SPEAKER_01 0:03

Welcome to the Future Ventures Podcast. It's Kelly McClary. On today's episode, we're joined by Stanley Wee, founder and CEO of Pine AI, autonomous AI platform that doesn't just answer questions but actually takes action on behalf of consumers. Pine AI can negotiate bills, cancel subscriptions, file complaints, resolve disputes, and even navigate insurance appeals automatically. Before launching Pine, Stanley held leadership roles at the gore and invested in AI through Hill House Capital. Today we're going to explore the future of AI agents, consumer empowerment, and what happens when AI begins operating in the real physical world instead of just chatting about it. Stanley, welcome to the stage. Thank you for having me, Maxim. It is my absolute pleasure. I'm looking forward to the conversation and get your perspective in terms of like what pushed you from like investing in AI to actually like what did the conviction come to actually go in and starting an A an AI company? Yeah, just walk us through like your your thinking.

SPEAKER_00 1:30

But actually it's not, right? So it's uh, you know, but uh for but but it's kind of but but it when it gets to like 2020, 2021, right? So when like I I believe like it's like Dali, when Dali came out, and uh I believe like everything changed. They might be the true trend. This might be the real one, right? So and uh I'm a deeper believer that AI 100% is gonna change in like you know, it's gonna be the final revolution of uh human industry, right? So at that point I had you know I started to believe like you know, so the Jen AI is actually that you know, I mean like the ADI might come. Uh so I have to find a you know, I find a way to be to be relevant to this like industry, like the paradigm shift. Um, I guess you know, by you know, so by accident, I had this, like, you know, I had the this bad experience, you know, while I was traveling around, you know, especially when I was like uh you know manage like the this like uh uh international company, have to travel between uh US, uh Singapore, and the UK all the time, right? So I have so many issues, like so many, so many issues I have to deal with all by myself, like the how to make phone calls and uh you know, write emails, like or like you know, uh, or even like log in the website, fill up a lot of forms, right? So that's my personal pain points. You know, what if like can be done by using AI, right? So we started this company around 2024 when the model was not there yet, right? So it's you know 100% take a hike and to build an agent. You know we tried to we tried to build an agent that can you know make phone calls, call emails, orchestries, like all the agents to get something done, right? It's uh really, really difficult back then. Uh and uh, but you know, so when we actually, so by the end of 2024, so everything became like a possible, right? So especially there's so many research papers about agents, etc. And um, you know, the model become like a stronger, and uh we actually build something that actually gets things done, and um, you know, that become like um, you know, maybe everything become possible.

SPEAKER_01 3:26

I mean I love the story because isn't this how most founders uh get started? They define an idea that's irking them, and they're like, I'm just the intention is I'm gonna look out to see if there's a solution, and typically there is no word. The solution that exists, it's not exactly on point. So, like hell with it. I'm just gonna go and solve it myself. Uh so that's fantastic. Um Stanley, what um what are the biggest use cases now for um for voice agents?

SPEAKER_00 3:57

Like, yeah, I mean like uh for now, um, I think voice agents, you know, so there's so many potentials about voice agents. And it's not only just like you know, making phone calls or like you know, just uh chat with you, it's also some kind of like an interaction. I mean, I will be I'll be I I believe like you know, so voice actually is the most like uh the most smooth and um the you know the natural way of you actually interact with like you know machines, etc. Right. So especially if uh recently and you know there's a demo about from sync, you know, sync machine, uh and also like from grok, right? So there's uh you know, voice agents that make you know can make two calls and uh can like you know interact with like people very naturally. So you know, I mean like Jarvis from you know Aaron Mine, you know, is uh truly possible now, right? So you know, think about it. Why you know one of the backstories we had is like our chief scientist, like recent, uh just like they published like a paper um I think like uh last week, right? So because you know there generates generated a lot of hype on LinkedIn and X. It's using like they he came up with this like uh um non-compressible like a knowledge. They use that you know to predict, like to predict like uh, you know, uh the size of the model, like of those, like uh those like a saltar or salt models, right? It's uh you know, he he actually didn't write a he didn't he didn't code, he didn't write a single word. And it basically just like used a voice agent the agent talking to like a claude and to write the code, do the experiments, just submit the paper and get things done, right? So that's the I think that's the future of how you're supposed to work with machines, right? So you don't really have to sit down there, type like a tap everything, and uh, and it's not even just like you know, voice into text, right? You you just speak what you want, and then you know, it's just like send it as a command. It's supposed to be something like you're working with a multiple virtual assistant at the same time, right? So you you support you know the human itself and them themselves, the you know, can be the orchestrator, you know, sitting the it's like conductor, like sitting in the middle, sitting in the center, and they're telling agents what to do, right? It's uh you know, I mean the only band, the only limit might be the bandwidth of like, you know, humans like recognition from NOAA, right? So that voice part is actually more efficient, just like you know, you tap everything, you know, on the computer. And um, you know, that experience actually, so you know, I'll I believe like, you know, so that's the whole voice. I I would say like voice agents like developing in the future. Uh of course, like you know, now so for for our use case, and uh we we we use a voice agent to make phone calls, right? So if you deal with like customer service or like you want to deal with a garment or like you know insurance company, etc. And uh you have to make phone calls. And a phone call is one necessary like uh piece of uh solving the problems, right? So it's uh but a phone call itself is not you know it's not it's not enough. You you gotta make a phone, you know, you gotta like send emails and uh even talk to their chatbot on the website, or like fill up a lot of forms like on the website and to get the thing done, right? So it's just like you know, so you have to like put everything together and to make things work.

