#417 Revolutionizing Productivity: Yohann Abitann on AI-Powered Executive Assistants

[00:00:00]

Mehmet: Hello and welcome back to a new episode of the CTO Show with Mehmet. Today I'm very pleased joining me from France, Yohann Abitann. Yohann, thank you very much for joining me on the show today. The way, as I was explaining to you before we start the recording, is I like to keep it to my [00:01:00] guests to introduce themselves.

Mehmet: Tell us a little bit about You know, your background, your journey and what you are currently up to. And then we're going to take it from there. Just, you know, a teaser for the audience. We're going to talk about something really nice. It's about how we are integrating or how Yohann actually is doing that AI with productivity.

Mehmet: So Yohann, the floor is yours. First of all,

Yohann: Mehmet, thank you so much for having me. It's a pleasure to be here and with your audience. So a little bit of background about myself. I'm, I'm Yohann Abittan. I'm the founder of Newmail AI. I'm Swiss French entrepreneur. I've been working in artificial intelligence since 2017.

Yohann: I originally did my Bachelors in computer science and cybersecurity at NYU. And then I started working in cybersecurity. But I also had an experience during my bachelor's where I worked in the Sanford Research Institute, which is a big research center in California. Actually one of the oldest in Silicon Valley has a, has an exciting history.

Yohann: They were [00:02:00] responsible for creating the computer mouse. They participated in, I think they were the fourth node on the initial internet, which was called ARPANET at the time. And so I was in this institution and I had the chance to work with STAR Labs. They were a speech analysis lab that had created innovation that turned into Siri.

Yohann: And we were a kind of a second wave of startup working with them, licensing their technology. It was called auto dot AI. And we were doing real time speech analysis back in 2017 for call centers to monitor sales conversations and generate real time insights to the contact center people so they could have a better mapping and understanding of the conversation and know what to say, when to say it.

Yohann: And when we worked on that project, I just fell in love with ai. I saw this incredible technology that could take some rich, unstructured data, like a voice in this [00:03:00] case speech, and, and break it down into its components, extract value, and then do this at scale. And the magic of auto was this scalability, the ability to not only understand, conversations in real time, but to do it for millions and millions of people. And at the end of the day, that company was acquired by unity the big gaming company to monitor the voice channels of, I believe it's over a hundred million video game players keeping them safe with the technology is now called safe voice.

Yohann: And so this was what made me fall in love with AI. And after my stint in cybersecurity, I decided to work in AI again. I did my master's in AI products at Carnegie Mellon, and then I started working as a first product manager and then head of product of a company called load star, which was no code computer vision platform for the enterprise.

Yohann: The team was from NVIDIA. And so, we were working with these big a 100 [00:04:00] servers and we could actually create labeled data sets of very high quality video data for the purpose of detecting objects, detecting defects in the context of manufacturing, and we also applied it. to protection of biodiversity.

Yohann: And I can go more into that a little bit later. And after that experience, I decided to come back to Switzerland. I'm actually based in Switzerland now and start my own company which is called new mail and new mail is the AI executive. Assistant for entrepreneurs for leaders. And what new mail is and does is an agent that gets to know you.

Yohann: That's very easy to train a natural language. And then you mail will actively manage your inbox for you and manage your calendar. So that you can focus on what matters most to you. Entrepreneurs especially have big challenges around their inbox. You know, that's where so many opportunities lie. And yet it's still like, the 1960s.

Yohann: [00:05:00] You know, you have a stack of letters and it grows every day and everything looks the same and it's mostly bills. And it's not so fun as it used to be once upon a time when your friends would write you nice handwritten letters. And so to modern challenges, you know, we need modern solutions. So we're bringing your AI executive assistant that works 24 seven for you.

Yohann: I'm making sure that only the most valuable, important information is presented to you. And along with that, the insights that you need to make the right decisions so that you can act swiftly and make the best of your inbox, make your inbox work for you, leverage the opportunities and don't leave any value unexploited.

Mehmet: Wow. What a what a journey and what an experience you are like, really, you know, very rich. And I don't know where to start to ask you, actually. But let's let's start from from Newman, right? So, how this fits because one of the biggest [00:06:00] challenges people Talk about especially when it comes to email is achieving inbox zero, right?

Mehmet: Maybe i'm different from the others, but there's no One I know at least in my close circle, you know, everyone has smartphone, right? so when they bring their you know phones the first thing I have I have a look at is You guess the email, you know, I can and how many ones are there and i've never seen someone who's less than You 500, 600, 700, right?

Mehmet: And there's some guys like, when are you going to have time to go over that? They said, we don't know. So with what you are currently doing, Yohann, and with the power of AI, as we were mentioning, how this going to make people's life easier. And you know, what are like maybe some other use cases is like you talked about the stack and you, you, you're right.

