever it is, and you have to kind of understand what each person's doing so that you can like, properly pay them and then also see how much time you're spending in each function so you can fix it. And so what the those warehouse managers used to do is that they had to like manually

set everyone to like, they're not actually packing something there. There's an outage and they're just kind of waiting around. And it was really important that I was organized. And so I'm like, AI, I'm going to make an Excel VBA macro that's going to also fix everyone's time. Just put your warehouse code and your credentials and it's going to fix it. And then I'm like one week into this like new manager and I had a manager called Carol and she was last, she ever gives a story, but she ran the sheet and I ended up just

Pretending like no one that day clocked in. I just erased about, what was it, a couple thousand people's clock-ins for the day, completely, which led to maybe five times the amount of manual work to fix it. With a rush because it was pay week, so they would have not gotten paid for the time that now did not count because they never came in. ⁓ Yeah, that's a work-related funny story.

Ilan (02:50)
Hahahaha!

So, Rami,

you're the living Dunning-Kruger effect. All right, well, why don't we get into that? How did you come into product management at Amazon?

Rami (03:15)
Yes, absolutely.

⁓ Yeah, so I think I'm I think I'm part of that wave of PMs that just like kind of fell into being a PM You know, like I know that there's a lot of kind of a PM roles and some universities are in the offer like PM degrees but back when products managing was becoming a thing like you kind of fell into it by just being a really good at one thing and then needing to kind of Manage that product in a sense. So I thought off at Amazon actually like working and warehouses so I would ⁓ actually run like a shift

for warehouse operations, for like outbound operations at logistics facilities, like fulfillment centers specifically. And I was always this like really nerdy guy looking into data and always like, how can we make this process faster? How can we process more per hour? How can we process more destinations? Like whatever it is. And there was this like team of like more industrial engineers that would like, so Amazon's were after their products. They literally have like generational names and they're all photocopies. And so.

I could do little changes for like the one I worked in, but there was a team that would decide what the actual process looked like for every single warehouse across the world, the US, the EU, et cetera. So they had like these features and actual templates for all the different sections of the warehouse. And so one of the guys who was responsible for the one I was working in, like outbound, really enjoyed some of the experimentation I was doing. And he kind of just like plucked me out of there, made me an industrial engineer, even though I studied economics. And I just spent like two years doing nothing but figuring out how to improve like the...

templates of Amazon warehouses that were going to be opening over the next few years. And I got really good at it. I like I knew every package I got scanned across any section of the Amazon network. I knew exactly what database got updated with what role, what every system was, how all of them played together, what team controlled what system. And when you're talking about like a behemoth like Amazon, just knowing who owns all the different sections of something can make you such a powerful person. Cause I could get something done like in a couple of days.

And then Amazon had designed these new type of warehouses. They're called the 11th generation, which instead of using like those mechanical ways to organize packages, there was like a ton of robots. They're like a big robotic arm, pick up a package and it would put it on a little Roomba that would take it to where it should go. And the template said that I think something like it should do 3000 packages an hour or something, let's say. And it was doing like a thousand. It was way below the design standard and Amazon had spent half a billion dollars.

Ilan (05:32)
Mm-hmm.

Mm-hmm.

Rami (05:52)
So I got tapped to become a product manager of like own this new template end to end and make it hit the design standards. And so I was more like a hardware and software PM out of nowhere where I was having to work with like the software teams that designed the actual software that these robots ran on to hardware to like what were their physical limitations for them to move faster and have the right scanning speed and all those things. It was very fun.

Ilan (06:14)
Wow.

Rami (06:17)
And then my last year, I wanted to get more like B2C software product experience. I moved to the London team, the Alexa International, where I owned product for, as part of like a small product team, all of Alexa's features outside the US.

Ilan (06:24)
Mm-hmm.

Okay.

Rami (06:32)
So

Alexa was put into Alexa US and Alexa International. And the UK had the second highest market penetration, so the team was based out of here. And for a year I got to spend time, how do I improve this global product, one of the most popular hardware consumer products that existed, which was also very fun.

Ilan (06:48)
All right, so you joined the Alexa team. I asked you about this once in...

a non sequitur and a call we were having, but does Alexa record everything that everybody says so that I can start suggesting those products to you in advertisements?

Rami (07:07)
⁓ no. It only records things after you say the wake-

Ilan (07:11)
Are you?