SPEAKER_01 7:07

So I I I have a personal story, but I also have a follow-up question. I mean, I couldn't agree more with you that voice is the new interface. I mean, like um I I use an app called Whisperflow, so I I rarely type anymore. So I was just uh I signed up for it 100 days ago, I was looking. I have spoken 350,000 words in 100 days. My uh input speed has gone from like I'm not a fast typer, from like 75 words per minute to now 140 words per minute. And so I couldn't agree more than the voice is the new interface. Now, in order for agents to take action on our behalf, it's predicated upon building some kind of a memory about us or know having knowledge. Like, I I don't doubt that they can take the action, but kind of like if if if a voice agent is taking uh an action on our behalf, where does the data reside? Like, is there like a rack that's being built around uh around the agent? Because like like let's say that they're disputing uh a bill, obviously they have the bill, they must have access to the account. Kind of like walk me through kind of like the architecture in terms of where does the information arrive for the agent to take the right actions?

SPEAKER_00 8:19

Um yeah, I think that's a super good question, right? So you know that's uh I think I believe that has something to do with the harness part, right? So and then you know, so you know, I mean like models actually halluciness, right? So if even you even you even if you use like strong models such as like you know, so from Claude and the latest model from Claude and uh use claw code like do do something. It just like uh you know, you have to like literally tell them like to do it again, right? So that's uh it's uh that's a kind of a you know a bummer. Um and uh what what what you know our approach is actually to and uh we build like a really strong harness layer on top like the execution, right? So the it actually can, you know, there's actually multiple agents like an architecture. So basically, you know, they check their each other's like you know, homework and uh you know simultaneously. So to make sure, like, you know, so you know you don't actually simply just forget something, forget about something, right? And uh there are also like a you know uh stronger model or lightweight models, like you know, all lightweight models offer like a really responsive like uh feedback, but also like in a stronger model, actually they they they offer like a deep thinking and so that you don't make like you know, uh you don't really make like a mistake, right? So in our use cases, right? Most of the most of the use cases are like a really mission critical, right? So you uh negotiate bills like on behalf of the user, and uh if you screwed up, if you screwed up, like that means like a dollar lost, or like you know, their accountant got locked out, right? So you have to be extremely careful, like what information you bring in for the authentication, but also at the same time, like you know, so when you talk about like, you know, are you going to accept this like a $20 discount, or are you supposed to like reject that and and you know and and ask for more, right? So that that everything like you have to make sure like the model follows like the instruction, and also you know, also it doesn't really forget about like you know the details about like you know, the detail about how much like how much a discount like it's supposed to ask for, and um and uh you know what's also you know all the information it's supposed to take and uh what uh you know the the strategy is supposed to follow, you know, uh to to negotiate with the other party, right? So that's um you know that's that's that's how we deal with it.

unknown 10:35

Yeah.

SPEAKER_01 10:36

So if if if if consumers have agents, it's only logical to assume that companies would have agents. So it's is going to be like a gentle execution all the way where uh a consumer agent is talking to a company's agent and and two agents are negotiating on each other's behalf?

SPEAKER_00 10:56

Um, you know, so I we we do see like, you know, so it's not it's not you know, for either for like a negotiations or like you know, bill-related issues, it's you know, is that's considered like a mission critical and uh very rare uh for a company to um actually use it, you know, uh you know, agents, you know, to handle the situation like that. But most cases, like but if you if a simple matter such as like, you know, so oh I need to reset my password, or like, you know, so can you tell me working hours? Sometimes you you you know you rang into some like agents, right? So it's kind of funny to see you know agents talking to agents and uh you know uh you know back and forth. And you know, of course, like they they never know each other, they are at you know, they are this you know belong the same kind. Uh but it's kind of hilarious.

SPEAKER_01 11:43

What um what do you think is standing in the way for for for more people and companies adopting agents?

SPEAKER_00 11:52

Uh you know, I mean like a Yarra company.

SPEAKER_01 11:55

Uh just in any company, like for example, like let's say that you could tell, or let's say that you're a consumer, like why would why are people not adopting it faster?

SPEAKER_00 12:07

Um I think the thing is like, you know, so you think about the world is actually is uh you know is uh is uh is very divided, right? So on one people don't know anything about technology, right? So on the other side, you know, part of like a spectrum is people deeply understand the technology, but you know, without any awareness of like the problems, you know, you know, in in specifically industry, right? So those people, you know, it's very rare to you know to find someone who actually understands tech, but also as you know, at the same time, like understand like you know, the industry pinpoints, you know. It's just like you know, it's gonna take time, right? So when there especially something monumental happens in that industry, they're gonna, you know, especially there's uh it's supposed to be a like a tipping point. Um, you know, when everybody thinks, okay, we need to adopt this time of you know technology or else we're gonna be eliminated by the by competition or like in by other customers, right? And um, you know, at that point, I mean like so the adoption rate will would accelerate 100% for sure. You know, think about this, right? So even for you know, in until this day, so people still talk about like digital transformation, right? That's a think about like you know, so internet was like a you know from almost like 30 years ago, right? So now we still talk about digital transformation. That means like you know, so it's uh it 100% gonna take time, you know, to to penetrate like you know through like entire industry, right?