Mehmet: It's like the old style, you know, I would have on my desk bunch of, you know, [00:07:00] paper letters. So I want also you to tell us more about the use cases is there and you know, how, how your solution can let us have a better productivity.

Yohann: Yeah, absolutely. Absolutely. You mentioned that when you look at someone's phone, they all always have 500, 600 emails.

Yohann: I actually offer a bounty to whoever can prove to me that they have the high score of the most emails I've seen up to 150, 000 unread emails in an inbox. Wow. You know, modern challenges, email was built for a time when that was a channel for important communication. And like we are in the attention economy today, everybody uses that channel to try and get a piece of your attention.

Yohann: And we have this concept of spam filters, you know, this idea that, If a company sends you an email that's for a million people then we should [00:08:00] detect that and we should throw it away or if some scammer is trying to to steal some information from you and they put some dangerous links then that is a spam and you need to throw it away, right?

Yohann: I think that was the The like 2000s, 2010 way of thinking about this email problem, but beyond that, we're way beyond that today. Even if you can catch all this spam, there's still this issue of clutter. And especially for leaders, anybody who wants your approval in your organization or who just doesn't want you to feel that your ego is slighted because you didn't see the information, you end up being cc'd.

Yohann: Copied on hundreds and hundreds of conversations that are not spam, that are not necessarily written by robots. And yet like you said, when are we going to have time to get through this 600? The idea is if you think you have to get through every single email, you'll never have the time. But this is a problem that's existed [00:09:00] for a long time.

Yohann: And high profile people have always had many, many people fighting for their attention. And the traditional solution was to have a person who is your executive assistant who acts as a shield, Between you and this external world that really wants to encroach and get some of your focus some of your insight You know, I I really don't like the sentence people use sometimes.

Yohann: I just want to pick your brain my my brain is a valuable thing and Don't pick it up. You should You can but you have to extend it Exchange some, some value in return. And so I think this same model of having a shield that makes this distinction of what is high priority for you? What is the right time to show you what information I call the guardian of your attention.

Yohann: New mail is the guardian of your attention. New mail in a few sentences of natural language will actually ask you, what's your role, what do you care about? What do you see too much of in your inbox? And then there are many advanced [00:10:00] features I can get into, but that's the basis. And with your answer to these few questions, Newmail AI will then priority rank your inbox so that you have time sensitive and important communications on top, things you need to act on and, and that are relevant to your role.

Yohann: Then you have Important threads that are relevant to your role could contain valuable information that you might need now or in the future. And, and those will be marked as important. And then beyond that you have things that are interesting. You know, maybe it's a newsletter that you subscribe to, but it's not on the specific key topic that you care about.

Yohann: So it's not important. It's not clearly important, but maybe it could be interesting. And it's clearly not time sensitive. Right. And then beyond that, you have what we call clutter, which is all the things that you've identified, this type of thing, there's too much in my inbox. Maybe it's cold calls, maybe it's marketing, maybe it's too many newsletters from the past.

Yohann: Maybe you're a CIS admin and you have [00:11:00] a crazy amount of notifications from a specific system that keeps getting triggered and it's difficult to. Disambiguate, you know, what you want from what you don't want to see well, you can just tell it in natural language to your AI assistant and your AI assistant will make sure that all that clutter is not even in your inbox is marked as red, and you can make sure that all these interesting things are sorted appropriately and summarized and will be delivered at a time that you have time for it.

Yohann: Same with everything that's important. Everything that's important is summarized delivered on a daily basis It's also highlighted and then what's urgent and time sensitive is most highlighted And then beyond that we also create a to do list that's integrated inside of the google ecosystem because We started with gmail Gmail has some good sophistication that we've built on to have some nice advanced features.

Yohann: So using Google Tasks, we have an integrated side panel, and this side panel tracks all the [00:12:00] communications where you still have actions that need to be handled, linked to the specific email. So you don't have to do a search step. All you have to do is quickly scan through your list, see anything you want to work on, hit the link, and then in one click jump into that conversation, and then you can manage it.

Yohann: So that's the basics of what email does for you

Mehmet: fantastic, you know, You triggered in my mind. I prepared different sets of questions, but I I always love when when you know, it becomes like more Casual conversation and spontaneous one. I remember because you gave one use case, which is very interesting, right?

Mehmet: So and I was hearing on on the radio a couple of days ago because you mentioned about you know, the the scams and you know, the the These kinds of things that happen over email. So yeah, I just remember, so it's here in Dubai, you know, the guy who's responsible for finding companies in the Dubai financial center, the [00:13:00] IFC, he was talking about like how, you know, these guys, they became better with AI because they are drafting their emails in a very, very, you know, good way because traditionally these people, they are not You know, English is not their first language of course.

