No point in thinking about three-year plans in the AI era. Building an adaptable, resilient team is key: Notion’s Akshay Kothari

Kothari, who was initially an investor in Notion, joined the company in 2018 as chief operating officer when the company had just eight employees. He later earned the title of co-founder.

The $10 billion startup, which counts Sequoia, Index Ventures, and Coatue Management among its investors, aims to challenge Microsoft and Google's dominance in the workspace with its AI-powered tools.

During the interview, Kothari also spoke about how Notion is rewiring its products with AI, its ambition to build a productivity suite that better connects users’ work lives, and how India is shaping up as a key market for the company, both in terms of customers and its employee base.

Edited excerpts

Can you give me a sense of how things are unfolding in AI right now, especially as someone sitting in the heart of Silicon Valley?

It is pretty dizzying to be honest, it's moving really fast. A day feels like a week, and a week feels like a month. The last two years have been probably the most exciting two years of my career.

People are saying that around 90 percent of software could be written by AI next year, and I think that might actually turn out to be a true prediction. It's going to be a pretty different world.

Notion focuses on knowledge work, which is everything from people writing docs to managing their projects and taking meeting notes. I have a feeling that even knowledge work itself will basically become more or less programmable. If I can give all the context I have to Notion AI today, it probably makes better decisions than I can already.

One of the things I’ve been thinking about is the concept of knowledge workers, people who essentially think for a living. There are about a billion knowledge workers in the world today.

In the future, it's possible that there will be no concept of knowledge workers. It becomes knowledge managers. The idea is that everyone, like you and me, all of us will have agents that just do things. The knowledge manager’s work will be to coordinate across them.

It applies to every function, whether you are in sales or marketing. If the person has agency - the will to do things - they will be able to do a lot.

How do you plan for the long term in an environment like this then? Does it even make sense?

It is a dynamic world. There’s no point in thinking about three-year plans anymore. I think you just need a sense of where you want to be in a year. Even that might change a lot.

There are things we're using today; that whole category didn’t exist three months ago. Even trillion dollar companies like Google and Microsoft are moving so fast.

In some ways, you have to build a team that is quite adaptable, resilient to change and is willing to lean into it. For Notion itself, AI started as a small side project with five people and now everything we do — whether you're writing a doc, designing a database, or searching Q&A — touches AI.

I'm not sure if I keep up fully with it too but we've hired a lot of people who are quite good at it. So I think of my job a little bit like an editor.

I’m not sure I come up with the best ideas, but I look at people’s plans and edit things towards where I think the world is going.

If you look at the startup world, there are so many new products being launched. But the interesting thing is that the products still feel prototypy - only a few actually work reliably and consistently well.

Has the pace of shipping increased? Notion has had a really busy year — probably 10 or 15 launches, right?

It has gone up. The most crude way to look at it is by code commits per person. We grew the engineering team by 30 percent last year, but our code commits doubled.

This year, I think the code commits have already probably gone up by 30-40 percent. So in general, productivity per software engineer has gone up a lot.

How do you think AI coding agents are changing the way we write software?

It's changing quite drastically. I think it’s basically gone from writing every piece of code yourself, or getting help from Stack Overflow, to coding agents autocompleting for you as you write.

Where we're headed is that you don't even have to do that. You can literally come in and say, “Hey, I’m trying to build this thing,” and let the agent run. “I’m trying to do this other thing,” and let that one run too parallely.

What about the entry-level people coming in now? What should they know, or what should they be learning in this environment?

I actually think people who are graduating now might actually be well positioned, because they don't have any baggage or have a way of doing things. So if you graduate from a college today and take a software developer job, you will probably just start with agents because that's how the world has been operating over the last year.

The challenge is more in the mid-tier where you are not an experienced architect and you're not an entry level native AI programmer. You've been coding for the last 10 years in a specific way, but so much has changed. The question is will you be able to unlearn and learn this new thing?

We've also changed our recruiting to focus on those two edges — we want experienced architects because we want Notion to scale to millions of users.

And then we are also hiring a lot of people fresh out of college, because those people don't know any other way, they only know the AI way of coding.

I'm generally bullish on people coming out of college now because they are also a little bit more open to adopting the newest tools.

I have two young kids and the thing I want to teach my kids is agency to do things and resilience to change. That’s pretty much the two things.

As long as you have the agency to want to build things, there’s going to be incredible possibilities.

People talk about this idea, which initially felt stupid, but now it’s possible. I think someone is going to say, “You know what? I don’t want to hire any employees. I’m going to have agents and build a hundred-million-dollar revenue business.”

Because now you can. You can have a designer agent coordinate with a software agent, which coordinates with a sales agent. I think that’s going to happen before the end of the decade.