Rami (07:13)
actually I actually get to this like no there was an activation word and especially like before LLMs it was like an engineering marvel to process natural language and understand it and it was expensive. Like Amazon would go bankrupt that was actually processing because just someone saying words like you need this machine to understand what you're saying right or you have to have like one person per Alexa somewhere in the world hidden in a cave listening and like transcribing so yeah yes the mechanical Turk great story.

Ilan (07:35)
The mechanical Turk.

Sorry.

Rami (07:43)
That's a great story, Mechanical Turk.

Ilan (07:45)
Yeah, do you want to tell it quickly? I don't know if the whole audience knows the story.

Rami (07:50)
Yeah, mechanical Turk is a common modern phrase to talk about something that looks kind of like software in the front end, but it's just manual in the back end kind of thing. And it happened because there was this marvelous machine, think it was either between the 1600s to the 1800s, a while ago before machinery, where it was a chess robot, a chess machine that could beat anyone at chess. And it was called the mechanical Turk, I think, literally.

It would play anyone, this machine, and it would actually move things around and everyone was astounded that this machine could beat everyone in chess. It was like a world marvel until one day it turned out that there was like a midget hidden under the chessboard in like this little box actually moving the pieces. ⁓ Very funny story.

Ilan (08:34)
Amazon actually has a product called Mechanical Turk, right?

Rami (08:37)
Yeah, we use a lot of mechanical turks in different places. So like, for instance, when it comes to putting products away that like into inventory, people can order. There is a camera that's actually, this was like in the last five years, it didn't be like the way before. It used to be that you'd have to scan the package you had and you would scan the shelf you're putting it in. And to make it faster, they changed it so that you would just like scan the product and then just immediately put it in a shelf. And this camera would capture what shelf you put it in.

And whenever it had like less than I think, I think an 80 % confidence, it would send it to like some remote team, honestly, I think in India that would confirm which one it was in. And it went from like 80 % of the inventory placements being mechanical turks to like 10 over time. Yeah.

Ilan (09:15)
Mm-hmm.

OK.

about at Querio? Did you guys use Mechanical Turks in the early days? OK.

Rami (09:25)
I forgot that, yes, yes, yes, yes, yes.

Still till today. All the time. ⁓ my god, my poor co-founder. ⁓ it's

like, we're even considering putting some flows in of like, if you request this really complex thing to become a dashboard, then it just needs a bit of manual love. Our AI's working on it. Come back in like an hour. Or we'll notify you when it's done. Yeah, I think a lot of startups do that.

Ilan (09:58)
Mm-hmm. I wonder if

I wonder if Replit does that because every time you try to create a site on Replit, know, it's always like you can get notified when your site is done and so actually we're sending it to our team in India to build your site for you.

Rami (10:08)
Who was that

company that ended up being a mechanical Turk and just shut down like builder.ai or something? That's crazy. I met a founder actually last week who was like, I paid this company a year ahead to build the backend of my software. And it turns out that it was just all like fake and now they have shut down and I cannot recoup my money and they're like a bootstrap founder. Crazy. Yeah.

Ilan (10:33)
my God, that is crazy.

Okay, so let's pick up the thread of the story. You're at Amazon, you're on the Alexa team. That's right, let's keep weaving here. So what happens next?

Rami (10:48)
and like kind of going more into query. So my superpower was always data. I was really good with Excel. was really good with SQL. I was really good with Python. I had access to all the databases. Like I could statistically prove any feature I had actually had the impact measured. And I could also identify what features I had to build out because I could identify the bottlenecks and like where we're dropping off users or where we're dropping off like throughput in operations process. And at Alexa, it was much worse.

of a problem than even an operation. Like my fellow product managers would have to wait like six weeks for any data answers from like this data team because the UK didn't have a dedicated data team. So it's all in the US. And so they were like really bottom of the, they really like at the bottom of the food pile for that. So yeah, you're talking about a $3 trillion company having a six week wait time. you, it was like three, four weeks in the US. It wasn't much better. It's a humongous company. So I was like constantly tasked to like

I tried the SQL classes, tried the Python classes, or like tried to help my other like product managers. And then at one point, you know, there was all like those kind of restructuring the tech companies, everyone's worried they're going to get fired. And I had this like coworker, she called me, she's like crying, like, ⁓ I don't know SQL, I'm not as good, I'm going to get laid off. Like, it's not your job. You're a product manager. Like your job is not to know how to code. Like relax. And then there was GPT-3 at the time, wasn't yet, it was still like the playground.

