Ascend: Stories of Scale

Balancing Automation and Human Insight in the Age of AI

• AScaleX • Season 1 • Episode 4

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 24:54

Any feedback would mean a lot for us. We hope you continue to enjoy the episodes!

How do you build AI products that solve real problems instead of just adding more features? Yan Grinshtein, founder of Cepien AI, shares his perspective on startup resilience, customer understanding, and building AI tools that deliver meaningful value.

On this episode of Ascend: Stories of Scale, Angelique and DJ sit down with Yan Grinshtein, founder of Cepien AI, to discuss building startups in today’s AI-driven world. 

From resilience and founder mindset to AI-powered research, the conversation explores what it takes to create products people truly need.

Yan also shares why understanding real customer needs is a key competitive advantage, and how Cepien AI is transforming research synthesis by turning weeks of analysis into minutes while keeping human insight at the center.


Yahn Grinshtein

CEO/Co-Founder, Cepien AI

Yan is a global design leader, startup builder, and entrepreneur whose journey spans immigration, homelessness, and multiple successful startup exits. As the first designer at four startups, he helped shape products and systems that scaled globally and contributed to lasting company growth.

Now serving as CEO/Co-Founder, Yan brings a grounded perspective on leadership, design, resilience, and building products that genuinely matter.

Thank you for listening! This podcast is a passion project of AScaleX. Know more about us through the links below:

SPEAKER_02

Funding does not necessarily translate always to people or right people. You can throw money at the problem, but the problem is not gonna get solved just because you throw money at it.

SPEAKER_01

Growth stories often start with funding rounds and product launches, but sometimes they begin with survival.

SPEAKER_00

In this episode, Jan Greenstein, founder of Sapien AI, shares how personal resilience shapes his founder conviction and why insight and not just capital is the real growth bottleneck for modern product teams. Thanks, Jan, for joining us.

SPEAKER_01

Thanks for having me.

SPEAKER_00

Welcome to Ascend, Stories of Scale. I'm Anjalique.

SPEAKER_01

And I'm DJ, Summon for Tracy.

SPEAKER_00

Yan's journey isn't conventional. Yan, how did instability shape your risk tolerance as a founder? Maybe we can start with that.

SPEAKER_02

I think it gets you to a lot higher risk tolerance, essentially. I mean, it it depends, right? So when you have a stable job and you have nice income coming in, right, you you feel more safe. And obviously, coming from homelessness, that's that's kind of the direction you want to get to eventually. And so when making a decision to start a company definitely was a very long kind of deliberation. Should I risk that stability and comfort level? But once you're once you're in it, understanding that building a startup is to some degree is not as harsh as living on the streets. So it's you know, you're like, meh, all right, I got it.

SPEAKER_00

Sounds good. But can you tell us about your story and a little bit backstory about yourself?

SPEAKER_02

How far do you want me to take this?

SPEAKER_00

As far as you want, and which part you you're um very comfortable sharing?

SPEAKER_02

Uh I'm comfortable sharing with anything. Like I'm not I'm an open book, I don't I don't mind sharing with people. I've been doing this for quite some time now. I was born in a scary place, as people, a lot of people think or may think, Soviet Union, uh, Ukraine, but at the time was still part of the Soviet Union. Uh I grew up in Israel, um, and then uh I did military service, and then right after that I moved to Canada. I was in Canada for six and a half years. That's when I uh went to college. I kind of started my career as a as a designer, if you will, or try to at least. And then I found myself in New York in 2006, and I don't want to go into too many details, but if anyone wants to read the details, they can always find my uh blog post, an article about my homelessness. But I found myself on the streets. That's the kind of like the backstory of how I got to that place. But then within the first, I would say seven to eight months, one of the largest New York State uh law firms were my clients. Honestly, lucky, I guess, I don't know, but an incredibly successful design career across agencies, across different countries and different continents. Eventually, right after being in agencies for about a decade, I decided to stop because I I burnt out. Agency of life is not for everyone, and I'm an ambitious person, and so I really wanted to be an art director at a certain point, and so I really I got there super fast. Uh I was, I think, at least at the time, I did not know any art director that was 27 years old, and I was. I burnt myself to the ground, quit the agency life, uh, took some time off, and then moved back to New York in 2014 and went working for startups, and then spent the last decade plus building startups alongside founders. Uh, first design hire at five startups in the last decade. Three of them got acquired.

