…I really like the idea that everyone can do things. And here’s what I mean by that. I have a lot of really close friends who are incredibly smart, but for some reason, they don’t feel like they can go out into the world, build things, and make an impact. I honestly think everyone can. Deep down, I believe everyone has the potential to be an entrepreneur in some way. They can build things — not just companies, but for themselves, for others, for their communities. With AI, for example, I feel like the barriers are coming down even more. Everyone will have tools to create and solve problems, and I want to see more people stepping into that role. I think it would be amazing to see more entrepreneurs at all levels, solving problems in innovative ways. For example, take something like the LA fires. Imagine if those fires inspired hundreds of people to start building solutions — firefighting robots, drones, sprinklers, or some other kind of technology. That would be incredible. And the thing is, anyone can do it. It doesn’t have to be limited to just a small group of people in elite circles…
I had the pleasure of talking with Kevin Yang. Kevin is a technology entrepreneur and artificial intelligence innovator, recognized for his contributions to developing AI-powered tools aimed at enhancing productivity and streamlining workflows. He is the co-founder of Serif.ai, a company based in the San Francisco Bay Area, which focuses on building AI-driven executive assistants designed to simplify tasks such as email management and scheduling. Serif.ai, founded in mid-2023, represents Yang’s vision of a personalized AI assistant that acts autonomously to help users regain time and reduce stress in their daily lives.
Born in North Carolina and raised in Arizona, Yang’s early interest in technology was shaped by personal circumstances. As a teenager, he began building software to assist his father, who lost his vision in an accident. His first projects included creating web automation tools to simplify online navigation for tasks such as replacing a lost credit card. These formative experiences laid the foundation for Yang’s enduring focus on automation and usability, inspiring much of his later work in AI.
Yang attended Stanford University, where he met his co-founder and began exploring entrepreneurial ventures. During his time at Stanford, he gained hands-on experience through internships at Meta and HealthTap. At Meta, he contributed to the Applied Machine Learning team, where he worked on AI systems to detect and interpret visual content. At HealthTap, he focused on improving patient adherence to medication by leveraging natural language processing. These roles honed his technical skills and introduced him to practical applications of machine learning and AI in real-world contexts.
While still at Stanford, Yang co-founded his first startup, include.ai, a no-code web automation platform that aimed to simplify online workflows. The company operated from 2019 to 2022 and marked his initial foray into developing tools designed to enhance digital efficiency. His entrepreneurial spirit continued to thrive through freelance consulting projects, where he worked with organizations such as Starbucks’ Siren Labs to explore geospatial AI-powered marketing solutions.
Yang’s personal projects have also gained attention in the open-source community. In early 2023, he developed LLMParser, a library for extracting structured data from unstructured text using large language models, and EmailTriager, an application for automating email responses. Both tools reflect his dedication to reducing manual effort and improving the accessibility of advanced AI technologies.
Throughout his career, Yang has been drawn to the challenge of creating systems that integrate seamlessly into users’ lives. He has emphasized the potential of AI to transform productivity by transitioning from tools that require manual input to proactive agents that can anticipate and execute tasks autonomously. His vision for AI extends beyond efficiency to inclusivity, aiming to empower individuals who may lack the technical skills to fully leverage existing AI platforms.
A vocal advocate for the transformative potential of AI, Yang has also expressed concerns about the industry’s future. He has highlighted the growing challenge of distinguishing between human and AI-generated content on platforms such as Reddit and TikTok, which raises questions about trust and authenticity in digital spaces. He has called for the development of advanced safeguards, likening the need for AI-specific safety measures to the introduction of seat belts in cars.
Yang is optimistic about the potential for AI to democratize access to tools traditionally reserved for large enterprises or wealthy individuals. With Serif.ai, he seeks to expand the market for executive assistance, making it accessible to entrepreneurs, small business owners, and professionals who could benefit from automated support but cannot afford human counterparts. His approach combines cutting-edge AI technology with a deeply personalized user experience, setting it apart from larger, generalized platforms like Google.