And I ended up using it to make like a small little pilot of what query was today, which is like, I gave it understanding of like a few key tables and it's like, the PMs could kind of get to working once they have the first initial SQL query written. So I made them like a bit of a tool to help them write their SQL and they really loved it. And I was like, you know, I've always been quite honest earlier and I'm thinking in my head, I'm like, man, like this is really going to enable so many more people to just become a lot more self-sufficient when it comes to the data work that they need. ⁓

feels like it's at the cusp where the technology is. It's my superpower, so I've always been good at it. I've known at this point probably over 100 product managers in my own career that I've helped with this exact issue. And I think between obviously getting paid an Amazon salary mid-20s and saving it, having a really good idea, feeling that timing was really there, it really gave me the courage to just kind of like...

quit Amazon, get rid of the golden handcuffs and just go full time to kind of pursue this idea that I thought was, that I know, I know it was amazing. Yeah.

Ilan (13:18)
Alright, so where does that bring us to? What's the timeline here?

Rami (13:21)
This was summer of 2023 is when I decided to leave Amazon. Querio really started kicking off probably the end of Q4 2023, really like maybe January 2024.

Ilan (13:34)
Okay, so how does that start? So you have the idea at Amazon, you know that data is your superpower. How do we dot, dot, dot Querio now exists?

Rami (13:47)
Dot dot dot. I had a really good friend I met in French class at UT Austin, where I went to university, who ended up going from a machine learning data person to a product manager, also really heavy on his data. He actually worked for two startups back to back that got acquired. And so he was an early stage operator who went from heavy data into heavy product. He was the first product hire, the second company. And we really shared the same passion of like, have seen both sides of this aisle and the problem.

And we really think that this is a place that we can have a lot of impact and fix it. And so the second company I worked for had just called and acquired. He was free. He decided to team up and start Querio. It took us like a month to pick the name. And then I don't know if you know Jason Calacanis. He's quite a popular investor on Twitter. He went to Alden's podcast. He was doing these like 25K angel checks for people. He posted about it on Twitter. I DMed him on Twitter. He said apply. He writes us a check like a month after we had decided to start it.

⁓ Now he's our biggest, he's reinvested two more times actually. He's our largest investor in the company. And thank you. And so we had some extra money. ⁓ We didn't take a salary the first year, but we used that money to get a part-time designer and to hire a founding engineer. ⁓ And with that, we got prototype outs over the next few months.

Ilan (14:51)
Congrats.

Rami (15:08)
learned a lot about we can go into all these other startups, but like fundraising, want to do it and fail to get it trying again, getting your first customer, legal stuff, like just things I've never touched, know, things I really, really never touched. I'm a lot of growing pain, but one thing after the other, we had a bit of money, a small team made a product, learn how to get a handful of early customers. And then it kept snowballing.

Ilan (15:17)
Mm-hmm.

That's a really cool story. I think a lot of product managers aspire to start their own companies. What was, I'm curious. So I wanna go into Querio more in a minute, but

what's one thing that you wish you would have known that might have shortcut some of those early learnings that, you maybe something that's not something that we traditionally do as product managers ⁓ that would have helped you in the early days.

Rami (16:04)
more of a product manager problem or more like an Amazon problem.

I wish I had built way less product early, built the ugliest, simplest thing possible, and gone straight to market and just tried to find every possible way to get feedback on the early product because I think about nine months in, we just deleted half the code we had written because it became extremely obvious that this major assumption that we had like wasn't true. And I think obviously product managers shouldn't have assumptions, should validate them.

into like Querio for instance, like that's great. But when you don't have a customer base, you don't have the money to pay for like external research. I think it could be quite easy to track to just like, there is a degree of taste and assumptions you have to do as a founder. When you don't have a customer base to ask and no one to actually try it out and things like that. ⁓ But I think as product managers, like we want to deliver great experiences, right? And we don't want to shift half-assed products that don't feel finished.

⁓ and I think you need to put your ego aside and your standards aside and especially like maybe the first six months and just try to get and this is much easier now with vibe coding for instance, but like just get the most basic flow of the product you're trying to build out to at least get people using it and just even for free just let me know how you feel about this, right? ⁓ And I think yeah, I think that'd be one of the major hopes.

