"We're doubling every couple of weeks now." While most startups struggle to find their first hundred users, Keith Peiris and his co-founder Henry Lee at Tome achieved something extraordinary: they built the fastest productivity tool to reach 1 million users.
Tome isn't just another presentation platform—it's a complete reimagining of storytelling for the AI age. By leveraging large language models to help users craft compelling narratives, they've tapped into something deeper than just removing the pain points of PowerPoint. They've created something people actually love using.
In this candid interview, Keith reveals why most productivity tools fail to create genuine excitement, how they convinced an OpenAI founder to join their board before they had a product, and why he believes the future of knowledge work will turn us all into directors rather than creators.
Watch the full interview now on EO's YouTube channel! Below is the complete transcription of the interview. Minor edits have been made for clarity and readability.
Key Highlights:
"Building a compelling PowerPoint presentation just requires a lot of time and effort to design the slides. The biggest problem is that it's very hard to start a story, but we found that using large language models, we can actually help people with that, because it turns out that most stories sort of follow a similar arc."
"We started to grow exponentially in December. We're doubling every couple of weeks now. We'll be getting to millions over the next couple of months."
"I think we're actually moving to a world where people aren't going to create work themselves anymore, they're going to be directors. They're going to ask algorithms to create different types of work, and you just evaluate it and shape it."
"We started to grow exponentially in December. We're doubling every couple of weeks now. We'll be getting to millions over the next couple of months."
"I think we're actually moving to a world where people aren't going to create work themselves anymore, they're going to be directors. They're going to ask algorithms to create different types of work, and you just evaluate it and shape it."
Breaking the PowerPoint Prison
What's fundamentally broken about how we create presentations today?
Keith: There's a whole lot of problems with PowerPoint. You have to go back to the origin story. Original PowerPoint that was made in the 80s was built for physical slides. You would send physical disks to a print company, they would print your slides, and then you would take your slides, feed them into a projector, you would give your presentation.
When you think about it, that origin, PowerPoint isn't interactive, and it doesn't take advantage that we're looking at presentations all the time on digital screens and on our phones. The other thing is that building a compelling PowerPoint presentation just requires a lot of time and effort to design the slides. If you work at a company, anytime you're designing the slides so that your idea will be considered and understood is just kind of a waste.
But we found that using large language models we can actually help people with that because it turns out that most stories sort of follow a similar arc and we found that this is a really interesting opportunity if you get a large language model, you should be able to generate the starting point to build your story on top of. Great storytelling connects facts and concepts together into a narrative that we can understand.
When you think about it, that origin, PowerPoint isn't interactive, and it doesn't take advantage that we're looking at presentations all the time on digital screens and on our phones. The other thing is that building a compelling PowerPoint presentation just requires a lot of time and effort to design the slides. If you work at a company, anytime you're designing the slides so that your idea will be considered and understood is just kind of a waste.
But we found that using large language models we can actually help people with that because it turns out that most stories sort of follow a similar arc and we found that this is a really interesting opportunity if you get a large language model, you should be able to generate the starting point to build your story on top of. Great storytelling connects facts and concepts together into a narrative that we can understand.
Finding the Perfect Co-founder
How did you and Henry decide to start Tome together?
Keith: I think we started to believe that this idea was valuable, but I knew that I needed sort of an extraordinary co-founder, which is to say someone to help imagine multi-modal storytelling tool from the future. I think with Henry it felt very good, very fast. He had just redesigned all of Facebook Messenger. He was actually about to reject some big other Silicon Valley job.
Henry: It was really easy to brainstorm and generate with Keith. That was probably the most standout quality to our interactions. When I was still working at Facebook, I was talking to my friend Seth and Seth mentioned that Keith was thinking about starting a company. We started to work together and have these sort of jam sessions on exploring this problem space of presentations and why they're not great anymore and like why people don't really like using them.
Those were enjoyable for the reasons I mentioned, it felt really generative and we decided to start this company together from the perspective of both having an intense focus on the kind of business it would become and really being able to create a company with unbounded potential as well as a product that wouldn't compromise on the details, its execution along the way.