Outside of his professional work, Yang’s vision extends to inspiring a broader culture of innovation. He hopes to see more individuals empowered to build solutions for pressing global challenges, from environmental crises to public health issues. By lowering the barriers to entry for entrepreneurship and innovation, he believes AI can play a pivotal role in shaping a more equitable and creative future.
Yitzi: Before we dive in, our readers would love to learn about your personal origin story. Can you share the story of your childhood and how you grew up?
I started getting into the idea of building companies a long time ago. I was born in North Carolina, but we moved to Arizona when I was about two or three. When I was a kid, my dad had an accident and lost his vision. Some of my earliest memories of working on anything tech-related were actually from high school, when I started building different kinds of software to help him.
If you think about not being able to see, it’s incredibly difficult to do things online. For example, I remember one time we lost his credit card, and it was such a challenge to navigate the Chase website, click through all the menus, and order a replacement. That’s when I started figuring out how to create web automations to handle tasks like that.
The idea behind what I’m working on now is really an evolution of those early ideas. There are so many things you do on the internet, or on your computer, that you shouldn’t have to do manually. You shouldn’t have to write every single word yourself if AI could take care of parts of it for you. And there are tasks that could just be automated entirely by AI. That’s been a big focus of mine for a long time, and a lot of it has been inspired by helping my dad.
On the company side, I met my current co-founder during my first year at Stanford, and we’ve been working together ever since. This is actually our second company, and it’s been such a fun journey. You’re catching us at an interesting moment because our servers are a bit overloaded right now, so we’re in the middle of trying to fix things. It’s kind of a crazy, fun time!
Yitzi: Before we get into AI too much, can you tell us some amazing stories from your work and your businesses? Could you share one or two of your favorite professional stories — ones that really stand out in your mind?
I’ve got some hilarious ones! During college, we actually started our last company in our final year at Stanford. While we were trying to get users to try out the product we were building back then, we figured out a way to get the emails of all the students at Stanford. Then we emailed them about this tool we’d built to help them automatically enroll in classes.
The product let you pick your classes, click a button, and it would repeatedly hit the “enroll” button for you. We got enough students using it that when enrollment opened, it completely crashed the Stanford server. No one could get into their classes!
We ended up in quite a bit of trouble for that, but it was our first glimpse into how some of these automations could actually make an impact. It worked for that one very specific use case — some students got the classes they wanted, but everyone else had to wait until the servers came back online.
That’s one of my favorite stories from those early days. Back then, we weren’t even a formal company yet — we were just experimenting, trying to automate little tasks here and there. It’s hilarious to look back on, and it’s one of those moments that really stands out to me.
Yitzi: So let’s turn to the main focus: AI. To begin, AI is the biggest buzzword right now. It’s changing everything. Can you list a few things that most excite you about the current state of the artificial intelligence industry?
Yeah, I’ve been following it closely for a while. I think 2025 is going to be the year of agents. Everyone’s talking about it, but what excites me the most is what I call personal agents. Instead of agents blindly going out into the internet and clicking on things for you, imagine if they actually understood your preferences — what you want — and could act autonomously. That’s the part I’m really excited about.
Up until now, we haven’t had fully autonomous tools. For example, you still have to go into something like ChatGPT, write a prompt, get a result, and then manually copy and paste that result into something like Twitter to post a tweet. Not that I’m thrilled about the fact that so much social media is becoming AI-generated — we’ll need to figure out how to deal with that — but the shift toward full autonomy is what I think we’ll start to see this year. That’s when AI transitions from being just a tool to becoming something that can truly 10x your productivity.
Think about this: being able to have something work for you while you sleep, and then waking up to see the results, is a completely different experience from having to log in every day and type into ChatGPT yourself. That’s the kind of transformation I’m excited about.
I’m also personally very excited about extending this to people who could benefit immensely but don’t necessarily know how to use it well. Right now, you often need to know how to prompt these tools correctly to get the most out of them. If you don’t, you miss out on their full potential. I think this year, we’ll start seeing products that proactively work for you without you needing to prompt them.
Email is a great example. Instead of you having to tell it how to respond to an email, it could proactively create a draft or schedule a meeting for you. All you’d need to do is review the work that’s already done, rather than having to actively create it yourself. So, the two things I’m most excited about are autonomy — feeling almost human — and proactivity. These are the things that are currently missing from today’s tools, but I believe we’ll start to see them emerge this year.