Ilan (17:25)
Mm-hmm.

I think ⁓ that's really interesting. David and I both love this quote from, I'll have to find who it's from, I'll put it in the show notes, but it is, ⁓ if you're not embarrassed by the first version of your product, you launched too And I think both of us come from, well, we come from much smaller companies, neither of us have worked at Amazon. And I think that there's probably a burden of

perfection at Amazon that maybe doesn't exist in a smaller company that allows you to iterate a little bit faster on your products.

Rami (18:07)
Yeah, that's a huge burden for like perfect products at Amazon, the amount of people you're releasing it like staging rollout. I think I just thought about something else. Like the second thing managers, they definitely care about user growth and feature growth and things like don't care much the initial users to come into the product or they don't usually...

play a role in that, especially like in mid-size companies. It's just not where they play a role, right? Like you are working with the customer base that's coming to you, or maybe you're working on someone that's bringing the website traffic and you are making sure the onboarding flow is perfected. I think it's really hard to stop, like as a product manager and you turn it to a founder, you're going to end up doing the least of the work that you love to do and that feels your strongest suit. And figuring out...

how to get revenue and how to get people onto your platform should take up 80 % of your mental bandwidth and the product like 20, if not like 10. That is, especially in the early days, like nothing will solve a startup's problems more than revenue. It is a silver bullet or just traction, like just nothing, even to an embarrassing product, whatever it is. I think it's...

It's important that product managers accept that quickly. And it's going to be annoying. It is annoying.

Ilan (19:28)
Yeah.

Okay. So ⁓ let's, let's use that as a, opportunity to talk a little bit about What is Querio?

Rami (19:39)
So yeah, Querio is an AI business intelligence platform. Primarily a business intelligence platform, which means it helps people pull insights out of the database. Things like Looker or Tableau. And it's AI native, meaning there's AI agents that can actually write code to your database, query data, and also Python to let's say chart it. So it's how do we enable people to, in a self-service manner, generate dashboards, insights, reports, answer their questions from the data that they're collecting at the company every single

Yeah, to give it a bit of color, like if you ever looked at a dashboard on Looker or Tableau or something like that, you probably have to have an engineering team or a data team make it for you. Why? Because most people don't know how to work with databases. They don't have any visual interface. You have to write some code, usually SQL to actually pull that data, where the tables are and what they're called and how to join them. just there's a lot of complexity there. And so it's one of the last kind of functions at a company where 100 % of people need questions and answers from there. And like,

2 % of the company can actually get it. And it's just a humongous bottleneck for so many teams. And they either will run blind or wait a long time or do a lot of manual things. And it's very frustrating. so yeah, Querio is, because of LLM's unlocking an entirely new kind of product category, we can now build a data analytics product that from scratch allows anyone to just ask a question there like they would to a data team. And Querio's AI agents themselves will actually write all the SQL and Python and generate the insights and charts that you're looking for.

in like a minute, two minutes instead of two to six weeks.

Ilan (21:10)
Can you give us an example?

Rami (21:12)
Get ready to get your mind blown, ladies and gentlemen, because we've got Querio this is actually the main page of Querio. We've tried to keep it very simple of just ask for what you want. ⁓ And so I don't think this data is going to relate to most of you, but I you're going to get the gist. So I have a Dunder Mifflin database. Hopefully, you've seen the office. But Dunder Mifflin obviously sells paper products. So we can look at some tables we have here. have the prop dip.

paper products that we sell, the different accounts of the account managers, all the orders we're having, maybe some of the reviews. And, I don't know, maybe as a product manager you want to know who are your top accounts because you want to make sure.

which accounts kind of bought the most amount of product in 2024 by like revenue or something. So I'm going to ask that question. Usually you probably have a dashboard list after some time. Let's say it's early, don't. Querio will take your question, look at what it understands about the database, make a plan for actually answering your question. So here it says it's going to write SQL query. That's great.

talk about the inputs it's going to include, and it's going to make a table. It's going to do that first step. In this case, it's only going to be one step because it's only going to be a SQL query. It's wrote it. It's now running it. ⁓ And so we will now get a response in about a couple seconds. OK, perfect. So the top accounts by revenue, the first one is Heaney LLC for $2.2 million, Oberbrunner for $2.1, et cetera, et et cetera. So we asked one direct question to an AI agent that was directly plugged into our actual database, which wrote code.

ran it, gave back the result, and gave a description, a summary of the actual results, which is super nice. And that took, I think, 29 Versus if you were to do this at a company today, it would probably take you a couple weeks for someone to get to it and actually make it. Yeah. What do you think, Ilan?