Mehmet: So they are using AI to proofread and enhance and so on. And I just remembered also, if you remember, Yohann, a couple of years ago, there was this chatbot in New Zealand that it was designed to waste the time of the scammers. So, but it's good use case because you know, like if, if I can remove this also from my inbox and I don't have to go with it, Fine.

Mehmet: Now, the second thing I wanted to ask you about, so this is a very good use case. The second thing regarding, so you mentioned newsletter and you mentioned marketing stuff. So sometimes, and I'm curious how the AI would be able to learn this. So sometimes I go and make an inquiry [00:14:00] about something, or I subscribe to this newsletter.

Mehmet: Many times, to your point, I find it like after some time, it's not interesting for me, but I forget to unsubscribe. So. But sometimes I need to go back to it. So how AI would, you know, would help me here in understanding, you know, really, you know, what I mean is how it's gonna judge, really. You know, this is really interesting.

Mehmet: Is it by how I open, how much I scroll or is it like by the content itself? So I'm curious about, you know, how, you know, the, the model will be able to, to judge, like, if what I'm doing is really, sorry, what it's doing for me is really what important or not, if, if I am clear on the question.

Yohann: Yeah, yeah, absolutely.

Yohann: So there, there's two key pieces that you bring up. One piece is about learning. How does it learn from your behavior? What kind of signals is it, is it paying attention to? And then you bring up is it about the [00:15:00] content or, or, or what is it about the data that, that gives the AI this understanding of whether or not it's relevant to you?

Yohann: And the answer to both of these kind of gets to the heart of how we built NuMEL, which is. I come from a cyber security background and I come from an AI background and I always realized the AI is only as good as the data that you put in it and that's basically a function of trust. Then beyond this, I think there's a lot of talk today about just cramming your AI full of so much data.

Yohann: And I'm, I'm actually against that. I think it's about putting in the data that's really valuable and sensitive. That's going to enable your AI to have that competitive advantage to understand what you do and how, what you do is different. But built into that notion, this is very sensitive.

Yohann: Professional data and how to manage it in a responsible way. So part of that it directly goes against this idea of what I call data maxing. Just saying, [00:16:00] oh, just throw all the data in there and the AI will, will learn and we'll figure it out. That's never been a good way. To work in AI, I guess open AI and the scaling laws has made people think, you know, we need internet scale data set, but that's not true at all.

Yohann: If you want an effective decision making AI, and then what you need is the right data and the right labels. And in this case, we're talking about sensitive professional data. So that data needs to be handled in a, in a very privacy conscious way. So we do look at content, and that's what makes it different than the traditional.

Yohann: We'll just look at the headers and patterns in the headers and also like open behavior. So we actually go to the next step, and your AI will read the actual content. So, I can determine in a newsletter. So we do have like context about sender level, and we have context about content email level.

Yohann: So, on, on the one hand, we'll see if over time your behavior with respect to that sender, if that sender [00:17:00] sends you multiple low quality emails, then we're going to proactively suggest to you, we think that you're actually not interested in this, should we unsubscribe you and on the flip side, if somebody sends you multiple high importance things, then we'll suggest you, you add them as a priority contact.

Yohann: But apart from this one behavior. Almost, I would say everything else in our platform was designed to be very transparent to the user so the user can see it, manage it, and delete it anytime they like. And what I mean by that is, I told you we answer a few natural language questions. That is what we call the preferences.

Yohann: And that's the one thing that we store of user data. On our systems, preferences are answers to these seven questions with some extra settings. And this is stored encrypted in our system and then everything else, the signals that we take are once again, explicit. So we have an AI where in, in the beta on this aspect of the system, which is [00:18:00] through a conversation, you can tell the AI what matters or doesn't matter to you.

Yohann: So if it delivers to your daily briefing and says this, this newsletter should be interesting. And you say, actually, I'm, I'm not so interested in that one anymore. All you need to do is exactly what I said. Tell your AI assistant, that's not interesting to me anymore. And then we will understand and recategorize because we could learn implicitly, but what we decided to do is.

Yohann: You need to know where every single piece of your data is. So we don't want to have feelers and analytics that are not open. Everything that we do is open and transparent. So it's very easy to audit. That's how we were also able to voluntarily take on the highest level of security audit that Google offers for workspace applications.

Yohann: And we, we keep being very proactive on, on the aspects of compliance. You know, we're European based, so GDPR is very important. And GDPR, the foundation of it is every [00:19:00] user needs to know where the data is. They should be able to manage it, delete it at any time. And so that's, that's the key for us. So we learn in a very.

Yohann: Explicit way that is the same way an assistant would learn from you traditionally is through a conversation When there are signals that come out about decisions that have been made that seem like they could be enhanced we call it guidance and I think that's the right way to evolve and work with ai systems today It's not full automation, but I would say it's almost automation with guidance or with supervision.

Yohann: I think that's the best way to let the AI run, not need to prompt it every single time but then be able to say, actually, this would be a little bit better if you can do it this way. And then that way, you know, it's extremely clear what learning happens, when and why. And you can edit, understand, explain it.