Yitzi: That’s amazing. So, conversely, can you share a few things that concern you about the state of the industry now? And what do you think could be done to alleviate those concerns?
Yeah, that’s a great question. I think the main concern, like I touched on earlier, is that it feels like social media is becoming a dark forest. What I mean is, on platforms like Twitter, at least for my account, it seems like almost everything is sponsored AI content. Instagram Reels and TikTok are starting to feel the same — just flooded with people pushing AI tools or AI-generated user content promoting itself. That’s something we’re going to have to figure out.
I used to trust Reddit as a space for more genuine, human interactions, but now even Reddit comments often feel like they’re written by bots. It’s hard to know what’s real anymore. I thought I could trust TikTok because you can see people’s faces, but a lot of those creators are also pushing sponsored content now.
So, I think one of the biggest challenges is going to be figuring out what’s human and what’s AI. A lot of communication on the internet is becoming heavily AI-driven, and that’s going to make things tough. It creates this question of trust: How do we navigate a world where so much of the content we see might not be coming from real people? That’s something the industry and society as a whole will need to address.
Yitzi: As you know, many people are worried about the potential for AI to harm humans and the dangers of AGI. From your perspective as an insider, as a leader in the industry, what should be done to ensure that AI stays safe?
Yeah, that’s a really good question. It’s interesting because there are so many different theories on how to keep AI safe. For example, OpenAI’s approach is to reach AGI and control it, ensuring limits can be put in place. On the other hand, you have companies like Facebook and proponents of open source who are pushing toward an open-source model, where anyone can download and run the technology.
I think, whether we like it or not, the world is probably going to lean toward that open-source model, where this intelligence becomes accessible to everyone. That means we’re going to need much smarter firewalls. It’s similar to when cars started getting faster, and we had to create seat belts and other safety features. Right now, we don’t have those safety features for AI. I call it the “AI firewall.” We’ll need to develop something entirely new, like an advanced antivirus system specifically focused on AI.
One example of this is within email. If you’re someone who gets targeted by salespeople a lot, you’ve probably noticed an increase in AI-generated cold emails that feel very personalized. Most of those aren’t written by humans — they’re generated by AI that scrapes your LinkedIn profile or other online information and crafts an email designed to get you to respond. To combat this, we’ll need AI that can identify and filter out that hyper-personalized spam so you can still get to genuine, human communication.
Right now, Gmail does a decent job flagging obvious spam, but it doesn’t have a filter for these more advanced, personalized emails. We’ll need to develop entirely new systems to tackle that level of sophistication.
Yitzi: That’s great. I’ve seen a lot of discussion about the threat of model collapse because all the training data is being used up, and now they’re working on synthetic data. Can you explain to our readers what that fear is? Because surely more books are being published, more scientific research is coming out, and new ideas are being developed. So why is this an issue if human knowledge keeps growing?
Definitely. At a very high level, here’s how these large language models (LLMs) work: they’re trained on internet-scale data — essentially an enormous amount of information pulled from across the web. But the thing is, we’ve already used much of that data for training, and finding new data on that same scale is incredibly challenging. Even though new books, research papers, and other content are being created every day, it’s nowhere near the volume of data we’re talking about when we say “internet scale.” That’s why there’s been a shift toward using synthetic data to fill in the gaps.
The concern with synthetic data, however, is that it can create a feedback loop. For example, if the content being created — like SEO articles or online resources — is AI-generated, and that content gets fed back into training the AI, you’re essentially teaching AI to learn from itself. This can degrade the quality of the models over time, leading to what some people call “model collapse.” It’s like an echo chamber, where the system loses diversity and originality in its training inputs.
To address this, a lot of the focus in the field is shifting from just collecting more data to improving “runtime compute.” This is where models, like ChatGPT, are designed to think more actively during interactions rather than relying solely on pre-trained outputs. For example, instead of instantly producing a response, ChatGPT Pro might take more time to “think” and process your query in real time. This shifts the focus from needing endless new datasets to making better use of the data the models already have.
