Podcast: Applying AI to human intimacy (with Offer Yehudai)

On this week’s episode of the podcast, I am joined by Offer Yehudai, the co-founder of Arya, to discuss the intersection of artificial intelligence and human intimacy. We explore the development of AI-driven companions designed to improve relationship health and the delicate balance between automation and human expertise. Among other things, we discuss:

• Whether the lack of human judgment in AI interactions encourages users to share more intimate personal details
• Why a hybrid model combining automated intelligence with human oversight remains necessary for maintaining consumer trust in sensitive topics like intimacy
• What role long-term contextual memory plays in developing sophisticated AI tools that truly understand complex human dynamics
• If the current explosion of low-quality AI applications will lead to a permanent erosion of consumer confidence
• How utility app developers can leverage community roll-ups to achieve sustainable growth in a competitive advertising market
• When the integration of commerce and physical products will become a standard component of digital wellness subscriptions

Thanks to the sponsors of this week’s episode of the Mobile Dev Memo podcast:

• ⁠⁠INCRMNTAL⁠⁠⁠. True attribution measures incrementality, always on.
• Xsolla⁠. With the Xsolla Web Shop, you can create a direct storefront, cut fees down to as low as 5%, and keep players engaged with bundles, rewards, and analytics.
• ⁠Branch⁠. Branch is an AI-powered MMP, connecting every paid, owned, and organic touchpoint so growth teams can see exactly where to put their dollars to bring users in the door and keep them coming back

Interested in sponsoring the Mobile Dev Memo podcast? Contact Mobile Dev Memo advertising.

The Mobile Dev Memo podcast is available on:

• Spotify
• YouTube
• Apple Podcasts

Transcript

Eric Seufert: Welcome to the Mobile Dev Memo podcast. I am your host, Eric Seufert, and I am joined today by Offer Yehudai. Offer, welcome back to the podcast.

Offer Yehudai: Hi Eric, great to be here again.

ES: We had you on about 18 months ago, and a lot has happened with you and Arya, your company. Let’s start with you reintroducing yourself to the audience and telling us about the exciting developments that have happened with Arya.

OY: I previously sold Fyber to DT and about four years ago, together with Tomer, my co-founder, we started Arya, which is an AI couples companion for all life stages, starting with intimacy and relationship health. Think of us as Samantha from Sex and the City who knows you and your partner and makes sure you have a good relationship with support, content recommendations, and commerce, shipping you what you need to bring those recommendations to life.

ES: And you recently raised a round?

OY: Yes. We have raised to date about $17 million in equity and recently took a cohort finance from Bitkraft, which is a great tool I recommend for consumer founders. If you have healthy cohorts and you do not want to give away more equity from your cap table, finding a good partner like Bitkraft to finance cohorts is a great way to grow.

ES: Healthy couples, healthy cohorts. I do need to record an episode dedicated to cohort financing because it is fascinating how the entire VC paradigm for mobile apps, particularly in gaming but even with utility apps as with Arya, has shifted. There is a new complement added, which is non-dilutive, and it has become a big category.

But that is not for today. I want to talk about Arya and applying AI to intimacy. When I had you on the last time, we talked about the origins of Arya. It has been several years now that you have been operating the company. It has been successful and is growing. Talk to me about what you have learned to date. What is the core insight that convinced you that intimacy issues are solvable with AI, and how has that changed? AI as the dominant topic of conversation is an 18-month phenomenon, but you started Arya before that. How has the journey of the company changed in this new environment, and what is the core insight you are operating against today?

OY: Arya is really my own personal story. I am happily married with a great wife and four kids. We are a great couple. I work out, my wife is a yoga teacher, and we know how to build a healthy routine for our lives. We have a smart mattress that yells at me if I drink too much. But when it comes to intimacy, love, and relationship, you are on your own. Four years ago, my wife and I wanted to build something that can really help us and many other couples. That is how Arya came to life.

Back then, AI was not the story. But here is what we have learned. Stigma around almost everything in life has collapsed in the last 10 years. Think about mental wellness. No one talked about their anxiety or depression. You did not see athletes on billboards in the subway talking about their mental wellness challenges and how they are coping with them. That changed. Amazing apps like Flo really took the stigma away from women’s health and period tracking, and you have seen billboards about that as well. The same goes for fitness, nutrition, obesity, and GLP-1.