Yohann: I used to be obsessed with AI explainability and natural language models sometimes are very easy to explain because they can [00:20:00] talk to you. And sometimes they're very mysterious and we're trying to challenge that notion.

Mehmet: Very cool. Just out of curiosity, very quickly. So your solution can fit individuals as well as businesses.

Mehmet: Am I right?

Yohann: Yeah, that's right. So we, we say that we work with leaders and teams. And we recognize that founders, entrepreneurs, CEOs, they're good target audience because they have the strong pain. They have the most emails. On average, they receive 3 to 500 emails and they send more than 100 emails per day.

Yohann: So there are a lot of roles that don't do as much of that. Also their email, the value per email is quite high and there aren't so many excellent software tools out there targeting this area. They also have many contexts. So it's more complicated to deal with an entrepreneur's inbox than to deal with a marketer or sales was more defined role where the opportunities look more.[00:21:00]

Mehmet: And out of curiosity again, can I connect multiple inboxes?

Yohann: Yeah, absolutely. So that's one of the aspects we're working on is making the unified inbox experience more seamless. Currently we're creating one assistant per inbox, but very soon we hope to unify all of that. And I think that the next evolution that we do will, will really facilitate that because we're going towards this more agentic model, like I mentioned, of having an AI that you can just chat with conversationally.

Yohann: And so I think we'll disambiguate based on the, on this aspect of one phone number manages multiple inbox accounts, and that will be how we identify you. And then. All your inboxes shall be made easy to, to synchronize and unify through that assistant, because as you mentioned you know, the inbox is evolving and how do we make the best of this inbox?

Yohann: 10 years ago, the solutions that were developed it was all about get through your inbox, super fast, [00:22:00] get to inbox zero. We actually believe that with AI, we can get people out of their inbox completely and into. Just give them an assistant and that assistant becomes the interface and you can control your digital experience through that assistant.

Mehmet: Fantastic. You know, I'm looking forward to see because you know, like I know a lot of people and you know, sometimes I'm into that spot also as well. And this is why I asked you, Yohann, about, you know, multiple inboxes. So let's say I have my personal inbox. Maybe. I'm also having the work inbox. I know some people they would have like maybe advisory roles with multiple companies also as well.

Mehmet: And they need, you know, to, to connect to multiple places. This is why I asked you now, let's discuss the calendar because calendar, in my opinion, is one other underrated part of our productivity. Of course there were tries. To automate things [00:23:00] like with all the solutions that you can do scheduling with it, but still, I feel there's room for improvement, especially with the AI.

Mehmet: So tell me, Yohan, what, what are you building for that use case?

Yohann: Yeah, absolutely. As you mentioned, you know, there, there have been like a whole family over the last 10 years of scheduling tools. Send me your link. I'll, I'll book on your calendar which makes it useful for, for many use cases. But it's still not the most seamless because you know, there's still billions of people using email every day.

Yohann: Gmail is 1. 8 billion daily active users. And I don't know how many people are using Cal or Calendly. But good salespeople know that you don't want to leave the effort on the other people to, to go and find time when you're trying to close a deal, right? So, What we do is we go to the next step, and when somebody wants to schedule a meeting with you, we identify that, and then, I didn't mention, but, you know, you said that scammers use AI to make their draft, their emails better.

Yohann: We do the same thing, right? We generate reply [00:24:00] emails, high context, based on your role, responsibility, the context with the person, and in the conversation, and then we write a high quality draft for you. So in that draft, we're going to include availabilities in your calendar to meet that person. And then through what I mentioned of the agentic interface we can ask you to validate, you know, like we found these three slots or three of these good to share with that person.

Yohann: And you can say yes, or you can say, actually just share these two or give me three different ones. And then we can take care of looking through your calendar for you, finding that availability, sending it by email, and then all they have to do, they don't have to search and look at 10 slots and look at their own 10 slots.

Yohann: They have three and then they say, yes, yes, no.

Mehmet: Cool. That's super cool. So even the remove the friction of sending the link and then waiting them and then coming. And I gotta tell you something really funny. So I had to struggle a lot, you know, in because [00:25:00] maybe when you booked this, you've seen like how much time I put to customize the, the invitation.

Mehmet: Because, you know, if I go with the default mode, it's saying it will gonna say, Hey, Yohann and Mehmet meeting. That's it. So, and I don't have any context. And actually the person who brought this to me is one of my guests, right? So he said, Mehmet, like, okay, I'm seeing this calendar. I have no clue. What is it for?