SPEAKER_00

Sounds good. Why sapien?

SPEAKER_02

It didn't start as sapien, it started as inside. So when when I was starting the company, I wanted to be clever. And coming from an agency world where I used to work on branding, like there's a company, there's a multi-billion dollar company named Tabula today. I was the first one to work on their branding when they just raised their first round year that decades ago. And so I wanted to be, you know, something that it's not as simple as Insight, but at the same time representing the core business, if you will, or core technology that we're building, or core uh solution. And so I went with Insight. The official name of our company is Insight Intelligence. Now, what we realized when we launched the first MVP and people started using it, but we realized the name had a problem. The problem was A, autocorrect. Autocorrect would always correct it either to insight or night. We had a lot of quote-unquote miss opportunities for people trying to find us because people don't type in domains anymore. People go to Google and they just search for things. When you get autocorrected, then you realize that okay, I need to do something about it. And also people misunderstood the name. I would say mid-25, when we realized this problem, we went on a rebranding and renaming journey where we explored slightly over 1200 different names for what we're building. And eventually we landed on Spient for several reasons. One is the fact that it represents the mix of human and artificial intelligence because Homo sapiens essentially represents the wisdom, the thought process, the fact that we're human, we can think for ourselves. And so the next evolution of what we consider of intelligence is some sort of a mixture between human know-how and uh artificial intelligence. So the logo that you see on my head here, the little thing that is coming out, that's what we kind of see as a breakthrough of the next generation or next level of intelligence. So that's why Sepien.

SPEAKER_01

And going back to just some of your journey, you lived through a lot of uncertainty, moving a lot. How did that change how you approach product decisions?

SPEAKER_02

Moving really gives you starting over, right? I think that's that's the key key point of immigration, is you're essentially starting over from almost scratch. And you have to rebuild everything from almost scratch, right? And so I think that portion essentially creates a different mindset in you as a human where problem solving is not as of a cognitive overload anymore. I've learned, and I and I observe this in other immigrants as well, is that you learn to find diamonds in the crap. Like whatever crap is happening to you, you learn how to derive some diamonds out of it. Case in point, when we launched our MVP to public, we did like a soft launch to the public, not private beta. We took it out of private beta for a little bit to see what will happen. And it was a flop. It was a massive flop. To other people and other founders, that will be considered a massive failure. You spent months working on the product, you have validated the product with your potential ICPs. You have companies on the platform, you have companies paying you for being on the platform, and you lunch to the public and you have less than 1% conversion rate. That's a massive punch on the gut. It was a punch in the gut, and it took me exactly 15 minutes to realize that this failure is actually our success moment. And I went back to my team and I said, we just succeeded through failure. And here's why. We just got the most, the cheapest version of incredible amount of lessons that anyone can get in the shortest period of time. We rebuilt our product, we redesigned the entire system, we rebranded and renamed our product, we reintroduced our solution, we reconsidered our entire messaging platform and single failure.

SPEAKER_01

Absolutely. I I loved hearing that because they often say you need to fail fast so you can learn, and I can see how your background having to move and start over, figure things out, be a problem solver, how that translates into you being a founder now.

SPEAKER_00

Let's talk about something we see across growth uh stage companies. Why do companies with funding still uh ship blindly? Why do you think and also do you think that's a deal breaker for a lot of products today that they're doing that versus having funding?