SPEAKER_01 13:40

So I I and I mean I completely agree with you. I think that they got my first email in 92, 3, 4, I don't remember, something like that, but it was it was a while ago. Uh so that's 30 plus years, and yes, there's still companies that are operating in in land worth. And when I see this, it completely surprises me. So besides customer service, uh Stanley, like what what what's the next foray into for agents? I mean, obviously customer service makes sense. Um, because it's like what what comes next?

SPEAKER_00 14:12

Um I mean, like for us or for the industry? For both.

SPEAKER_01 14:16

I'm I'm curious, kind of like what is what lies on your old map and what do you think is gonna be the the next adoption case?

SPEAKER_00 14:23

You know, for for us, right? So um, you know, we do we started with uh you know customer service. It's actually my personal pain points. And um, but but I but going forward, right? So when when we actually release the product, you know, we see that so many use cases is you know, not only just like dealing with uh customer service, but also like you know, uh doing all kinds of other type of chores. For example, like you know, so the make reservations, right, manager their appointments, or like, you know, uh, you know, third-party communication, people actually many users use us like for information inquiries. For example, like uh uh find me the cheapest like plumber within three kilometers with three miles, you know, within three miles like where I live, right? Yeah, so that's uh you know, it's a very, you know, so basically the agents just go to Google Map, find like uh, you know, all the plumbers within like three miles of my place, and they're gonna make all the phone calls like with all the plumbers, try to find the cheapest, right? So it's uh actually so it's uh it's uh it's a fascinating to see how users like use uh use a pie to solve all kinds of problems. You know, uh the other day we just like talked to uh you know a construction contractor and they use us like to do you know to do many like a garment-related issues. For example, fill up a form on the website and asking for like uh you know uh the progress or application of permits and uh or like just you know handle like their internal communications, etc. Right. So you know, I mean it's fascinating like uh to see uh how user you know you know they are so smart and uh they can leverage like the all different like AI tools, you know, for different purpose. You know, we did an interview with with that guy, so he he actually tried almost all possible like you know, AI tools are available in the market, right? Yeah, so that's super, you know, I mean that's like a super like impressive. And uh I would say it's like farm from you know in our case, like you know, so our uh product roadmap is pretty clear, which is like, you know, so uh we either you know in the very beginning, we just launched the tool and it it it it does email, I make phone calls, like it does fax, or even hire somebody from Fiverr or like Appwork to get something done in the real world, right? So, and uh that's the all the capability to offer to everybody. And uh, but you know, we actually observe the user behaviors on the on the platform, and it was you know, we started to see like you know, there's a you know, there's so many use cases emerging on a daily basis, right? We took like uh some of the use cases, like you know, so we analyze, like you know, we analyze them and uh we try to optimize them like to you know uh to be more successful, right? So you know, for example, like you know, uh the government, you know, communication-related cases, like you know, uh after our optimization, so we pretty much uh we can achieve like a 95% like success rate. Now that's gonna make like the heavy users or power users get super happy about the use case, right? So I do see like um you know in the industry, I think some you know, uh if you look at Chat GPT or even like for uh some like uh even Microsoft co-pilots, right? So they help you dealing with uh you know the PowerPoints and um you know PowerPoints and uh you know spreadsheets, you know, that actually you know really helpful like when you are professional, right? But uh but things like you know, it doesn't really connect so much like with the physical world, you know, because the physical world is like is uh is very different from the you know the digital world. And uh there's uh the infrastructure is beautiful for humans, not beautiful agents, right? But uh, you know, but we do but we 100% believe like you know the penetration actually started, you know. But phone call is actually the you know, I I'm I'm I'm not sure. Have you watched like the movie Matrix, right? So the phone call is actually the the channel between the real world and the virtual world, right? So we want you know you know we have to make a phone call to actually you know travel between these two worlds, right? So it's kind of for you know, in my opinion, it's it's kind of true, right? So our agents actually call the real world and uh to get a problem resolved, right? So that's I think that's the very beginning, the kind of the very first step, you know, for agents to get something done in the real world. It's very it's totally different from like you, you know, build PowerPoints or like a spreadsheets or like you know, simply just a you know do a desktop research, right? So uh you know to me that's kind of monumental. I believe like you know, so all the agents, not only just us, but also like you know, the all other agents, their goal is supposed to get things done, right? So you know, and uh they have to like uh uh you know create like a real value, not just like you know, so come up with a strategy, you follow up and then you do it. You know, it's now it's like the agent come up with the strategy and do it themselves.

SPEAKER_01 19:05

Um is there particular um industry with the the the the first adopter um or where this will become most widely accepted? And then what other industries would follow?