Intimacy was the last frontier, the last taboo that no one talked about, but it is changing now. We as a society are now talking about how Millennials and Gen Zs are feeling lonely even within a relationship. Divorce is on the rise, people live in a sexless relationship, and Gen Z are more lonely than ever. The time is now, and AI allows us to take all the knowledge that we have and make it more accessible to many people.

ES: Do you feel like AI has accelerated a lot of stuff? I think in many of those conclusions, people are attributing too much to AI, but I think a lot of this does have to do with AI because people are probably more open to sharing personal information with an AI tool than they are with a psychologist or therapist because there is no human judgment. Is that the case? You mentioned that the stigmas have collapsed, and that is probably because there is no barrier to discussing this with an AI system. The reality is, even if there is a licensed professional on the other side of that conversation, they might be judging you or you might still be embarrassed. There is nothing to be embarrassed about if you are talking to a chatbot. Do you think that is the case for some of these topics where AI has achieved what people are stating that it has achieved?

OY: AI is great, but AI is not the reason that everything has changed. It is more about a cultural phenomenon and how we as people understand stuff. I will be honest, at Arya, we are part AI and part human. I do not think that will change, not because of the technology, but because our audience is split. Some people feel more comfortable knowing it is AI on the other end, and there are people who still prefer to talk to a real person.

The real way to do it is actually to blend them together. There are answers or content recommendations where AI can do an amazing job that will work for all audiences. For instance, my wife and I can consume the same content like body confidence or how to initiate, and we will just consume the same essence differently while AI will do the work. But I want to know that there is a human on the other end that has made that recommendation. Even if it is not a one-to-one chat with a real person, I want to know there is an intimacy coach or a certified sexologist that made that happen for me.

We have couples all around the country. When you go out from LinkedIn and Twitter into the real life of people in Nebraska or Colorado, you have to speak the language. Arya has a very small team of about 15 members. The concierge team has about two humans, and we support tens of thousands of users. It is not one-to-one coaching, but you do have to have that part human in the loop. That may be counterintuitive.

ES: Has there been any resistance from consumers about any of the AI components of Arya?

OY: 100%. You definitely have people who say they are not comfortable speaking to a bot about their love life or about their husband. What we found out is how to interface. How do you make sure they understand that there is a real person? Sometimes they will get a faster answer because the AI can make it more accessible for them, but this is not the ChatGPT experience for the user.

You know the feeling when you talk to a chat and it is always on your side. It is like going to a bar with your friend and talking about your life, and no matter what, your friend will support you and say you are right. In real life, if you go to a coach or a sexologist, they will put a mirror in front of you and tell you your point and that there is more to see here. That is what we try to do by being part human and not relying on just the public LLMs and the experience you get there.

ES: The sycophancy issue is a relic of this early phase. We are probably two years into the consumer acclimation to chatbots. There was a funny tweet I saw once that said just remember that the dumbest person you know is seeing this every day, and it was a ChatGPT response saying that is a great insight and you are right. I imagine with relationship issues, that is a very dangerous potential.

OY: AI lacks intuition and EQ. It may be great in coding, complex engineering, and design. But Eric, when your wife comes back home from work, you know in a split second how her day was because you know her. It might be a subtle way she puts her stuff down. When you are at work and you look at your colleagues or in a business meeting, you get signals beyond just text. If you have good intuition and your EQ is in the right place, you can feel it.

That is a very important part that applies to relationship and intimacy that AI lacks. That is where the part human is so important—that intuition and that split second that can make a good coding design great or turn it into very bad relationship advice.

ES: I woke up this morning and saw a profile of a woman I used to work with in the Wall Street Journal. It was an article about people who have started seven-figure businesses as solo founders. She makes the point at the end of this profile that it is easy to build this stuff, but people overestimate how much impact just building the thing had versus how much impact her own experience, intuition, and ability to distribute, market, and position it had.

I think a lot of these tools just make productive people very productive. But all of that developed experience and the whole framework for thinking about these problems and how to actually accomplish tasks in that domain is critical infrastructure for success with any of these products. Without that domain knowledge and expertise, you are just pointing a chatbot at anything. Particularly with stuff that relates to intimacy and is more personal, it is just going to fail. You truly need that foundation of domain expertise with a topic like this.