Henry: It was really easy to brainstorm and generate with Keith. That was probably the most standout quality to our interactions. When I was still working at Facebook, I was talking to my friend Seth and Seth mentioned that Keith was thinking about starting a company. We started to work together and have these sort of jam sessions on exploring this problem space of presentations and why they're not great anymore and like why people don't really like using them.
Those were enjoyable for the reasons I mentioned, it felt really generative and we decided to start this company together from the perspective of both having an intense focus on the kind of business it would become and really being able to create a company with unbounded potential as well as a product that wouldn't compromise on the details, its execution along the way.
The OpenAI Connection
How did you convince Reed Hoffman, one of OpenAI's founders, to join your board before you even had a product?
Keith: Early on it was just us working in the Greylock office. At some point we realized, all right, this is enough analysis paralysis. We just have to make this thing and then iterate on it instead of sitting in this room thinking about can we get the perfect product for the perfect customer. Some businesses are built that way, but certainly not productivity tools.
The way we attracted Reed to Tome, I'm very grateful that he's on our board and he gives us advice. He's one of the founders of OpenAI. Because usually you can't get someone like Reed to work on your seed company idea with no product. We just kept emailing him, we kept having Zoom chats with him and showing him progress. And at one point he was like, OK, you've made enough progress, you should go start the company. And that's when we put out the ask. We're like, well, if you think we've made enough progress to go and start the company, would you jump on the board with us?
The most memorable thing I remember Reed saying was making sure we were building a networked tool, meaning not just something that was really useful to one person, but that could result in them sharing it with other people and then adding them and bringing them into the creation process and being collaborative and just having several touch points for network effects to occur.
The way we attracted Reed to Tome, I'm very grateful that he's on our board and he gives us advice. He's one of the founders of OpenAI. Because usually you can't get someone like Reed to work on your seed company idea with no product. We just kept emailing him, we kept having Zoom chats with him and showing him progress. And at one point he was like, OK, you've made enough progress, you should go start the company. And that's when we put out the ask. We're like, well, if you think we've made enough progress to go and start the company, would you jump on the board with us?
The most memorable thing I remember Reed saying was making sure we were building a networked tool, meaning not just something that was really useful to one person, but that could result in them sharing it with other people and then adding them and bringing them into the creation process and being collaborative and just having several touch points for network effects to occur.
Beyond Removing Pain
What's the difference between building something people need versus something they love?
Keith: A startup is dead by default. The sort of default path is that you run out of money and you sort of have to pack up your bags. Doing everything well doesn't necessarily mean you will sort of escape that inevitability.
I think early on we had this view of, we should write perfect product specs, we should build everything as well as we can. We should force people to show up really early and work really late, and I think those things are necessary but insufficient. You actually need to tap into a deeper human need and you need to build something that's like awkwardly shaped that people could love, and I think that was sort of a realization with Tome of the absence of pain isn't joy, and if all you do is remove pain, you aren't going to necessarily create something compelling that's gonna bring people in.
Really early on, it requires a lot of imagination for someone to understand where it's going and what it could be. First year at Tome, I think the feedback we got was that this is cool, but it's not ready for me yet. That started to change last year. I would say around July when we realized people are actually using this for something.
I think early on we had this view of, we should write perfect product specs, we should build everything as well as we can. We should force people to show up really early and work really late, and I think those things are necessary but insufficient. You actually need to tap into a deeper human need and you need to build something that's like awkwardly shaped that people could love, and I think that was sort of a realization with Tome of the absence of pain isn't joy, and if all you do is remove pain, you aren't going to necessarily create something compelling that's gonna bring people in.
Really early on, it requires a lot of imagination for someone to understand where it's going and what it could be. First year at Tome, I think the feedback we got was that this is cool, but it's not ready for me yet. That started to change last year. I would say around July when we realized people are actually using this for something.
Explosive Growth
When did you realize Tome had reached product-market fit?
Keith: We're starting to get feedback like, hey, I'm trying to build a pitch deck with this, and we just realized that the quality of feedback was getting really high because people were trying to use Tome for real things.