SPEAKER_00 19:20

I think you know, so I mean a customer service actually is um you know is uh is proven to be a very valid industry, right? So uh if you see like there's a many big companies like you know, you know, it's already like uh fast growing companies like in this uh in customer service, right? So it's on the B2B side, right? So they help like the enterprise, they help companies like you know, business, you know, to uh reduce cost like by adopting like uh agents, yeah, right, in the customer service. It's a proven use case. You know, think about this, right? So even for um, you know, 100% like you know, solve like the scalability issue of a customer service, right? So, you know, it's uh the I mean like especially the you know the you know um if you can't it can you know almost like uh achieve like 100% scalable, like you know, so if you're using like an agents in that in that use case, right? Um I'm also um there are also other like a vertical vertical like uh you know industries. Uh for example, like uh you know, uh I know there's a couple companies working on the voice part, right? So, you know, you know, especially in the traditional like industry, like people, there's no software, right? So the basic is making falls with each other, right? And uh the information is very like a standardized, like very typical, just like for example, like you know, uh the hospital and the clinic and an insurance company, and uh you know, the reception always try to call the insurance company to verify that person's identity and uh policy, right? So and everything is done by phone call, right? So it's not it's not worth it to actually build a system. You know, if we build a system, we have to embed it like with other systems and to interpret with other systems, right? But uh, you know, but phone call is so flexible, it's you know works with almost everybody, and there's I believe there's a company actually uh building phone agents, you know, for them to communicate with each other like easily.

unknown 21:12

Yeah.

SPEAKER_01 21:14

Yeah, I mean, I agree. I um we're building we're building an app for one of our portfolio companies, and uh we're using voices as the interface. So in this case, rather than having to type a purchase order there on the go and they just speaking into um like using the voice to to create a purchase order on the fly. Like, okay, I need this or to create tasks. It just it it's become so natural and on the fly, and you you speak faster um than you type. So it just it just makes sense.

SPEAKER_00 21:46

Yeah. I mean, like uh, you know, so there's a it's a and I mean there's a phone call is actually the best way to reach out to the real world for agents, right? So it's kind of like you know, like I said, it's a channel. connect like the two worlds. Yeah.

SPEAKER_01 22:03

Yeah, no, it makes sense. Um I noticed on on your website your pricing is based on on plans. Like you have the free plan, you have the starter plan, you have the enterprise plan. But it's cost has a real cost like why like kind of like what is your decision? What is your thinking process in terms of arriving at the pricing strategy?

SPEAKER_00 22:27

Yeah, the you know that's a very good question actually. So if we are using like a cost based like cost based like a pressing so you know we we probably like you know have a estimation of uh like how much it costs like a um for each of the tasks like on average and then we come up with like you know so we give like a you know enough like credits for a user to do like a you know uh a number of tasks tasks like our on monthly basis right so then we plus like you know we we add a margin on top of that.

SPEAKER_01 22:54

Of course okay okay so you you estimate it and is is the estimation based on a number of uh of tasks they want to complete in a month so it's it's essentially you've kind of tried to reverse engineer outcome based pricing based on kind of like what uh how many tasks we want to be completed in a month.