Mehmet: I forget. I said, it's a, it's a podcast recording. I said, okay, why you don't have that? And then, you know, I've built my, I had to do it myself. And then, you know, I start to put also the link for StreamYard and, but it was all manual work, but if you can automate this more, that's fantastic. Before I ask you about the agent part of it, just out of curiosity, and maybe I'm reading the minds of some of the audience, how this will fit in a world where, you know, the [00:26:00] providers themselves, they are trying also to push their Quote unquote, co pilots and their Gemini co worker and so on.

Mehmet: So is it like you are seemed like a threat, an opportunity? How do you look at it, Yohann? Yeah. You know, that's a, that's a great

Yohann: question. I think that. In the nature of startups, we're so different than the big giants that what we do, the nature, the DNA means that we find different opportunities.

Yohann: We're agile in different ways. So I've been to meetings to see the Gemini offering. I've worked with companies, done consulting for companies that were using copilot and I've seen what it's strong at. And I've seen where it is weak or slow. And these are not perfect solutions. Because big companies are not perfect and if you've talked to startup CEOs before like you have There's this thing that we say that if you were not going to do a software because google could do it better than you [00:27:00] Then there would not be any software being developed, right?

Yohann: but the reality is is different the reality is these are wide ecosystems that these players have microsoft and Google and what they're building is an ai that can serve as the connective tissue You Between their different platforms to help you move the data. But what they're not building is the best possible executive assistant.

Yohann: And that's what we're building. And I think always in startups, it always comes down to focus and it also comes down to customer obsession and fundamentally, you know, These days, when I work with CEOs, I work with teams we see eye to eye and I can, you know, they have my WhatsApp number and, and we have this very close relationship.

Yohann: I don't know if you've had some support issues with Google product recently, but I spent two months trying to get ahold of someone and after four failed AI conversations and 16 emails and support I [00:28:00] finally solved the basic issue. So I think there is room to, to grow and exist. I think everybody wants AI and the big companies, they wanted their market cap to go up.

Yohann: So they all said, now we have AI and everything. And I don't, I am going to be opinionated and contrarian, you know, I had a conversation with Gemini recently and I asked Gemini, what can you do for me? And Gemini said, I can write stories. I can come up with narratives. I can help you write things. And I said, come on, Gemini, this is not very innovative.

Yohann: And Gemini said, you're absolutely right. So they still have work to do. They have great people working on it. You know, there's this rising tide element. I think so many tools, so many things are coming out. Anthropic just recently put out a couple of days ago the MCP, their model context protocol made for [00:29:00] AI systems to collect, connect more easily with external data sources.

Yohann: There's so many things being done in the space. And it's becoming easier and easier to make better and better AI assistance. So I'm very excited to be building in this space. And yes, there are mastodons and these giant players but I'm obsessed with this problem and I'm not going to let it go. So we'll, we'll see what happens in the next few years.

Yohann: It's going to definitely be interesting. And I think, you know, it's there's no better way to participate in the future than to try and build it. Absolutely.

Mehmet: Absolutely. And, you know, of course I knew the answer, Yohann, but I wanted you to share what you did absolutely fantastic way. Explaining why, when you are obsessed in the problem, you are authentically and what I like to call it design thinking perspective, because you are building the [00:30:00] empathy with people who are going to use this.

Mehmet: Don't get me wrong. I'm not saying like People at Google and Microsoft, they don't have empathy. I mean, on as persons, but because they are giants and they are, you know, you know, building much, much bigger things, right? So few details it might be missed. And this is exactly where people like yourself, you know, and building these kinds of startups, it get me excited because you are obsessed.

Mehmet: And this is the, the, the keyword that you use over there. You're obsessed with the problem. I love this. Now, Just because you mentioned WhatsApp, again, it wasn't on my list. There are a lot of things that happens and let me, you know, maybe some of the audience in the U S they know I'm in the, in Europe, in Europe and in, in, in the Middle East, usually we are close in terms of how we communicate.

Mehmet: So there's a lot of business conversations nowadays. You're one that happens on WhatsApp, right? Then I said to tell the guy, Hey, send it over email. And [00:31:00] then, you know, in the email, Hey, based on our conversation on WhatsApp, Hey Give me a break. Give me a break. I'm not able to handle email. So now you're telling me go to the WhatsApp.

Mehmet: So is there any vision to try also to bring, you know, other communication tools to the, to the picture and not only WhatsApp. So I'm saying WhatsApp because it's very famous. At least I know in Europe. I mean, even in here in the Middle East, WhatsApp me this. You know, telegram is also on the rise, by the way.

Mehmet: In the U S there, yeah, they still like kind of I message and these kinds of things, but I mean, still there are a lot of things that happens outside of the email related to business. So what's your vision on that?

Yohann: I think you're very perceptive and these are really good questions. Actually the, the be the beta agentic vision that we have that we're currently building it is to, to bring your assistant to WhatsApp so you can have that personal one-on-one chat, synchronous conversation with your assistant and to, from your WhatsApp, be able to [00:32:00] pilot your inbox, your calendar, your contact, and the rest of your digital life.