So, even if a publishing company is producing 1,000 new books a year or more research is being published, it’s still just a drop in the bucket compared to the scale of data required for these models. Human knowledge is growing, but not nearly fast enough to keep up with the demand for internet-scale datasets.
Exactly. To give you an idea of the scale we’re talking about, think about how many hours of video are uploaded to YouTube every single day — it’s mind-boggling. That’s the kind of scale these models are built on. It’s incredibly difficult to maintain that level of data for training, and that’s part of the challenge the industry is trying to solve.
Yitzi: Okay, amazing. So, please tell us about what you’re doing at Serif. Why does everyone need to get your product?
Yeah, so we’re building what I call an AI executive assistant. The idea is that everyone can save a lot of time — and not just save time, but actually achieve freedom — with an executive assistant. That’s the main vision. If you’re able to give a founder, a business owner, a manager, a writer, an author — anyone, really — an extra person who helps manage all the things that aren’t central to their day-to-day, it’s a huge help. Things like email, booking restaurants, scheduling — those are big ones. If someone could take care of those tasks, you’d feel like a burden has been lifted.
It’s not just about saving time; it’s about reducing stress. When I wake up, I don’t want to feel stressed about opening my inbox or dealing with scheduling. Our goal is to build something that feels like a human assistant — someone who can completely handle the tasks you give them. And not just handle them, but handle them the way you would. That’s why we spend so much time building the knowledge base. Anytime a new user joins, we automatically generate documents and folders to capture how they think, how they write, and who they care about. This helps us draft emails that sound like them.
The goal isn’t just to draft emails, though. We’re starting with email because it’s one of the most painful things for people. Let’s be honest — no one enjoys spending their time in email. I’ve never met anyone who says, “I love email.” But after email, we move to scheduling and then expand into all kinds of tasks.
For example, if you want to call five restaurants and book a meal today for your team of eight, we should be able to handle that automatically. We should know your preferences — like the types of restaurants you like. We should already have your schedule and your team’s schedule, so we know when to book. Because we securely store that personalized information for every user, we can actually start taking tasks off your plate entirely.
And that’s the whole goal — to completely take tasks off your plate so you can achieve more freedom.
Yitzi: That’s amazing. So you’re saying it’s not just about email, but more like an agent? That’s great. As you mentioned, Google has a small degree of similar functionality that helps you write an email. Can you explain to our readers how and why yours is different?
That’s a great question. So, Google is a great company — a huge company — but they build for millions of consumers. Their products are more like tools. For example, you can use Gemini to draft an email, but it’s never going to get as personalized. It won’t know your specific pricing information or how you handle certain support queries.
On top of that, Google is never going to fully automate tasks end to end. The way I see it, Google is a great tool, but what we’re building is more like a labor replacement. We’re trying to get as close as possible to having a human virtual assistant. That’s our main differentiation.
To achieve that, the product has to be far more personalized than Google will ever be. Google, by nature, cares about building things that work for everyone and avoiding anything risky that could impact their reputation or stock. So, they’ll never design tools that deeply cater to the specific needs of business owners, entrepreneurs, or managers. Their focus is too broad to achieve the level of UX and effectiveness we’re aiming for. That’s where we stand out.
Yitzi: It sounds like you’re describing the advantages of a lean organization — like the flexibility and focus of a startup versus a massive company like Google. Is that right?
Yes, I think it’s both agility and focus. Google is focused on serving everyone — they want their products to work for everybody. We’re focused on serving a specific group of people. I call them entrepreneurs, salespeople, managers, business owners, or creators — people who are building things in the world.
Because we’re targeting that smaller subset, we can make product decisions that are amazing for them, even if they wouldn’t work for the general population. That focus gives us a huge advantage in delivering something truly effective for the people we’re trying to help.
Yitzi: That’s amazing. So, this is a huge question that the whole world is grappling with. You’re creating a product that’s directly aiming to replace humans, right? There’s an entire field of executive assistants, and you’re building artificial executive assistants. For example, I used to hire someone for $60 an hour to transcribe interviews and edit them. But now, I use software for that, and I had to let that person go. Similarly, you’re replacing jobs, and those people will need to find other jobs. This is a big societal question — how do we balance the advantages of progress and new tools with the elimination of jobs?