SPEAKER_00 23:12

Yeah yeah so we do that actually is we do the estimation of like uh our users it's a power law distribution right so it's like you know so it's always like you know so 20% like a user generates like 80% like uh you know tasks right so we we actually you know we have to like smooth that out like uh by putting like a somewhere like a you know uh you know just like take something like in the middle yeah what's your process in terms of thinking about scaling the company uh getting more use more consumers to adopt this like how are you going about it yeah I mean like uh you know that's a that's a super important I mean like an you know think about it right now so you know web coding like you know so every agent's pretty much do like all the coding now and um come you know building a product is not a isn't it's not that difficult as before right uh you don't have really have to like uh you know own uh a large engineer army to build a like a fantastic product and now it's just like maybe two person you know you can build a very good product but the problem is like you know so but the you know the the heavy lifting actually you know goes to go to market right so you have to like you have to think about like growth all the time and I think like you know growth now is more important than product itself nowadays yeah right so if you look at there's a there's it's actually think about this it's actually structural change on the supply side right so because like agents you know doing coding is so much you know so much better even sometimes is you know I mean like it is better than some average like engineers nowadays right you know building product is easy so there's a think about this it's gonna be like infinite like an you know like a infinite like a supply you know right but uh but it but if but think about the demand right so there's no structural change like on the demand set there's only Google and meta right there's nothing else right maybe like in OCS maybe you let only Google and meta that's pretty much you know that's that that's everything right so there's no Google there's no new meta it's just like Google and meta that those guys just keep you know increasing the price of a CPC CPA right so making like a you know uh acquisition as you know the cost of acquisition is uh super expensive right it's it's it's so expensive that it's already become like a not sustainable for most of the companies right so there's a so you know it's a super critical for you to hit the critical mass of the users first and uh that's why like you know so think about the United States you have to like think about like a growth on daily basis and that becomes like a number one priority for almost everybody. And and and how are you guys thinking about because I mean you absolutely right Google controls a huge swath of of products that that get the consumer eyeballs and and meta does the same so how are you guys thinking about um um acquiring the users like who what are you what should be called to market strategy yeah and you know think about this right so um you know our product is it's not something like you know so it's it's a it's a consumer product and uh and uh it's the most consumers they don't even know what ai can do right so you really have to understand like you know so what's your product's value proposition and uh how you can sell you know what's your selling points right for us it's like you know so our top like three selling points like you know first it does like comprehensive research about the matter it can come up like a very like a detailed like uh detailed and smart like a strategy second is uh you know it's a very good negotiator and uh you know it doesn't really hallucinate it you follow instructions like 100% right third is like it's very you know it's consistent it don't just like it doesn't just like you know give up or easily and it will be like it could constantly follow up and uh try to get a sense that in the end. Right. So and it has to like uh weave every like a setting points like into your um you know advertisements or like you know your videos like on the social medias right so to make sure like you reach out to as many people as possible right so I mean for us or even for most of like AI products right so awareness I think is the most important thing you gotta educate the user right so um you know you first you have to make sure they understand the pinpoints second is like you know so there's a solution in AI that can solve your pinpoints third is kind of like uh you know this is a pain point you know uh you have to believe that you know it actually can right even people realize that pain points understand you have a solution but is they don't they might not believe that that your AI can actually solve the problem right so that's called that's kind of the trust how you can build trust that's the actual layer you have to you know offer or like you have to build like a you know uh you know uh from you know it's very different from like a what it was from the you the mobile internet you know you know a face right so you know it's a you know so at least like you know so five years ago if you want to build a mobile application you don't have to educate the users you know anymore because everybody knows like oh this is the application and it it does it right so I try it it works but I you know AI people say it's like what can I even do this you know or like you know so what's the what's you know is there what's the catch right so and you know also like there's so many AI products nowadays right so you know everybody's can you know try to catch eyeballs and uh but uh you know if you talk about one pinpoint there's maybe like 20 000 like uh applications like doing the same thing right how you stand out right so it's um you know it's a very very challenging and and how do you stand out the sense like you know so I mean uh fortunately so I don't think there's as many people are trying to do the same thing as we do and uh 100% like you know so I'm pretty confident that we offer the best like solution in the market and uh you know if you try it it actually gets things you know it can help you get things done right so I save like you know you know tens of a thousand dollars like you know some you know on on on a yearly basis like by using this app right so you know for some you know most of the time most of the time like you know so you don't know your rights are you know are violated until it's super uncomfortable right but now it's kind of like you know so every time I feel like a little bit uncomfortable with like the business or whatever I'll talk to to my talk to Pine and uh it does the research and you know sometimes that defends for me and there I don't really have to worry about anything right so it's about like you know so uh should I uh should I be arrogant you know arrogant or decent enough like not to like you know you know fight for like a two dollars or like you know should I just like you know uh you know raise my voice like you know to you know and uh uncomplain right something like you know so easily give up and now can be like uh obtained like a fiz without you know fit you know uh you know doing more efforts you know why why not right it's um you know that's uh uh you gotta be like you know it's the the the strategy part is still the same as uh as before right so either you have uh you know very differentiated like a product or you're very good at selling right so you gotta be good at one at least one aspect of that well it I mean you you have the product led growth you have the sales led growth sales is typically in in enterprise community led growth becomes really important if you have really really good product that's differentiated um and people people pick it up and love it then that community is kind of what drives the the the the growth um so uh now can you uh for for people that haven't used a voice agent kind of just can you just kind of outline the process of of getting set up kind of like what happens when you initiate a call like do you are you on standby like is the human in the loop that kind of intervenes or does the agent execute from end to end and the outcome is the outcome yeah so um so the things like for easy for easy phone call for example like a make a reservation uh with a restaurant that's easy right so you know the agent just could make a call and say like okay uh you know do you have like a you know a table for two like a today tonight at 6 p.m right it's easy right so you don't really have to involve the user anything but if you have to like negotiate with like a for example your your uh your landlord or like you know so you have like uh some hospital bill issues right so that's uh you know for compliance you know uh on for compliance like a purpose and uh the the agent you know I mean the representative need to need to like speak to uh you know the the user right so uh for authentication purpose and uh you know authorize us like to speak on his behalf on their behalf and so that like you know so so that as a case authentication is uh you know it can be passed right so uh or even for like some you know significant decision for example like you know so uh the user said I want 30% like a discount on this like uh on this bill and uh but uh but it you know but uh but the business only you know can can only offer like a 20% are we going to accept that or not right so we usually make a phone call you know we are kind of like a you know we kind of like try to smooth it out like with the representative or like a you know on online but also like uh separately we actually make a phone call to the user seeing like hey man so uh you know the you know the guy on the on the we're on the phone with uh somebody and uh they're pretty tough i only offer like 20 are we going to accept that or like you think uh we can do better you know for with the next with the next try or something right yeah and um you know in that case like you know so uh you know for significant decisions like that you always uh