Ilan (23:05)
I think it's pretty cool. Can we, as you said, you know, this is maybe one of the first dashboards that you would create as a company, right? Who are our top accounts? So let's dig a level deeper. What's like another maybe discovery question.

Rami (23:21)
Yeah, so let's follow up on this and say maybe ⁓ for the number one customer, what product did they buy the most? Actually, instead for the number one customer, show me their top five purchased products as a bar chart.

So, Querio has kind of contextual understanding of what we've been asking about. I just said number one. It knew who the number one was from before and the account ID for them, as you can see in the reasoning. We use Claude for our AI, if you guys are curious, because they actually let you see the reasoning tokens, whereas OpenAI will not show you what the LLM is thinking. And so, this is really good for kind of digging into what it understands about the database, what could have been getting wrong. Whereas with OpenAI, we just get no visibility, so we just can't really use them, sadly.

So over here it's gonna do two steps. It's actually going to first collect the data we're looking for and then it's going to chart it, write the Python code for that chart. It had an error so it just realizes itself and just rewrites it which is really nice.

we got our chart back, very rich, responsive chart immediately. It looks like the Hammermill copy plus copy paper is one that they bought the most. And so it's a decent distribution, but these top three products really make up like most of the sales. And so, you know, this is where someone, I don't know, my logistics hat or like my manufacturer hat is like, is like, okay, I don't have this data in here, but how much does it cost to make these? How can I the margins on my top selling products? You know, and that'll give me an idea of something to kind of go down.

Ilan (24:48)
Mm-hmm.

Right. That, or you could look at ⁓ maybe low selling products across your, ⁓ your user base or segment those, which is something that I love in Querio.io is I'll often just ask it to make logical segments, like segment this data, make logical segments. And it's super helpful for doing at least a first pass on

you know, what are the core tiles of customers who have a certain feature about them or, you know, certain spend patterns.

Rami (25:26)
Yeah,

spend patterns and maybe like product size mixes or anything like, yeah, exactly.

Ilan (25:33)
Love it. What's the favorite piece of feedback you've gotten about Querio.io recently?

Rami (25:41)
Honestly, it's really nice when it just magically clicks for some people the first week they have it and just like, I just saved seven, 10 hours of my week. It's crazy from either trying to do things myself, struggling with chat fatigue, copy, paste the SQL back and forth, or I had like one old dash, but I had a lot of raw data and then was just trying to manipulate it in Excel just to get the final thing I was looking for, like whatever it is.

but that's my favorite kind of moment.

Ilan (26:06)
Can you...

Can you save these charts? Like what if I wanted to make this my, one of my key dashboards?

Rami (26:13)
Yeah, so I actually have already saved this and let me save the one we had before just for something to work with. So in Querio we can also, you know, we have our save, we have our history, which is everything we've ever done in our workspace, but let's go to boards. So I'll make a new board here.

And I'll keep the one I just saved. Let me grab one that I had before. March 24th by customer. That's fine. Dwight Schrute's sales sure I'll put those three here. We have this like free canvas just to make what we want.

Then I can just click save and now I have like an auto refreshing board that will have up-to-date data every time I log Just like that. No tickets, no JIRA, no back and sub three minutes and you are off. Yeah.

Ilan (26:56)
Okay. So this is maybe a good segue into something that we were talking about just before recording, which is about un AI-ing your product. Right now it is so hype to put AI on everything. We're AI washing entire, ⁓ business segments.

but

What do you, let's talk a little bit. Let's dig into your thoughts there a little bit. And maybe let's start with, you talked about unai-ing querio So what happened? What made you make that decision? And then let's dig a little bit into AI and product management.

Rami (27:40)
Yeah, so I'll showcase something on AI in the product, for instance, and then I can stop sharing my screen and we can go into more details. But I think as a product manager or as a founder with both of the same goals, I want to give my customers the most delightful experience.