OY: We said AI lacks intuition, but there is also the matter of accountability—not just responsibility, but accountability. If I am working with a couple and two partners separately in a relationship, I want to make them happy. We are not a therapy product; Arya is a wellness product for great couples who want to become even better. But there is still this accountability part where I want to make sure they are doing what they are doing and that they are following their journey. It is one thing to complete a very complex engineering project versus being accountable for a couple’s love life relationship. From there, you expand to so many other parts of life.

ES: You know more than anyone that the App Store is rife with scam apps, despite what Apple says. If you lose consumer trust, you are toast. There is no coming back from that. The failure state is one article getting out about some horrible experience or someone getting scammed or any number of things you read about happening with AI apps now where the company did not take that seriously. How do you think about trust and curating that relationship with the consumer, ensuring that they feel comfortable and that trust never breaks?

OY: Trust is key in everything. When you build a consumer brand in such a sensitive topic like relationship and intimacy, it is really about giving your customers control. It is one thing to tell them what you think as the concierge, which is how we call our intimacy coach, and then making sure they understand that they have control and they can change. This is not a rigid journey that you have been put in.

Because we have this real-life commerce platform, the concierge can say, let’s work about your body confidence, and here is a great exercise. I am going to ship you a couple of products to try because I think it is going to help. If that first experience works well and I try something that I would never try myself, the trust level will jump dramatically. This service has opened up something completely new for me.

At the same time, if that first experience does not hit, you will say to forget this and they do not get me. I will share that one of the biggest reasons for churn is when our algorithm or the concierge is making the wrong recommendation. If we did not get you and your partner and we recommend the wrong thing, you might give us a second chance, but I just lost a lot of points in your trust level.

In the age of AI and people shipping apps like crazy where the barrier of entry is non-existent, I want to beg to differ. It was never about the tech. It was never about compiling an app and publishing to the App Store. The barrier to entry is to really understand your customer, speak their language, and make the right recommendation. That is how you gain trust. Trust is really about control and understanding your users, along with the foundation of keeping the data and making sure they understand that you are never going to share this with any third party. Our concierge is one-to-one, so whatever I tell my concierge, my partner will not know unless I give permission. It is about respecting and giving control to customers.

ES: You remember in the early days of AdTech when playable ads were introduced and got into the mainstream. We know today that playable ads are a very powerful ad unit, but in the early days, you had playable ads that were not properly labeled and people got really furious. You had playable ads that do not really represent the actual game. You play a version of the game and then the game is actually something quite different. Going back to trust, it is about how your actual experience is and if I have control over everything. You might increase your top-line metrics, but the bottom of the funnel just will not work if you do not have that level of trust. You are basically borrowing against the future and deteriorating the whole foundation of DAU accumulation.

OY: You will pay a price in retention.

ES: What is the biggest misconception people have about how consumers interact with your product, and what surprises you most about the data?

OY: Often I hear people say that after one month of Arya, they can just do it themselves. Why would consumers stay and keep using the product after you gave them the juice? It really makes me laugh. I have a personal trainer and I really know all the exercises by heart, but it is really about the accountability and pushing me harder. I know the drill, but the same goes for your therapist or actually every hobby you may have.

I think it is funny when people say maybe after a month your retention drops, but the data says otherwise. People stay with Arya, which is about a $60 a month membership. It is not a free or a very cheap one to forget about. You see it in your monthly statements. People stay with Arya for a very long time, and our LTVs are in the hundreds of dollars. It is really about building a habit and that still surprises me that people think you can just solve it by giving them the recipe.

ES: How do you measure success? Is it equal engagement from both people, or how do you engage with the customer? As a consumer founder, Tomer and I have defined the North Star for the company, which is our CAC payback. At the end of the day, you want to have healthy cohorts. CAC may change based on the plans, but how quickly can you break even on each cohort? That is the North Star. But I want to answer from a different angle as the founder and CEO of Arya. Our head of science, Nicolas, who is a grad from Washington University in Seattle, conducted the first academic research at the end of last year on users who have actually been on the Arya platform. We were trying to find statistically significant proof that we are actually making an impact on people’s lives.

Even if our CAC payback is amazing and cohorts compound, I want to know if Arya actually works. We have found that people who use Arya for more than six weeks report two very important things. Number one is feeling more connected than before. That word connected in the research is more about communication and closeness. So many couples in the US report feeling more like partners than lovers after several years together running a household. People are feeling more connected and people are feeling more intimately satisfied. That is both for emotional and physical intimacy.