try to consult with like uh you know the user first makes sense um i worry that the the the lms and the super apps are going to encroach in this space i mean i started to use uh like to do this uh use scenario i i i think i use open ai's agent backing generally to make a restaurant reservation for me i i and and it and and they have a voice input so i worry that the lms are gonna come in and take on that space yeah i think that's uh you know the the things like um you know that's very actually very very good questions that's yeah that's something like giving you a week like in the middle of the night uh all the time right so something like open i throw like a throw something like a bomb right so you know just destroy all the startups like you know so in the space uh it it is it's it's you know so uh I have like two uh two counter argument uh first is like uh you know um make a reservation is easy right but um but it actually try to do something more complex such as like you know the the bill negotiation right so try to make a sales call or like you try to like do a uh user survey or like you know try to like a manage your vendors or like you know uh you know have like a really complex you know conversation for example like information inquiry or like you know with the garment etc right complex scenarios like that open I and or even growth they cannot do it just like you know you know sometimes it's because like hallucinations or like sometimes it's about a you know pronunciation right you can cannot uh they cannot handle like really complex like you know conversations right for now at least uh second is like um you know uh you know there might be um you know uh I might be wrong but you know I heard about like you know so uh anthropic doesn't really have like a voice model themselves right but open AI and um you know I don't think open I or even for Gemini uh they have like their you know focus on focus like um on multimodal uh multimodal like um models at this point right you know think about this right so um text based like language model like I mean for text based language and models and uh there's so many training training like data but if for like multimodality like um you know models like data it's uh there's pretty much very very few right the only reason we can do it better so we train our own model using our own data right we're showing again so we actually you know take a lot of the test data and uh you know so we actually so we label it like and you know we agree you know and then we train our own model and uh so that you know to make it you know make it very different you know so it's a it's a top down it's like a bottom up like approach right but if you if you think about open eye or like you know anthropic they actually try to build the model from a top down approach right so they kind of take in all the data everything all together train like a universal model that is kind of like a universe for universal purpose right but if you think about like a making phone call making phone call just one aspect about the entire spectrum right so there's a white agents about like chat with you only right or taking orders command control right so you talk about you think about like a making phone call is a very special case is you know interaction in making phone call is super important than other voice agents right so interaction you have to optimize interaction to make it make it supernatural so that you can so that the conversation can be carried out in a natural way right but it doesn't really if you don't optimize that using like you know Gemini or like even open AI the interactions will be like a super awkward for example interaction right so interruptions and um while you're speaking you keep saying like yes you're right 100% right so I will not be an interrupted because I think I consider that as kind of encouragement right yeah but uh but if for models like so they they they don't know what the difference between that right sometimes just like it stopped right so it just stopped like you thought are you going to say something? Um not really right so it's become like super awkward and uh you have to optimize model on that I don't really think like you know so the the first like you don't really have like a enough like a domain specific data you know it's it's a true for everybody second is like it's not their priorities like to optimize like a phone call based like a worse model at this point right yeah um the you you mentioned uh sales can you use pineai as as a as as a sdr um or yeah is the focus okay yeah it's it's it's not our focus at this point right but if we do see other users like try to sell something uh by using by using mine so and even people doing like user surveys uh so it's like okay i'm gonna I'm going to interview this you know uh 10 users for my for the product experience of our our ours and so it just like give us like you know a questionnaire and uh ask these questions if you you know take your liberty to ask like a you know question in depth you know uh as long as it is relevant to the topic right so or even try to uh help me sell this and to these 10 people and uh make 10 calls and you know summarize everything you know and uh get back to me that kind of thing right so you do use that so I mean I can see the different use cases that they're on your website in terms of lost luggage or late delivery or things like that. Like um are you seeing users pushing the product in in in cases that you had not anticipated and you're like wow okay I didn't know that uh that that's possible yeah I mean like there's a one use case that's super impressive right so that I remember there's like a uh there's one guy from a Wizabisk Bikistan right so he actually he you know he came to he came he came to the US like during the pandemic and like a found a job like in as a you know front end engineer whatever right so but you know uh just like in 2025 I I remember it got laid off and uh he knows like a hit he's pretty much at the brink of uh broke of being broke okay right and um if he's like uh you know he's just ran into our product like you know just chat with like a pine very naturally you know you know talking about his dilemma you know he's thinking about like you know ideas about making money right so we just so after a few like uh you know you know uh around like conversations like you know uh they found a way so he worked with Pine to find a way that you know there's certain type of like a Japanese car you know in the dealer in in the in the in the rental car companies uh you know uh is for sale what was for sale at like uh uh $19,000 something like that. Okay. Same type of model since the same model of the car that can be that actually can be sold at the like the car dealer you know at a $23,000. So there's a $4,000 difference right so yeah actually so they work with client actually you know convince like a convince like the dealer and to loan him like you know $19,000 to buy the car from the rental car company and sell it to like a you know the the dealer dealer in the dealership right so he made like a three thousand after one thousand dollar handling fee whatever he made like three thousand dollars wow so it it was he was able to find the arbitrage okay in in this case can we just have a look in this case how did he know that the rental companies is has like is it like an auction side like how how did he know that the rental company had this car for sale and it and it was an arbitrage it was it was selling a at a lower price than what a dealership can buy it for. Yeah yeah it's kind of like you know so he knows that you know so he knows like there's actually the difference you know so you know so for rental car companies like this is the after certain mileage of the car yeah the URL will sell as like a a dis heavily discounted like a price right so and he know he has that knowledge right so he he used Pond and to call as many rental car companies ask different like cars like a price. Once you have the information and they call like a you know the dealer they call the car dealers and say like you know I have this car at this price you know do you uh do you want to buy right something like that. Is the arbitrary like between the you know the price difference actually made a you know make quite a few bucks and uh you know that's an amazing story actually uh and uh yeah I know that's uh and also like uh you know you know uh also there's like you know credit card interest rate is uh negotiable you know that right so there's a there's a one person who actually borrow like a five thousand dollar from the like from like a credit card company and uh but I cannot keep can cannot pay back anymore and uh you know pine actually negotiate like with the in you know credit card company like a for you know from a five thousand dollar to like a fifteen hundred dollar you know for one time payments and then all that clear right so they have this like a negotiation like you know so going on but most people don't really know and sometimes they're just like you know uh you know just let it go and then you know you you're screwed up like your you know your credit right so your credit profile and um you know or like simply just like you know you have to pay heavily you know think about like interest rate you know uh it's like a APR is like a 19% or like 20% something like that. Right. Yeah you know it's about like you know it's that's a thing right so if uh think about this right so you have an assistant you know sitting next to you pretty much knows everything about the world right all you need to do is just like constantly chat with the the assistant right you know there must be many good ideas like coming up right and uh but the only way is like you know so the interactions the interactions is super important right so if you have to type right think about chat easy now is only type or like you know the conversation is a little bit awkward or like a not natural or not smooth and you don't want to talk to the agent in the assistant. Right.