And I think it's really easy to be a bit lazy and use a bazooka to destroy a paper and use an LLM for flows that just don't require

For instance, let's say I wanted a line chart. Having to ask an AI to spend

a ton of tokens and a minute to convert this to a line chart seems crazy when you can just have style and just, my chart control isn't working for this one. So obviously that's not going to work. But let's say the text size ⁓ on like this thing over here, imagine you have to ask the AI to change the X axis or like to angle the numbers or like whatever. Like these, these little things that just shouldn't require you to use an AI to change, like even formatting a chart or like moving columns around to the table, like whatever it is.

Ilan (28:32)
Mm-hmm.

Rami (28:36)
When you rely on an AI for everything, make this a line, move this around, whatever it is, it's just, it ends up...

making the customer and the user experience worse because they know how to do that. actually, buttons are a much more efficient way of doing that. They have a problem getting that initial answer in charts. They don't have a problem changing the color or knowing how to switch chart controls or changing labels. That's not their problem. So I think having an AI that's writing the code in Python, you could just have it do all those changes all the time. It's so easy. I have to build no new feature. Because it can do all that by default by the fact that it can make one. It can also edit one.

Ilan (28:50)
Mm-hmm.

Right.

Rami (29:12)
Um, but I think you need to stop and be like, I need to invest the manual engineering time and effort to figure out what's off this 80 20 rule, for instance, maybe an AI, like a 90 10 rule is going to have the most amount of advantage from AI. But then what, what point does AI start degrading the actual outcome that my user is looking

Ilan (29:31)
That's such a good point. mean, as you said there, really, what are we doing as product managers or in your case as founders is we are delivering value to users. are solving problems for users and AI is one tool in our toolkit that allows us to solve problems for users. Another thing that just popped into my head about this use case that you're pointing out with chart controls is

Sometimes if you just give a text box for somebody to prompt, they don't know what they can prompt. They don't know what questions they should ask. Let's hear it.

Rami (30:11)
whole other rant, yeah. Whole other rant, yes, that's very

Ilan (30:16)
let's talk a little bit about AI at Querio.

Obviously, you've built an AI agent. So you're on the bleeding edge here of the technology. You've been doing this for a couple of years now. So really on the bleeding edge. And so I guess I'm wondering.

What's your, personally your philosophy about how your company should be using AI? And then from your perspective, somebody coming from product, is there something that you'd like to highlight that you think is a really cool use of AI for some of the workflows that you guys use at Querio?

Rami (30:58)
Yeah, so I think the philosophy for using AI in the company is the same philosophy I have for what AI features both for our customers, which is focus on your output and then find the best input for it. Don't just toss AI at things. ⁓ One place, for instance, I hate AI is copywriting. It's really good at doing a bunch of filler text. It's really good at summarizing things, helping you avoid like reading too many articles and like summarizing it. But if you're trying to

Ilan (31:25)
Mm-hmm.

Rami (31:27)
produce content for your company that sounds genuine and like me, it will never do that. I'm going to spend more time going back and forth with it than like, sound less AI, don't use dash. They're just not good at actual creative writing, in my opinion. ⁓ But we use AI in a ton of ways in the company. The engineers use co-pilots like Claude Code then Cursor.

Ilan (31:41)
Mm-hmm.

Rami (31:53)
Marketer does use help with some copy things. We also use it for like in Figma for instance, for Booker Designs. We don't have a product functioning yet, like me and my co-founder are the product team. And I know I talking about this a little for the podcast, but we don't actually do the We don't have sprints and design beforehand. We do our kind of product engineering work full base of a book called Shape Up by Basecamp, which is kind of like, so just as you...

never design things, you just actually describe the desired user flow or feature. you don't give how long you want to spend on it, you give how much you're willing to spend on it. So like, only want to spend a week on this feature and that's it. And then see what you can fit in. And so you never say like the UI should be this like a modal or a button, it's like, no, no, this is the required functionality.

Ilan (32:22)
No

Rami (32:38)
one way that AI helps like my product workflow from my product function as a founder, especially for my shapes, like it's really nice to keep them. It's a really good like mental sparring partner to like, okay, what technical risks would I not know about me, but I need like my technical part a bit more. What rabbit holes could this make it so like I only want to spend a week. What features, you know, helping you explore all angles. It's really, really good as like they're genius things. It's really, really good as sparring partner in that sense.