SPEAKER_02

Funding is not necessarily a the magic dust that will make your product successful or not. Look at OpenClaw. OpenClaw went public became a big thing, and the guy was just burning 20k a month with no funding. I mean, granted, he's a wealthy guy, he sold a company previously for $100 million, so it wasn't a big deal, but at the same time, making something successful is not about funding, it's about understanding the core need of your user or potential user. The reason, as you ask, like why so many companies are still shipping without truly understanding their user, it's because they're rushing. They're rushing or they don't have enough resources or they have enough people. Funding does not necessarily translate always to people or right people. You can throw money at the problem, but the problem is not going to get solved just because you throw money at it. Sometimes what I found in my experience working at startups for so many years is that you can solve a problem with a small team if you knowing if you really know how to hone and understand the core needs of your users. One of the examples that I can bring is I I was working at a relationship tech company some years ago. When I joined the company, they had a huge problem. They had a massive amount of downloads on the App Store, but about 20% completion rate. Which means people would create an account and only 20% would actually go through the entire onboarding process. That's a huge loss. They spent more money on marketing, communication, explaining the product, on education materials, on QA's. They spun all this money and yet nothing changed. What changed when I joined, I reshuffled the design team, I hired a little bit a few more people to the design team, and we literally took six months to basically roll everything back. It's essentially truly diving deep to understand what are the needs of the user, not just once. The difference between understanding what users want and need, there's a huge gap there.

SPEAKER_01

How did you gather that information? Was it through interviewing your audience? Was it through sources on the uh internet?

SPEAKER_02

In order to do this correctly, most companies are still relying on the same process. At large, it'll remain the same for the last three decades. And that is a what you just mentioned. You rely yourself mostly on user feedback. User feedback is typically broken down into two parts. One is the one that you're collecting as a designer product team, where you go out and you talk to users. You interview them, you ask them about their problems, you're asking about their needs, you give them your product or your wireframes to test out and see how they respond to it. This is what we call research. The other part of it is user feedback side of it, is user complaining, user asking. These are typically coming in through surveys, NPS scores, user complaints through ticketing systems like intercom, Zendesk, or you name it. All of these things are being collected and looked at, and like, okay, we have X amount of users complaining about this. Let's solve this. We have X amount of users asking for X features, then we have to build these features. The problem is, oftentimes users don't really know what exactly they need. They just, in their limited amount of knowledge of technology or abilities, they would say, I want that. As the famous quote, again, I don't know if it's a real one or around, it comes from Henry Ford. If I would have asked my users what they wanted, they would have said a carriage with three horses instead of two and not a car. Right? iPhone 1 was not born because people were asking, I want a screen that I can tap on. iPhone 1 was born because Steve Jobs and his team realized the core need of the user is not about having more buttons, it's about having a ability to complete specific tasks in a more delightful way. And so that's how they figured out the screen is gonna be a lot better. And it was. Users didn't woke up one day and said, you know what? I wish there would have been a device that had a big screen that I can top out. They didn't. At the time, they would wake up every morning and think to themselves, I wish my BlackBerry would have an extra button so I can do X, Y, and Z, or maybe it will be faster. That's it. And that's why most of the phone companies were racing to do that. Because they were listening to what their users wanted, but they never bothered to understand what their users actually needed.

SPEAKER_00

From what I've read also, Sapin reduces UX research from 40 hours to 10 minutes. I mean, that's not incremental, that's exceptional. How does that faster insight change iteration cycles and you bring it like 10x basically?