It was very important for us to see the data and make sure we actually make an impact. It is really about breaking the routine. Routine gives us comfort and safety, but routine also kills passion and makes us feel like partners running a household and not like lovers anymore. If you have this accountability buddy that makes sure that you break the routine in the way that works for you and your partner and gives you the tools and takes the mental load away by shipping you all the products you need, people will feel more connected and more satisfied. Making people happy is really the thing.

ES: How do you weight feedback from consumers against the metrics? I feel like there is probably a qualitative component here that is important to recognize.

OY: You always have to look for your super obsessed audience and see how they talk about the product, and also the ones who are leaving you and understand why they have left. When people leave, it is because we gave them the wrong advice or they really wanted it but their partner was not on board. That is the second reason why people would leave a platform like Arya. For the ones that are staying and obsessed, I get an anonymous digest of what people are sharing with the concierge. It is really amazing how it is changing people’s lives.

It is tied to retention. There is no free version of Arya. There is no freemium. You wouldn’t pay 60 bucks a month if you were not happy. We also track engagement retention. After about month four, it nicely plateaus. People in their ninth month or 12th month are as active and engaged as the ones on month three or four. People actually use Arya all the time. It becomes a daily or a weekly habit for couples who stay.

ES: Long-term context and personalization are very important. How much value comes from long-term memory versus just having a more capable foundation model for the applications that use that? You have this balance of AI capabilities and the human-curated and human-augmented tasks. How do long-term memory and context play a role here?

OY: AI lacks intuition, and that is also very true here. Yes, you have context, and intuition is so important, even the way people express or give feedback on their Arya experience. It is really important to understand how people talk about their experience and what they are willing to share. When you join Arya, there is an onboarding questionnaire and the concierge will ask you a bunch of questions.

There is a gap between how people answer about themselves and then how they actually act six or eight weeks later. There is a version of what or how we want to think about ourselves, and that is what we type in the answer, and then there is reality. If you are an LLM, I will take those answers and treat them as is. That is truth because the consumer just told me they want to feel more connected. When you have more intuition and EQ, you understand that this is what they aspire to be. That is not necessarily how they are right now.

Context here, and especially in a text-based interface, is very limited. That is something we also face with Arya because we do not offer video sessions or audio sessions. We are finding ways to overcome those limitations, but they do exist. Let’s look at their actual feedback on their experience and what products they have actually used and what the partner may have said about it. Bringing all of this together really helps. It does not really matter to us if the foundation models get very different or better; it is really about our training data and what we understand.

ES: How do you balance giving users what they want with giving them advice that is actually in their long-term interest? From a product perspective, how do you tune that?

OY: You do not just plug into an OpenAI API. You do not want to have your ChatGPT version. When we started about four years ago, that was before the spike of chat, so the very first version of Arya was not AI-driven the way it is today. But we have more than three years of actual unique training data. When you interface with Arya, you do not feel like you are chatting with ChatGPT. That is really important. We cannot just tell you that you are right or give you exactly what you want.

I can’t bench what I want right now; I need to get there. It is a journey. The concierge will put you on a journey to achieve your goals; it will not tell you what you want to hear. Go to ChatGPT if that is what you want. For us, it is really important to set the right expectations. If we do not do that, we will lose trust very quickly and they will churn. Having that unique training data and really keeping our honest voice here is critical.

ES: What have you personally learned about human relationships from building the company that was new to you or you might not have believed beforehand?

OY: There is a sentence that Tomer uses a lot. Arya is like cilantro; you either love it or hate it. It is not for everybody. What I have learned is how surprisingly quickly people open up once I bring Arya up. I have met partners in great VCs, and people immediately open up and start sharing their own experience. It is a topic that, if you give it the proper venue, people will share so much about their personal life. It is also true when we hire people. People who join Arya are really obsessed about what we do, and it is not for everybody. We have had people who said Arya was not for them. But when people open up and share, it is amazing.

ES: Five years from now, everything goes right. What does Arya look like, and how do you think the consumer behavior has changed?

OY: Here is the secret sauce. Here is what we believe will happen in five years. We have developed and are building three very important components. One is the concierge and the trust with the concierge. Having that trusted companion that I can talk to about intimacy means I can talk about other topics.

Second is the content recommendation—really understanding the content and giving you advice in the way I want to consume it. Then commerce—make my life easy and just ship me what I need. I do not have time or capacity to go and research and find whatever you just told me to buy. Those three elements together are what made Arya successful so far with intimacy and relationship.