SPEAKER_01 43:22

But uh but if it is a think about this is it's uh sounds like a real person and uh the conversation also is funny and uh interesting and uh you know you the more you talk to the agent so who knows pretty much everything about the world the more good ideas you can you know that can come up right so I I I'm I mean I'm I'm I'm a convert on voices interface like like we have we we have a dog and like majority of the talk people think I might think that I'm crazy because majority of the talk I'm I'm I'm on my uh on my butts and I'm constantly talking to uh

SPEAKER_00 44:00

uh either to different apps uh or i'm i'm taking notes or i'm just writing things um now i i get the album part can pineai answer calls on your behalf uh yes it it can so we recently just like you know so we are going to release this a feature uh I think it may be the week uh next week or week next week uh you know uh or maybe like uh you know maybe next a couple weeks yeah we we're ready to like uh you know launch like that that has a new feature fantastic okay um yeah I I can't wait to use the product I'm definitely going to uh give it a try and see how it works um what um what again our audience are as a scaling up founders and and other founders kind of like what could be and I know that you're consumer focused but what are the best use cases that uh uh founders can can can adopt in terms of using pineai um you talked about customer discovery um kind of you talked about a little bit on the sales side what other cases can can be adopted and think about this like actually so many YC company founders are using us as their like a personal assistant right so for example like okay so you can't just like talk to talk to the talk to Pine and see like uh okay uh you know that you know I'll I'll just describe uh how I use Pine to a formal yeah right so I usually walk to my works you know I I usually walk to my to my office on database I walk back and forth right so that's usually taking me about you know certain minutes certain certain minutes per trip so uh I usually on my way to my office I talk to Pine you know check my emails right so it prioritizes my emails and they help me reply all my emails and uh if there's an issue I say like okay uh there's an issue and uh for example I need somebody for example my CFO to pay attention to an invoice right so I'll just say oh hey Pine just like uh uh find Charles and tell him that you know tell him that you know so to pay attention to this invoice maybe if I have there no problem just pay it right so but he said but as before yeah I have to go to the office you know and maybe I sit down you know at like a Zoom meeting and talk to you know talk to like the CFO and say like you know something about this and that right so but on my way to my office I told my you know I just like talk to the agent going through my emails and I go through my calendar and try to you know and trying to give me even before our meeting it actually make a phone call to me it's like oh you are going to meet like a Maxim and uh he's uh and what is a podcast is about and now who is the Maxim yeah right so he give me like a background everything right so and brief me everything and um and on database on database on my way to my work right and uh we have like a lot of we we have vendors such as like you know uh AWS and a Gemini I mean like a Google you know or even Facebook a meta right so and we use Pine to manage all the vendors right so on on a monthly basis you know Pine inspects all invoices our company and uh schedule like Zoom meetings with all the vendors and the robot you know I mean like in your Pine actually enjoying the Zoom meeting talking to their salesperson or like their their solution architect and uh to report like any outage or problems or issues right even ask for compensation for ortages right so it's 100% fully automated and uh even you know so it you know um it actually helped me like organize all my expense reports and um you know so I have you know it connected with my other in a way you you uh uh we are using like a Xerox right so basically it connects to Xerox I just like um you know uh upload all my you know invoice you know all my like a receipts everything help me like doing reimbursements uh there's uh so uh you know in that case like you know so we don't um you know I I don't you know there's very very few meetings in our company so basically you know so just you know I send my agents agents like talking to other people's agents right so the agents were talking to each other and uh figure out like priorities like understand all the context everything inside the company like you don't really have to have meeting anymore right so because everybody shares the same context and uh all you do is just the information inquiry it's just like you query something else like hey I was uh you know what I I'm looking at this bug and uh you know tell me like uh when Dylan can actually fix it right so that's like you know so it I don't really have to walk to the to Dylan's desk and um you know they interrupt him from his like a coding and uh you know that that that kind of you know is a super disturbing to most of like coders right and uh but exact his agents would reply to me thinking like oh yeah he's working on that right now don't worry about it right so that's the best i mean i mean that's amazing like I I love this productivity hack so where is the institutional memory built part uh uh question one and is it within within pine kind of like you you you you spoke about context but kind of like where's that context being built and the second question is like you're talking about zero in terms of integration but like I'm assuming that is an MCP like I'm I'm able to connect whatever app I have with Pine and give it that uh interface uh we don't really use magic in our company but we use we use like CI and I mean the command line the command line is actually so no is so much better than than than anything it's uh okay I mean like it's uh is uh is uh works really well easy to reintegrate and easy to manage right um the context right you know so our our simple you know most of the context is uh is our we have like you know we we connect with Google Drive and uh even Microsoft like a Microsoft like you know so uh workspace and uh you know Microsoft like 365 and um uh oh we we also have uh like a repositories our like um uh our like even our our GitHub right the GitHub and uh and the some and the uh and the the also the bug