I think for the product managers listening, obviously you can vibe code lot of prototypes and I say prototypes, do not try to make something in production unless you want to delete your production database and lie about it. ⁓ But ⁓ I think for product managers listening, a really good way that this could help you, especially when you're trying to make more of an impact than early PM is if you're having a hard time showcasing the vision you have for a feature or a product ⁓ to your engineers, to your leadership team.

Humans are very visual features. So like making a prototype, on Lovable or Replit can really help sell your feature ideas. It can really help like the engineers kind of understand the feature you're imagining. It can help the designers get inspiration for what you're thinking. Obviously, the best way to get buyers from leadership is to be data-driven in and use Querio But aside from that, I think it'll really help your communication with your engineers and your designers. And I think it's also

really nice. I tell people that Querio will help everyone becomes more data literate. I think that vibe coding tools help a lot of product managers become more like coding literate and seeing how easily things can get messed up and seeing how the little attention detail matters and just notice, okay, oh, it's not just a button that has to exist. It has to be connected to the back end that does something and all these APIs. like, I think I know a lot of product managers are really smart and they're more technical, but I think you probably have like a 50-50 split. And it's just really nice seeing a lot of product managers upskill themselves.

more of like a secondary improvement But yes, from coding to marketing to design to customer success, AI touches all parts of our company. Yeah.

Ilan (34:39)
I

have a, so you've gone down this path and now I have a more like product process question for you. What made you guys decide to ⁓ work in this, in the shape up framework? We'll post a link to the book in the show notes, but it's become, I think the book was released about five years ago and it's become a really popular framework for thinking about developing products.

As you said, it's written by ⁓ one of the head of engineering, one of the co-founders of Basecamp. So what made you guys decide to move in this direction? Have you always done that, or is it something that you changed into?

Rami (35:20)
It's actually relatively new. It's about like maybe four months we've been doing it, three, four months we've been doing it. ⁓ I think a startup's greatest asset is speed. A lot of teams get as good asset is speed. And I think a lot of people love feeling ownership, right? This idea in Basecamp where the engineers and the designers own how something gets delivered and how it looks. It keeps them close to the customer. ⁓ It keeps them like with high agency to design and think about how things should look. It shows high trust.

It moves a lot faster. There's like less kind of, I'm going to think you're just going to be a coding monkey and do kind of thing. ⁓ I think it just really helped us be a much flatter organization and helps move a lot faster. And it helps increase the ownership and agency of everyone in the company. ⁓ And I think that alone just has so much value really, like just everyone kind of being plugged in, being close to the customer.

every single section of the company, knowing what's going on, understanding kind of who our customers are, how they perceive things, why we're doing things. And then also keeping in mind the appetites. like, okay, you're not going to work on this for three weeks. Like, no, no, let's decide how much we're to work on it in the current scope of what the customer really needs. And just delivering that, I think it just keeps us also very accountable in the last two chances of things just overextending. Cause the goal isn't to spend, to just deliver the features. I want to spend a week on this and whatever I can get done in a week, I will get done.

and the most MVP way to solve a problem attempt.

Ilan (36:51)
All right, I have taken up way too much of your time already today. ⁓

Rami (36:57)
You have it,

I'm sorry, it's one of the best pockets I've ever done, it's so fun.

Ilan (37:01)
⁓ This has been really fun for me too. Thank you, Rami.

I'll ask two questions for you. One, if people want to know more about Querio, where should they go?

Rami (37:11)
Our website, querio.ai, that's Q-U-E-R-I-O, it'll also be in the description at bottom of the video. ⁓

Ilan (37:20)
⁓ And where can people find you?

Rami (37:24)
me as Rami or Querio

Ilan (37:25)
You as Rami

Rami (37:27)
I'm on everything. So I'm on LinkedIn as forward slash data Rami. I'm on X as Rami ABH. And I'm always available for a chat if anyone wants. Honestly, you can just guys, you guys can have my personal account. It's calendly.com/ramiabh Tommy, you found me on this podcast and I will gladly have a chat with you about product careers, AI stuff, whatever it is.

Ilan (37:53)
That's amazing. That right there, that's like a huge value for every listener. I hope that you get bombarded with people coming from Prompt and Circumstance and booking some time with you.

Rami (38:03)
Me too. I'm

a very big believer in a rising tide lifts all boats

Ilan (38:07)
100%. Rami, thank you so much for your time.

Rami (38:10)
Thank you man, you're great.