Yohann: WhatsApp is. For 3 billion people is the biggest chat platform in the world. And I think there is this opportunity with AI to drop into existing workflows and, and reuse the same tools. You know, we, we really got saturated with apps. We all downloaded so many apps and then deleted so many apps. And now if.

Yohann: If I have to download an app to figure out why your product is awesome, you've probably already lost me. So we're working in going where entrepreneurs are today and where they're doing their work. And WhatsApp is huge. I do think that WhatsApp is an intimate channel. And I'm sure you've had this experience.

Yohann: I don't want to have too much business conversations, especially not initial conversations over WhatsApp. For me, that's a, that's a preferred channel, but we think we can earn trust with our users to earn our place in their WhatsApp as their executive [00:33:00] assistant, and then from their WhatsApp help them pilot the rest.

Yohann: When it comes to automating WhatsApp itself, it's still something that Meta is quite guarded about and with good reason, right? There's little spam on WhatsApp so far. It exists, but I've not been a huge victim of it. And that's because they keep it close to the chest. It's not easy to do automation on WhatsApp.

Yohann: So currently they don't allow bots or business accounts to be part of groups, for example. But definitely the next level agentic communication stack that we're building that many others are building. I'm sure that WhatsApp will fall into this. I know that Slack, owned by Salesforce now, they have many like AI features and you can imagine AI features for chat as well.

Yohann: It's pretty straightforward. And WhatsApp itself has WhatsApp AI. Which lets you talk to chat GPT within the WhatsApp [00:34:00] ecosystem. So I'm, I actually wouldn't be surprised if WhatsApp comes out with a premium B2C offering for their 3 billion users where for a few dollars per month, you actually get to have some more powerful AI features in your WhatsApp.

Yohann: So that being said, you know, we're, we're, we're going to be there and we're going to have all the elements and all the tools. And if it turns out that that's something that entrepreneurs are passionate about, then we'll try to solve that for them.

Mehmet: Absolutely. My number one, you know, feature that I want now, I don't know who going to do it as a summarizer for the WhatsApp group messages, like we, we have, you know, some friends groups and.

Mehmet: You know, we work during the day. We come, you know, sometimes I don't open the WhatsApp, you know, the whole day because I have meetings or maybe I'm recording now the episode and I come back and I found like 60, 70 messages and say, guys, what we're chatting about is this, is there an AI to just tell me what happened?

Mehmet: Wouldn't I miss? [00:35:00] And really it's becoming sometimes annoying because of course, you know, especially with the, you know, everything becomes kind of formal and, you know, Oh, what did I miss? Oh, should I have been informed about that? And it's overwhelming now because, you know, I want really to discuss this with you, which is related to the agentic approach you're talking about The audience knows by now how passionate I become when I talk about agents in AI because When I interacted with chat gpt the first time, okay, of course, it was like a moment I said, okay, this is something big because I remember when I was kid always I was you know Imagining someday a robot which I can talk to or type to, and it knows everything and it can, you know, inform me and educate me and which is ChantyPT kind of do it.

Mehmet: Now the second thing which I [00:36:00] started to think about, okay, if, if the model can generate text. Regardless of the T model, we can discuss the technicalities, but I mean, it can get some text in natural language, you know, and translate it to something that we can really think, actually we don't think, it is really something readable by a human and it has meaning.

Mehmet: Like even when you were saying about Gemini, telling you like I can write stories and so on, and yeah, you are right. It tries to imitate a human, which is okay. Now, taking it to the next level and taking actions is where Agents comes into the picture. And the first time I started to see, I think, It was early 2023.

Mehmet: First attempts, you know, people doing the baby AGI and

Yohann: yeah, yeah,

Mehmet: yeah. I said, hold on one second. Like this is okay. So if chat GPT was the thing, no, forget about it. Like this is the next big thing because now I'm having an army of bots or You know, agents [00:37:00] doing stuff for me and all what I need to do is just, you know, to give them the plan and you go guys figure it out.

Mehmet: So tell me a little bit more about this aspect, especially with, you know, the work you're doing currently on productivity and helping CEOs and founders and executives.

Yohann: Yeah. So just as you mentioned, right? Like, having an AI that's capable of being triggered through a conversation in a way, it's the evolution of the interface and it's very powerful.

Yohann: It's flexible and it's currently in its current form. It's limited because the AI is only being triggered when you ask it a question, right? So I'm sure technical people like yourself and like many in your audience, we've used chat GPT integrated to spreadsheets. You know, when you're able to ask 100 questions at once with all different parameters, then you're starting to have hours of time saved and not just minutes or, or seconds, [00:38:00] right?

Yohann: So I think on that aspect of how many questions Can we afford to ask AI? That kind of subselects which problems we can use AI for. And so we've been thinking about proactive AI. AI should know what questions you ask yourself before you do it. It should answer them for you and display the answers so you don't even need to ask yourself that question.