Yeah, that’s a great question. The way I think about it, there are two main things I want to address. First, I want to expand the market. I want to serve people who need an executive assistant but couldn’t afford one before. There are so many people who want the freedom an EA provides, but they can’t justify spending a couple thousand dollars a month. So, our product can serve those people and help them do more.
As for the executive assistant (EA) profession itself, I think EAs will always have a place. AI will take time to reach the point where it feels truly human, and until then, EAs can still step in. For example, if you delegate a really complex task, an EA could still handle that within our platform, Serif. The simpler tasks, like drafting emails, aren’t something most people — EAs included — enjoy doing. So, it makes sense to let AI handle those things.
Longer term, when AI does reach the point where it could replace more jobs, I think people in those roles will need to evolve. They’ll need to specialize — either by niching down into very specific areas or by developing expertise in fields like social media marketing, video editing, or something else. Instead of being generalists, they could focus on areas where there’s always demand for human creativity or judgment.
I don’t have a perfect answer, though. It’s something we all have to think about. Even as an engineer, I have to ask myself how much of my work could eventually be automated and how I should adapt. It’s a big challenge, and I don’t think anyone has all the answers today. But it’s something we’ll need to keep grappling with.
Yitzi: That’s great. Is there a story behind the name Serif? People familiar with Serif and Sans Serif might recognize its association with fonts.
Yeah, so, in all honesty, I wanted to come up with a name that was two syllables — like Google, YouTube, etc. — and Serif fit that. I also wanted the name to be connected to writing because, if you actually observe a lot of people at work and watch their screens for a day or even a week, you’ll notice that much of their work involves communicating. A lot of work is about gathering information, sharing it with the right people, and then doing some tasks. Not much of it is actually clicking around or automating tasks.
So, we wanted a name that was really tied to words and writing — something that created that association in your mind. We also wanted it to be super snappy and simple, so you could easily think, “I want to delegate this to Serif. Serif is my assistant.” We landed on that name, and, luckily, the domain was available, so we just went for it.
Yitzi: So, here’s our main question: You’re a successful technology leader. For those looking to get into the industry, can you share five things you know now that you wish someone had told you when you first started leading a tech company?
- Five things. The first thing that comes to mind — especially this year, more than any other — is that you should be actively experimenting with new technology. About every eight to ten years, there’s a major leap in technology. The last big one was mobile, and around that time, we saw the rise of all the big social media and tech companies we know today. Now, it’s AI. So, you should be figuring out what’s possible today that wasn’t possible before. That’s important because if you can do something now that wasn’t feasible in the past, you can solve problems that couldn’t be solved before. And solving those problems is where new businesses come from. That’s the first thing.
- Second, you need to focus on who you want to serve first, then figure out what to build for them. A lot of us, myself included, love technology and end up creating tools — kind of like building a hammer and then looking for a nail. But when you’re starting a company, it’s crucial to be really clear about who you want to serve. For example, let’s say you want to serve business owners. You’d ask: What do business owners do every day? Maybe they spend a lot of time in their email. From there, you can map out their workflows and identify their pain points. Then, you figure out what product solves their problems. So, the process should go from understanding the person you want to serve, to identifying their problems, and then creating the right product.
- The third thing is that distribution matters a lot. You can have a good product with great distribution, or a great product with okay distribution, and in most cases, the one with better distribution will win. That was something I had to really internalize because I used to focus so much on product. But the truth is, no matter how good your product is, it doesn’t matter if no one knows about it. So, investing in distribution is critical for driving the business forward.
- Here’s a more tactical piece of advice: keep launching. Just launch as often as you can. Your first launch might not work, and that’s fine. Launch again. And again. Keep building great things and keep getting them out there. Don’t put all your hopes on one launch, and then give up or completely pivot if it doesn’t work. It’ll probably take longer than you expect, but if you keep iterating and trying, you’ll eventually find something that works.