Now we are looking at other life stages. My wife and I went through several IVF cycles. Luckily, it was eventually successful and we have four kids, but when you go through that, that takes a toll on your love life and intimacy. Even when everything goes great, the first kid or any kid really changes the dynamics between partners. There are so many different life stages that we as couples are going through and we need help. We need a trusted companion, we need some content, and someone that will just ship me what I need because I do not have time for this.

In five years’ time, it is really about how we expand and support couples wherever they need. There is still the base of the pyramid: your relationship with your partner, your love life, and your intimacy will always be on the base of the pyramid. But on top of that, so many layers like postpartum, infertility, perimenopause, and menopause—all those stages impact relationship and we can take the assets we have built to make couples’ lives radically easier.

ES: I want to shift gears a little bit. I want to talk about scaling utility apps in 2026. My state-of-play assessment is that you have seen utility apps become the dominant advertisers in mobile. That was not the case three years ago. Up until about three years ago, games were far and away the majority. Now it is apps, the subscription model dominates, and web-to-app flows dominate. You just get a lot more opportunity to optimize on the web in the pre-subscription flow. Talk to me about the state of growing a utility app in 2026, because it is pretty fascinating to see how that has evolved.

OY: Tomer and I really believe from day one in a hard paywall—start monetizing super early, with no free trials. When you have a free trial, you start optimizing for the wrong users and the wrong metrics. Hard paywall and monetize early, and set the right price. Do not be shy. Set the right price. That is your edge because high LTV will allow you to be more competitive and get the right users you need in this super competitive space. If your LTV is low and your CAC cannot be competitive, that is definitely going to be more challenging.

I want to share something different. Like many others, I, being a very hands-on media buyer myself as CEO, rely on the big platforms for user acquisition, and that is always a concern and a challenge. When Meta earlier this year made an algorithm change, many companies felt it, including us. You have seen companies like Audity, a public company, reporting that as well, making a huge impact. That is definitely tough, so you need to have other venues to grow.

Something we realized, connecting to our larger vision, is that there are so many apps, communities, and high-intent audiences out there that we can actually purchase. We can actually roll up apps, services, and communities that do not have our strong monetization or our strong retention, but their audience definitely resonates with our vision. By rolling up those apps, we actually unlock high-intent audiences all at once very quickly.

We have just completed our first acquisition of Flame, which is a relationship app with about 250,000 users worldwide. Again, high intent, people who opted in to invest in their relationship, but monetization was lacking. We can come in with the strong monetization and the value proposition by Arya. I am having more conversations like this and getting more and more inbounds from different apps and services, not necessarily in the relationship space, but adjacent and other life stages.

For me, it is really interesting. Five years from now, take Bending Spoons as an example of rolling up and making those audiences more efficient. User acquisition efficiency, financing that we bring to the table, monetization, and retention, and the three core elements of trust, content, and commerce. Then user acquisition beyond just paid social gets really interesting when you unlock and open up those audiences with acquisitions. That is going to play a big role in how we are going to grow in the next couple of years.

ES: That foundation of domain expertise creates the settings that are needed for a viable company here. You saw a tremendous growth in the number of apps launched. In the App Store, there was an 86% increase in the number of app submissions in Q1 2026 versus Q1 2025. It is much easier to build an app now. Guess what? It is still really hard to monetize, and it is probably harder to monetize an app now. It is probably harder to distribute an app now because there are many more. You had a lot of companies that were launched into this environment where they can spin up an app very quickly and easily, but that monetization expertise becomes even more critical to actually growing a business. It is not just launching an app; it is creating a business. UA has gotten harder and harder every year since 2008, since the inception, but it has inflected now. People with true genuine deep domain experience have a massive competitive advantage operating in this space.

OY: I think gone are the days when you can just take an app and put like an annual membership and slap whatever $59.95 and just expect people to pay. Content is great, but content is accessible in other venues. Solving real problems for users, whether it is through commerce or other aspects, is how you can properly monetize, and that is really hard to scale and to replicate.

ES: Offer, this was fantastic. Any closing remarks you would like to leave with people? How can people learn more about Arya?

OY: First go to arya.fyi or getarya.com and you can learn more. Eric, let’s open calendars and let’s have another one in two years so we can check in and see how we are doing. I am loving this.

ES: I am in. Let’s do it. Offer, thank you so much for taking the time today.

OY: Thank you, Eric. Bye-bye.