I think we started to grow exponentially in December. We're doubling every couple of weeks now. We have hundreds of thousands of people creating in Tome, and I imagine that we'll be getting to millions over the next couple of months. And then a lot of people use Tome as a presentation platform. So that I think that's one of the reasons it's growing. It's just so useful across a wide variety of use cases, which is a very meaningful moment for us.
We just went down yesterday morning because there were a few too many people trying to sign up and use it and we had to scramble to get it back up and sort of create more scalability.
I think we started to grow exponentially in December. We're doubling every couple of weeks now. We have hundreds of thousands of people creating in Tome, and I imagine that we'll be getting to millions over the next couple of months. And then a lot of people use Tome as a presentation platform. So that I think that's one of the reasons it's growing. It's just so useful across a wide variety of use cases, which is a very meaningful moment for us.
We just went down yesterday morning because there were a few too many people trying to sign up and use it and we had to scramble to get it back up and sort of create more scalability.
The Future of Knowledge Work
How do you see AI changing the way we work?
Keith: When I think about the future of AI, AI being this collaborative partner with you, and that's a really profound thing. If you work in knowledge work, if you work in an office and you use a computer, it used to be that you need to create work yourself. I think we're actually moving to a world where people aren't going to do that anymore, they're going to be directors. They're going to ask algorithms to create different types of work, and you just evaluate it and shape it.
I think in some ways it's going to turn knowledge work into sort of being a good manager, except instead of managing lots of people, you'll be asking these systems to be generating work and you just evaluate it and shape it. For example, research could be really quickly automated or just retrieval of the right information to put something together. Iterations, let's say you had a mock-up and you wanted to make 5 versions of it, and AI could do that instantly and you might take 15 minutes.
The more data oriented and facthood oriented tasks are going to probably take us a little bit longer to get right with these models because the models aren't able to really understand the difference between opinion and true fact and also the topic of truth itself is complicated like who's the president of the United States could be talking about today or it could be talking about a year from now or a year in the past.
I think in some ways it's going to turn knowledge work into sort of being a good manager, except instead of managing lots of people, you'll be asking these systems to be generating work and you just evaluate it and shape it. For example, research could be really quickly automated or just retrieval of the right information to put something together. Iterations, let's say you had a mock-up and you wanted to make 5 versions of it, and AI could do that instantly and you might take 15 minutes.
The more data oriented and facthood oriented tasks are going to probably take us a little bit longer to get right with these models because the models aren't able to really understand the difference between opinion and true fact and also the topic of truth itself is complicated like who's the president of the United States could be talking about today or it could be talking about a year from now or a year in the past.
David vs. Goliath
How do you compete against Microsoft when they could just integrate OpenAI into PowerPoint?
Keith: If you sit around and ponder what makes humans special, I'd argue that the biggest thing that differentiates humans from other species is our ability to work together. In history, this is why we're able to build spaceships. This is why we're able to build supercomputers, because we can get hundreds or thousands of people all agree on something and work together.
If we were just building AI PowerPoint, we should go home. Sooner or later, Microsoft will connect PowerPoint to the OpenAI's models and I'm sure it'll work great. I think what we're building is completely different. We're reimagining how you tell stories in the age of the Internet. We're building our human computer interaction paradigms right now thinking through AI. We're thinking through how you need to have a conversation with large language models rather than press a button and it generates everything you want.
We know we are David in David versus Goliath, so it's a race. I think this age-old need for humans to be able to communicate to one another is going to persist. It doesn't matter how good technology is. My dream is that we can keep evolving and keep building core technology to help people understand each other even 100 years from now.
If we were just building AI PowerPoint, we should go home. Sooner or later, Microsoft will connect PowerPoint to the OpenAI's models and I'm sure it'll work great. I think what we're building is completely different. We're reimagining how you tell stories in the age of the Internet. We're building our human computer interaction paradigms right now thinking through AI. We're thinking through how you need to have a conversation with large language models rather than press a button and it generates everything you want.
We know we are David in David versus Goliath, so it's a race. I think this age-old need for humans to be able to communicate to one another is going to persist. It doesn't matter how good technology is. My dream is that we can keep evolving and keep building core technology to help people understand each other even 100 years from now.
Explore more