SPEAKER_02

We don't replace the actual research. There are some companies out there that are trying to build AI that will go and conduct user interviews and things like that for you. I am still very skeptical of how accurate it will be because one of the things that I've learned building AI before ChatGPT even became a thing is that AI, even in today's development and uh its level, still does not know how to read between the lines. A good UX researcher knows how to read between the lines. It's not about what people tell you in the interview, it's about what they don't tell you in the interview. That's the biggest point that makes a research successful or not. What we automate, we automate the entire post-research synthesis process. So typically the way this process works is I'm a researcher, I'm a UX, let's say I'm a UX researcher, and I need to understand something. And so what I do typically, I would go and look at our intercom. Maybe I'll go and look at our Zen desk. I'll look at my NPS scores, I'll look at some of the surveys that I've conducted, then I'll go out and they do user interviews. I'll actually talk like we're talking right now. I'll have a transcript, something will transcribe it. It's gonna be a whole transcript that I'll put somewhere like DAFTAL or Aurelius or whatnot, uh UX research data repositories. And then I'll go in and I'll spend some time going through all this piece of data, all of the transcripts, all of the conversations in Zendesk or Intercom or all of my surveys, and I try to highlight or tag findings. Pain points, requests, needs, annoyance, distrust, the list goes on. Now from there, once I've done all of that, I'll try to see what are the trending things. Like if you're working in Dovetail, for instance, you can put it in uh charts and it will show you how many pain points you discovered and and whatnot. Right. From there, you start collecting other pieces of data like behavioral analytics or whatnot, and try to see how this collides together. You try to find the trends. Once you find the trends, from there you can derive typically one, two, maybe three insights that are actually valuable. And based on that, you'll make quote-unquote recommendations. I found these three insights. Here's my three recommendations, what we need to do in order to solve for these insights. That is being turned into an actual action. A Jira ticket or a few or some along those lines to actually go and execute and build and ship. Once you build and ship, you take it out there, it's being used, you test it, making sure that you actually understood fully the need, understood the desire, understood the insight, and actually solved the problem. Now this entire process, on average, can take with AI today, we've learned in our studies, can take around 40 hours. It used to take between two to three weeks to a team, not just one person. Our system essentially does that entire process, like that entire piece boils it down to 10 minutes. We we say 10 minutes, but in reality, it takes about 10 minutes when you just set up the platform and it starts ingesting and doing everything. But uh after that, it's it's mostly real time.

SPEAKER_00

Yeah, that's amazing. Great, and thanks for helping us understand the product a little bit more. You also mentioned this during intro call, Yan, that you will be launching this Q1 in terms of removing sort of blockers for people to test out your product. Why remove the onboarding friction entirely? What are you seeing there in the market right now that you felt like you have to ship this uh to help them and remove that barrier?

SPEAKER_02

Every product in in our quote unquote category, they're all still acting like 2022 SaaS environment. You go to their platform, you go to their website, you look at their promise, you look at some of the messaging, you maybe look at some pictures of their of their product, maybe a video or whatnot. And then you do either you sign up for a free trial or and you try the product or you book a demo. And somebody like myself will get on the call with you, walk you through the product, show you how it does things, show you how to set it up, and and so forth. What we decided to do, we wanted to change this narrative. We wanted to change how people experience these type of products, especially ours. And so what we changed there is we have made our product so easy and simple that you don't need to get on the call with anyone to understand how to use it. The playground, as we call it playground that we decided to put out there, we just want people to experience what it looks like to have AI turn your data into an actual, actionable and valuable tagged user issues and insights and recommendations. Now, obviously it's limited. We don't have endless pockets to fund all of this, right? But we have a little bit, so we can do that. The playground will not even require you to create an account. You don't need to create an account.

SPEAKER_00

So you can just literally go in there and start using the product.