Yohann: Remove that, that weight off your mind. So we have questions like, is this email relevant for me? Is it time sensitive? Are there actions I need to take that need to be tracked? Do I need to write a reply to this? These are all questions that our AI answers every single time you receive an email. Beyond that, I mentioned to you, I don't think we're ready yet with the error rates.

Yohann: have general purpose solutions that can really go out in the world by themselves with no supervision and perform high quality work. So I think for me, one plus one equals 10. And we are in [00:39:00] the phase of AI, human, Assistantship or collaboration. I think co pilots have kind of stolen this thunder of this idea and maybe like not really delivered on it so far.

Yohann: What we're building, as I mentioned, it has aspects of what you say in terms of planning, what needs to be done and then going and doing it, but we still keep the feedback loop pretty tight. So, for example, I'll give you a sense of one of the things we're working on with our agent. If you say organizing a meeting with john next week, thursday afternoon, then we have to do a number of steps.

Yohann: So first we recognize this intent of Is this what you what you're saying you're saying? Okay, find john. So who is john? We look through your contacts We understand there are eight johns, but in your recent communications. This one is the top hit Maybe we ask you this john on this email. Is that correct?[00:40:00]

Yohann: And thursday afternoon you mean after you go to the dentist so around 4 p. m At earliest Right? So we have this type feedback loop where you express your intent. Then we will do the search steps. We will chew it up. And then what we want to do is present to you the decisions as yes, no decisions. So it's as easy as possible for you to make your AI do something, but you're still supervising each key decision step, because as I mentioned, I don't think in terms of where we are today, maybe six months, 12 months from now, there will be innovations and we'll be able to have more fully automated.

Yohann: And we're, we're building with, with that in mind that the AI will keep getting more competent. But today it's for me. A short leash and well defined tools and intent. But when you bring that all together, then you have an AI and it's especially interesting when you start to look at [00:41:00] multiple applications, right?

Yohann: So I mentioned contacts then we have checking the availability in the calendar and then we have, you know, creating an event and then sending them an email and saying, Hey, John, Here's the link. Here's blah, blah, blah. Let's, let's meet at that time and then summarizing perhaps the conversation or that email and placing it in the event, as you mentioned.

Yohann: So all these back and forth between different tools, if we can unify that onto one interface of your conversation with your agent. I think we're saving a lot of time, a lot of decision power, a lot of context switching without actually doing too many unforeseen hidden steps so that we can still keep that very, very high quality and that signal to noise ratio from intent to result or output.

Mehmet: So it's work in progress, Yohann.

Yohann: Yes, the agent that we're building, we're aiming to launch at the start of the year. But currently we're still building it. [00:42:00] And I think that agents as a whole is a work in progress in the sense that we're still in the phase of human AI collaboration and supervision. So when I work with companies and they tell me, Oh, I'm trying to understand how can I use AI?

Yohann: How can I benefit from this technology? What I tell them is identify problems. Where supervising the task is much easier than doing it from scratch. And one example I give is, if you're going to interview 10 people and you have 10 CVs, maybe taking those 10 CVs, giving them to AI and say, AI, rank the three best and then generate relevant questions based on their CV for me.

Yohann: And then you yourself can look at the questions and choose the ones you like best, and then ask them in the context of the interview. That means that you removed a lot of repetitive tasks and you didn't have to, to read the source data and then make your own determination and then [00:43:00] write it from scratch.

Yohann: You're just selecting, okay, actually, you know, your experience at Microsoft on distributed systems seems relevant to this project. How do you think it will help you? That is a personalized solution that you're delivering that lets you get deep with someone quickly. And it was much easier to have AI suggested than to come up with it yourself.

Mehmet: Absolutely. You know, like, I like, you know, when, when the aim is also to, to give us time, right. And, and remove the friction of doing repetitive tasks. People thinks that, Oh, repetitive tasks, you know, like it's repetitive. It's like a routine. No, actually. I know from myself and I noticed other people, the people who does a lot of repetitive tasks tends to do more mistakes.

Mehmet: And if you give people like things that change all the time, of course, there will be some mistakes, but I mean, they will do less mistakes because you know, they will, they have to think and they have to go [00:44:00] over it, you know, so they have to notice when you do something repetitive. And I know, again, I, you know, I give example from myself.

Mehmet: It's not like I start, I stopped, you know, noticing I said, okay, so, and then I figured out that I did a lot of mistakes because I didn't automate. And I'm, you know, the audience know, I'm big fan of anything automation. And of course, AI, I tell them the current AI is automation actually. So it's, it's going to, to help us, you know, automate things that, you know, we need to repeat it.