like repositories etc right so and um you know it's that's that's the that's that comp you know that's all the context we have in the company and also confluence we have confidence for like documents everything or even notion right notion right so and uh that's that's the context right and um you know and that's the most important thing is about like you know so you can't you can't trust like human written like a documents anymore you have to like uh go in our company and that you don't humans don't really put a code in the repository and the human don't write documents anymore and uh you have to go through the agent yeah and uh to write the documents and write code right yeah that's actually it's a it's uh agent friendly if a human started write something you know so uh the agency might not like it yeah no I hear you I hear you um that's amazing um any other hacks that you have yeah I think the um I think the more important thing is uh the most important thing about like you know how you work as a team right um it's uh you know so though actually so you know you know the way uh engineering teammate gotta manage is super different um than than than previously right um previously it's about like you know so okay we're gonna break down the tasks and um you know here's the responsibilities and uh we have we should think about like you know so everybody needs to lock in and right log in and uh you know handle like a workload from one one to the one one one to one one to the other and um that's how like people used to manage right so but but nowadays about like you know so uh in our case like one of the biggest lessons learned is uh uh we we have to hire the best people right everybody's supposed to have like a holistic role you know holistic uh view of the entire product right everybody can there's no ownership on code repositories and everybody's supposed to like a you can't change anybody's code as long as you run the evaluation set on each people's each person's like you know so repository right yeah you can't change whoever's like code code base as long as like you know you know what you're doing right and uh so that significantly reduces the time of uh collaboration think about like you know so I you know some some our employees like work from like uh 11 a.m to 2 a.m and some people actually work from like 6 a.m to 6 p.m right they don't like you know but how they work like you know so they can't just wait in the other person's time to like uh merge the PR right so they have to work they have to learn how to work as you know synchronously and uh in that way so if they have to like uh you know everybody knows the product every and everybody understands like the entire repository and uh you know and uh the bar is pretty high and so that you can't you can just like uh you know change whoever's code change what which with whichever part of the code and uh the do at at will um it's uh you know that in that sense like that that can be like a super efficient as a as an organization um you know which you know that's uh it's uh I'm I'm not sure we haven't hit that uh the block the the the little block of like a scalability yet but there might be a skin you know you know scalability issue once you grow up the team like to maybe like 20 to 50 people right so um I mean I'm not sure like whether it's this is this model is scalable and also is like limited by how you know how you can recruit like good people constantly right yeah awesome I know we're at time um just last question what's the buy what's the best advice you've ever received um I think I receive like good advice on a daily basis right either from our our our team members or like from investors or like you know so or our users right um I was thinking the most important thing is about like you know so you have to focus on growth I mean like most of the founders like I don't think most some of the founders have this belief that my product is the best our technology is the is the best but the thing is like even the best product dies because nobody pays attention to right yeah so you have to focus on growth all the time especially at this at this moment of the our you know of the entire industry user attention is uh very very difficult to grab you know we have to do whatever you know do everything we can to actually um you know to win to win over like user's attention it's a super challenge at this point yeah I I'm also I'm not saying I'm doing a good job and I'm I I I'm pretty sure I've been you know I'm I'm doing a really bad job on that and uh just like you know is it is it just a hard lesson is a lesson learning in a hard way yeah no I I I mean what we're seeing across different companies that we work with the the ones that really succeeded the ones that they're really focused on uh um building a community um building um um you know being obsessed with the customers um uh building a product that's really is highly differentiated and and amazing at what it does I mean like you like I was just looking at you at your website traffic you're probably one of the very very few companies that I have ever seen with this kind of traffic um because I'm just looking at like I mean it's insane it it's insane to see this kind of traffic for uh a a company um unless we're talking like a really really big brand but but that's that's different so congrats I'm obviously you're doing something something right and you're getting uh users' attention yeah thank you so you know so we're we're trying and um like you know so it's um you know so I I think that even me I I I stop you know I stopped doing anything else but uh growth right so 100% focus on that even our CTO uh is 100% you know it's like 80% you know focused on like a growth uh related like it related like a PR and uh I mean like you know we have to do this 100% you know yeah uh to to to survive I mean yeah absolutely uh Stanley it's it was an absolute pleasure to uh to connect with you and and learn about what you're doing uh keep on doing it because it's clearly needed and uh uh based based on the href stats you're clearly doing a good job in terms of getting uh people's attention thank you maxim thank you for thank you for having me i mean it's a it's a great you know conversation so i i i actually so it's a it's a it would it actually help me like a summarize like what I'm supposed to do I don't really have to do this monthly review anymore that's awesome I'm glad glad I was able to help yeah thank you thank you maximum yeah a good one yeah