SPEAKER_02

You're starting to use what we call a mini agent. So our entire platform is a lot more robust than what we put out there for free for people. Now, that one single piece will allow you to just see how easy it is to turn your mass amount of data. And again, we're still limiting that. You you won't be able to upload a document, a data CSV file long more than I think five or ten mega megabytes per session. We'll not give you all of your user issues, so all of your insights, because again, we're limiting because it's our burn. We're paying for all the APIs for all the data processing, all the AI tokens, like we're paying for that for people to experience, right? And so, but within minutes of you interacting with it, and it's designed in such a way that it will guide you through the process. You don't need to fully understand what to do there because it will keep asking you questions, and while you're answering the questions, it builds its confidence scoring. Then it will ask you for the data, it will ask you for your ICP document, it will present you product and business goals for you to select, which goals you have. Based on your unique goals, it will present you your insights. And you can just literally see what will be the impact on your business goals, what will be the impact on your product goals, what will be the impact on your usability goals with a specific insight. Free of charge, no account necessary. Just go ahead, give it a try. So we're technically launching uh at the end of this month.

SPEAKER_00

So you're launching, and we will be very eager to see the initial results that you'll be getting in that approach because uh I I think it's very interesting to see a lot of founders, at least in the AI world, figuring out the removing the entry of any barriers of entry for users. Because to your point, it's actually easier for people to play around with the tool, but then you have to spend money on tokens and stuff, so you also have to limit it. But you also reduce costs because you don't need to onboard people to do so, and it works both ways. So that's very interesting. And then I've also had conversation about at least from AI founders talking about their pricing model. It's not per seat anymore, it's based on usage and stuff like that. So it's really interesting to see how even the consumption, how people consume products today is changing. It's not just how we're retooling ourselves for what's available out there, but even how we consume these products is changing. So that's very interesting to see in the market.

SPEAKER_02

Everything is changing. When we launched, we launched with the traditional SaaS pricing approach where you basically pay us per seat, we charge for that. But then we also realize that charging per seat is not going to be scalable.

SPEAKER_01

And zooming out, what does sustainable growth look like in you know an agent driven world now?

SPEAKER_02

Some people would look at companies like Lovable and Coursor and whatnot and how they're rapidly growing. But I think sustainable. Sustainable growth will come from finding the right balance between user consumption, right? How they use the products and what they consume, your pricing model, how much you charge so it's not too much, and it's still giving you quite a bit of profit, the balance of tech. And I think what a lot of companies are starting to realize that in order to create this balance, you're probably gonna have to go the route of in-housing your own LLMs. We're doing that. Like we're we have our own LLM that we're training, and I I think we'll see a lot more of companies doing that, especially in the larger size companies.

SPEAKER_00

Sounds good. I think this conversation is really interesting because it just not shows the product that you guys are putting together, but also the shifts that we're seeing in entry point, pricing, positioning, and everybody's still early on. Everybody that I've spoken to, we're all testing, we're all trying to break things and learn fast, right? So I think that's very key in the um market that we're in today. So thanks Jan for sharing that. We'll probably just share more of your findings to us in the next few months here as you roll out your product.

SPEAKER_02

Yeah, I think the the entire industry is still learning and figuring out how to deploy AI, how to use it, what capacity. And I don't think we'll see an actual like, yes, this is how, this is where, this is why, probably until 28, give or take.

SPEAKER_00

That's a good prediction. So we'll definitely look at it. Don't hold me to it. Everybody because I think last year is everybody's like, okay, we have AI, and everybody's talking about how do we implement this, how do we actually embrace this technology in our company and stuff like that. This year is more on execution, at least from what I've seen. And I think a lot of people will try to jump very fast and have mistakes. Some people will probably do it gradually in a safe mode and then try to like augment their workforce and stuff like that. So it really depends, and then we will definitely see results uh by December of this year, how people did it, right? So I think that'll be interesting conversations to have. As we go through it, like I think we'll become understanding of what we're really dealing with, but I think we're barely scratching the surface here.

SPEAKER_01

If survival builds resilience and insights is building velocity, maybe the next era of growth isn't about raising more, it's about learning faster. Yan, thank you for sharing your story and your vision. Thank you. This was powerful.

SPEAKER_00

And to everyone listening, this is a sense stories of scale where growth is never linear but always intentional. Thank you guys for joining us. Thank you, Yan.

SPEAKER_02

Thank you for having me.