- Another tactical piece of advice, tied to the last one, is that if you want to build a big company, you need to focus on retention. It’s great to get users, but if they don’t stick around, it doesn’t matter. So, your first priority should be nailing retention. Make sure the product is good enough to retain users. Then, figure out how to get people to hear about it. You can start working on both in parallel, but retention is the foundation.
One more tactical point: simplify the user experience as much as possible. As an engineer, it’s easy to think users won’t mind a few extra screens or questions during onboarding. But even adding one extra step can cause up to 20% of users to drop off. So, you need to focus on streamlining everything. Make it as simple as one click — something that just works. If you’re targeting consumers, this kind of simplicity is critical for achieving efficient growth.
Yitzi: That’s great advice. I once interviewed John Sculley, the CEO of Apple during the time when Steve Jobs was fired. Sculley came from Pepsi, and he said that Steve’s philosophy was always about making the user experience as simple as possible. That was a core part of his vision.
Okay, this is our aspirational question. Kevin, because of your great work and the platform you’ve built, you’re a person of enormous influence. If you could spread an idea or inspire a movement that would bring the most good to the most people, what would it be?
I really like the idea that everyone can do things. And here’s what I mean by that. I have a lot of really close friends who are incredibly smart, but for some reason, they don’t feel like they can go out into the world, build things, and make an impact. I honestly think everyone can. Deep down, I believe everyone has the potential to be an entrepreneur in some way. They can build things — not just companies, but for themselves, for others, for their communities.
With AI, for example, I feel like the barriers are coming down even more. Everyone will have tools to create and solve problems, and I want to see more people stepping into that role. I think it would be amazing to see more entrepreneurs at all levels, solving problems in innovative ways.
For example, take something like the LA fires. Imagine if those fires inspired hundreds of people to start building solutions — firefighting robots, drones, sprinklers, or some other kind of technology. That would be incredible. And the thing is, anyone can do it. It doesn’t have to be limited to just a small group of people in elite circles.
I really want to see a world where more people feel empowered to take action, to create, and to build. I think that kind of movement would bring so much good to so many people.
Our final question is what we call our Matchmaker question. We’re fortunate that prominent leaders in technology and business read this column, and this is your chance to connect with them. Is there someone in the world or in the U.S. you’d like to sit down and have coffee with, collaborate with, or share a beer with? We could tag them on social media and see if we can help make the connection.
That’s a great question.
Honestly, I’d love to meet Mark Zuckerberg. I really respect what he’s built, and it must have been incredibly tough to handle so much public backlash over the years. But I think he’s a great leader, especially for the companies under Meta and in the open-source space. So maybe Zuck — but is that shooting too high?
Let me think of someone else. I’d really like to meet Drew Houston, the CEO of Dropbox. He built such a foundational, generational company at the early stages of the internet era. Dropbox really pioneered online storage and, what I think is so cool, is how they built a company focused on serving professionals directly, not just businesses. They cared deeply about the user experience, and I think there’s a ton I could learn from him. I’d love the chance to connect with Drew.
We’ll do our best to get his attention, and who knows? Even if it doesn’t happen right away, this will be out there on the internet forever, so somehow it might reach him one day. We once interviewed someone — not a famous person, but he had made a deal with Mark Cuban. The article mentioned Mark Cuban, and Mark Cuban actually called him, saying, “Hey, your article mentioned me.” It’s hilarious how things like that can happen.
Okay, Kevin, so how can our readers find out more about Serif, try it out, and support you?
Yeah, they can head to serif.ai to start a trial. If they email me directly at [email protected] and mention that Authority Magazine sent them, I’ll give them an extended trial so they can really dig in and test out some of the newer features. I’d love to support that. I always try to respond to every email quickly. If I didn’t, that would honestly feel like an embarrassment to my own product! So yeah, they can email me at [email protected].
I can vouch for that — you do respond quickly. Well, Kevin, it’s truly been an honor to meet you. I wish you continued success and blessings, and I hope we can do this again. Keep up the amazing work — you’re really doing something great.
Thank you so much for taking the time. I really appreciate it.
My pleasure, Kevin.
Kevin Yang Of Serif.ai On Five Things You Need To Build An Amazing AI Product was originally published in Authority Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.