It’s like intelligence as a service. For India, for example, imagine the barrier to entry is significantly reduced. It will challenge people who resist change, and also those who are kind of just waiting for orders.

How are you rewiring Notion's products with AI? 

For the last 10 years, we've built the core building blocks. We have an editor that is amongst the world's best editors. Then we have databases that people can use for everything from personal use cases to work ones.

Now we have an AI layer or a chat agent layer that you can ask to do things. What we are doing now is basically looking at each persona, breaking down their work into specific tasks. What do they actually do? how do I make it easier to assist them in all those use cases? How do I augment their work?

One of the things we did recently that was actually counterintuitive was that we removed the AI add-on as a plan because AI is everywhere now.

Basically any plan you get, you're going to get AI. The higher tiers just give you more premium features.

You’ve launched Calendar, now Mail — is the idea to keep going in that productivity direction? Like building a whole suite?

That's the goal. At the end of the day, it’s about connecting people's work. Initially we had the view of all the docs that you've read or written. Then with Calendar, I got a sense for all the meetings you have.

And now with Mail, I understand all the communication you're doing. Interestingly, none of the existing large players are connecting these experiences together.

There’s no player today where you can say, “Hey, look at all my emails, tell me all the trips I have, put them on my calendar, and remind me to prepare for them.”

I went to a conference recently. I uploaded the PDF with the attendee list and ran a research mode on it. And I said, “You know about Notion and what we’re trying to solve. You know these companies. Pick the 20 I should go talk to and tell me how I should start the conversation.” It actually did an amazing job.

Eventually, I hope people will think of it as a suite. But in the interim, it doesn't need to be the suite. I'm just trying to connect your work life better.

Notion has always called itself an anti-SaaS company. What does that mean in practice?

SaaS in the last 20 years has just exploded into very specialized tools, thanks to the internet. The number of tools people use at work every year is over 100 now, as per the new Okta report.

We think it happened because it's become easy to build software and make money from specialised products. I feel the pendulum has to swing the other way, where people are just sick of so many different tools.

If you’re a startup today, you need a search tool, a chatbot, meeting notes, email, research, a team wiki, and a CRM — that’s like eight different tools, each costing $20–$50. We give you all of that in one product, for $20 a month.

Anti-SaaS is a bit of a tongue-in-cheek. What we're saying is anti-specialized SaaS. you should have a few tools that just do a lot of the work.

What are your thoughts on distribution, especially for a company like Notion? Does AI have a role here as well?

I feel like a lot of product founders like myself, we spend too much time on the product. At the end of the day, the only way I can beat Microsoft and Google at their game is through distribution.

If their products are the default on people’s devices, I’m playing a pretty tough game. So I have to figure out: how do I take on the power of the defaults?

There's a bunch of different ways to do it. I think we got to over 100 million users partly because students in colleges prefer Notion over Google Docs today. If they use Notion as default, then when they go to work, they're going to demand Notion at work.

I also use ChatGPT a lot more than I use Google now. So, if I'm going to go to ChatGPT and be like, “Hey, what tool should I use to take notes?” we need to make sure Notion shows up there.

So, the distribution is shifting, and each company needs to figure out what they need to do to drive it.

How is India shaping up as a market right now, in terms of customers as well as an employee base?  

India is our only R&D center outside the U.S. We have New York and San Francisco and Hyderabad is the only other one. We got there through an acquisition a few years ago, but now we've rebuilt it. So I think we have around 60, 70 people. Just under 10 percent of our global employees are there.

India is a very fast-growing market in terms of usage. Younger people use it a lot, but we’re also now being adopted inside businesses. A lot of unicorns use us, so the business side is starting to take shape as well.

We have to do more. I think there are two key things we need to crack for India. First, the product needs to be faster, so that it works well even on 3G and 4G connections. The other thing is support for local payment options like UPI.

We're very bullish on the market. APAC in general is a big focus for us. India and Australia are probably our next growth markets.

Do you see a difference in customers from these markets as compared to mature markets like the US? 

For India, the usage is comparable. The monetisation will take time. Part of it is probably like we grew up never paying for software. But it’s changing and people are starting to pay.

I think the price itself may need to evolve a little bit so that it feels right for the market. But before that, I would want to support all the local payment methods.

What are the biggest challenges that you face right now? 

I think the talent market is very hot right now. So if you're trying to hire like the best of the best AI people, the going rate is extremely high.

There are very few people who are extremely good and understand this world of generative AI in a deep way. Some of these people are truly 10x people or 100x people and you can see why.

If you can get into the mode of having agents do your work, you can parallelise a bunch of different things. There's going to be a person who can literally do the job of 100 others because they know exactly the ways to parallelize it, which was not possible before.