Mehmet: You want to

Yohann: say something? I think, I think, you know, once again, it's that human AI collaboration, that one plus one equals 10, because all the tasks that we do, they have repetitive aspects and they have aspects where you can really bring value. And so what we do is we try to do the repetitive parts of the tasks so that you can add your personal touch at scale.

Yohann: And what that looks like in the world of emails is a little bit more empathy and a little bit more strategy, right? Yeah. That's it. If somebody asks you for 10 information points [00:45:00] and your AI can put a draft together that says all these things, does the nice greeting, does the nice sign off, then all you need to do is jump in and say and add one line and say, by the way, I hope that this other project you told me about is going great.

Yohann: Here's the information and that's what we say. Personal touch at scale. We're not trying to replace you. We can't replace you We're trying to give you more time to be you and let the machine do what's for machines

Mehmet: absolutely Yohann like as you are coming to an end just I have you know a question for you And you know with all the passion that you have and you know, I I am Kind of really appreciating the way you truly understood and and you you know exactly what are the pains and the problems that You know your ultimate I would say customers would have so And I told you, you know before we start the recording so part of the audience are people who are [00:46:00] aspiring entrepreneurs So what advice you would give them if they are looking to turn a technology idea into a market ready product?

Mehmet: Because really you are onto something Super, super, super hot, I would say. So what you can share from your experience.

Yohann: What I would say is it's very rare that you have an idea that turns into a product. I would say go, if you want to be an entrepreneur, go and talk to lots of people around you and, and find something really painful that is in a space that you care about. And if you see a strong pain in a space that you really care about, start asking people questions and understand that pain better, because the pain that you feel is going to be different than the pain that they feel.

Yohann: But if you, if you can find that, that core of people who really care about the issue then you can have that as that burning flame to, [00:47:00] to sustain you through this difficult, but exhilarating adventurous journey. I think most startups fall apart when there isn't that strong connection to the pain.

Yohann: And some of the smartest minds I met in Silicon Valley told me, and this will always stick with me, become married to the pain, not to the solution. I think that's the best advice I can give to, to entrepreneurs starting, find the pain because if there's no pain, it's, it's not worth it. It's only going to be pain for yourself and there won't be enough enough gratitude involved in the system to keep you going fast and far.

Yohann: And it's a marathon and not a sprint. And I would also say to aspiring entrepreneurs, you know, my door is open. You have my, you have my name. Now you can find me on LinkedIn. I'm more than happy to, to answer any questions, provide any guidance that I can with pleasure. And then maybe the last one I would say, find the community.

Yohann: Because being an entrepreneur is a lonely game. You can't always share [00:48:00] everything with your team. Others, your friends might not necessarily understand. Find communities of entrepreneurs. Find people who are building things. Talk to them. Find out about their lives. And and use that also when you decide that this is what you want.

Yohann: Because it's, it's good to be the boss. But it's also a lot of hard work and responsibility.

Mehmet: Absolutely. And yeah, so, so I, I hear this a lot and I was lucky also to work with a couple of in a consultant role with, with the, with the great minds at startups. And yeah, so I've seen them, as you said, I call it in another way, like falling in love with the problem, not with the solution.

Mehmet: And we have same background. So I, I used to be a pre sales guy or sales engineer, and I always, you know, Telling people don't open the slide on showing features and you know Why you are better than this because you have a faster thing Okay, these are important but try first to understand the pain of [00:49:00] your customer and then from there you can Your life will become much easier actually because guess what?

Mehmet: You don't have to talk the customer will start to tell them to tell you your pains I mean their pains and then You know, your, your, your life is much easier. And then you, you, you validate at the same time. Okay. The hypothesis that I developed is right. And now it's the time to, to show them, okay, I understand you have this problem.

Mehmet: This is what we have developed. Let me see if this gets something, you know, something you can benefit out of it. So a hundred percent, you are on this. And thank you for mentioning, you know, because I was going to ask you where people can find more about you and also about the new mail. So just, you know, The link, I gonna put it in the show notes, but yes, please.

Mehmet: Also, I gonna put email, mail,

Yohann: ai.

Mehmet: I'm gonna put new mail AI also in the show notes. You are really, really, I enjoyed this conversation. Like one of the, you know, always, you know, I would say exciting times for me when I meet people like yourself, passionate about what they're doing, understanding, you know, what's, what's happening around having [00:50:00] the vision for the company.

Mehmet: So I appreciate, you know, the time you shared with me today and with the audience and, you know, this is for the audience. This is usually how I end my episodes. So guys, if you just discovered this podcast by luck Thank you for passing by if you enjoyed what you have Seen or you've listened to today give us a thumb up Share it with your friends and colleagues and if you are one of the people who keeps coming Thank you for passing by.

Mehmet: I really appreciate it and keep your also recommendations and your You know comments coming I read all of them and I take them into consideration Thank you very much for tuning in and we'll meet again very soon. Thank you. Bye. Bye