#581 Akshay Kothari | COO Notion

00:00:00: Our guest today grew up in India, the middle of three brothers who all came to US to study.

00:00:05: He describes his life as a series of plans that did not work out and of turns ended better than any plan precisely because of this.

00:00:14: Instead one of his dream universities he went to Purdue.

00:00:17: instead of investment banking joined family office in Silicon Valley.

00:00:21: when his work visa didn't come through In two thousand eight started at Stanford Lehman Brothers collapsed.

00:00:31: A class that required students to ship a product within eleven weeks gave birth to the news app Pulse, which was later acquired by LinkedIn.

00:00:40: There he led international business SVP products until a chance breakfast with Ivan Zaho changed everything.

00:00:47: He had backed Ivans startup Notion as an angel investor in his first investment after selling Pulse.

00:00:54: For four years he watched the team search for its path and at times thought, his money was gone.

00:01:01: When Notion finally took off in two thousand eighteen he joined as employee number eight with a mandate to do everything except product.

00:01:09: today is chief operating officer has built support sales marketing customer success people legal and finance each over six or nine months before finding an experienced leader to carry the function forward.

00:01:22: He sees that people he hired as this real legacy notion.

00:01:27: We were especially touched by a writer.

00:01:29: he had shape.

00:01:31: Every Friday, someone at the company spends forty-five minutes telling their life story and it was even himself who went first!

00:01:40: For more than nine and half years this podcast has been exploring how work can strengthen people rather then wear them down.

00:01:48: in More Than five hundred eighty episodes we have talked with more than seven hundred personalities about what his change for them And watch still needs to change.

00:01:57: What can we learn from the life in which most important currents were never planned?

00:02:02: How does a company transform itself with AI from inside out and what that means for people who work there.

00:02:10: And why does CC, The biggest opportunity of our time In small and medium sized businesses?

00:02:16: One thing is clear To solve challenges today We need fresh impulses.

00:02:22: So keep searching for methods Role models experiences tools & ideas.

00:02:27: that brings us closer to the core of new work.

00:02:31: And from the very beginning, we have been asking whether everyone can truly find and live what they really want deep inside?

00:02:38: You're listening on The Way To New Work today

00:02:43: with Patshaya

00:02:43: Guthari!

00:03:01: Actually welcome to the podcast On The Way Too New Work.

00:03:04: I'm very excited about this and you know a little bit.

00:03:10: There is one question that we like ask everyone into this podcast, and it's a very easy one.

00:03:14: And you can answer in thirty seconds or three hours really depends on how did he become the person who are today?

00:03:23: It's great question.

00:03:25: I'll try to stay on the thirty second side rather than the three hour site.

00:03:31: but yeah i grew up in India.

00:03:33: an am the middle brother of three brothers came out here to the US, this study.

00:03:44: And yeah and I think in some ways my life has been a series of plans that did not work out but then the thing that worked because of that was ended up being better than the plans.

00:04:02: so i find myself much more open-minded about or opportunities that may just show up rather than wanting to be very precise about how life is going to play out.

00:04:22: If that makes sense, there's like several examples where I feel like i made all the plans did all of work and then... That didn't work out but then thing worked because it ended taking me a better path anyways.

00:04:37: So I don't know, maybe it's made me more stoic as someone who just turned forty and looking back at how life has worked out so far.

00:04:59: Yeah, so I

00:05:08: mean part of it is...I think if i actually rewinded even further back.

00:05:17: I remember when I was in high school I took my ascites and did well on math but didn't do as well on verbal.

00:05:25: And So I had this dream of maybe you know going to Stanford or Berkeley or MIT.

00:05:35: And and that didn't play out, but I ended up going to like a top-ten at engineering college in Purdue.

00:05:41: It just happened to be the right setup for me Like.

00:05:44: i think it really gained a lot of confidence at Purdue In those years.

00:05:50: uh...I think At the end of prudu..i was so obsessed with wanting To get a job in investment banking and consulting because this Was two thousand six two thousand seven?

00:06:02: The peak.

00:06:03: So every smart person I knew was getting into investment banking.

00:06:08: And then that didn't work out, and so that ended up bringing me to the Bay Area.

00:06:12: Uh...and i-I jumped straight to working at a family office investing in startups, and I loved their job!

00:06:19: ...And I wish my job could continue but life had other plans.. ..And so I-I didn't get like a work visa in two thousand eight.

00:06:29: uh....and as backup I had applied to Stanford and I got into Stanford.

00:06:34: And so, I started at Stanford September of two thousand eight.

00:06:40: the week Lehman Brothers went bankrupt.

00:06:42: So what an amazing time do join college financial crisis was happening?

00:06:50: Then at Stanford tried a bunch different things much startups that didn't play out.

00:06:55: but literally in my last quarter I started Pulse, which was a news app that LinkedIn acquired.

00:07:04: And then...I was happy at LinkedIn as like a VP product running international for LinkedIn and it chanced breakfast with Ivan where he said why don't you come help me scale this company?

00:07:21: This business became the thing that I'd sort of took on, right?

00:07:26: And so very unplanned a lot of these things but it sorta just shaped out that way.

00:07:31: How was like when you...I mean, I remember Pulse and I know their story because I'm slightly similar startup generation as you are.

00:07:43: So I experienced Lehman in my first start-up parallel happening.

00:07:54: your backup was Stanford, and then you started Pulse and sold it.

00:07:59: So like there is magic happening in between?

00:08:01: Like what's this magic thing happening?

00:08:05: they're going through the steps?

00:08:08: Is that by accident life happening or are also some pushing on pursuing in-between ?

00:08:19: To be honest ,it does feel happy accidents And I'm not trying to be like overly humble here, it is genuinely.

00:08:30: a lot of these things have been just very serendipitous.

00:08:34: Just to give you an example right?

00:08:39: Like the only reason we picked to work on news and media in two thousand ten was because we were in a Stanford class called Launchpad.

00:08:49: One of the requirements of the class within the eleven weeks of class.

00:08:58: And my co-founder and I had a bunch different other ideas like travel, data visualization... The only reason we picked News was because it felt as if we could build something in the quarter to launch it!

00:09:16: That just took us down an interesting path even with Pulse.

00:09:22: Right before LinkedIn acquired us, six months before that we were at a fork in the road where we had to raise our Series B and actually have tough time raising Series B back then.

00:09:36: We weren't able to... And it turns out that the LinkedIn acquisition was great for them AND for US like I think would've been hard business too scale as an aggregator like I think very few aggregation media businesses have done, you know.

00:09:54: Done really well right?

00:09:56: And so...I think that combination was also quite serendipitous and yeah in some ways feel blessed.

00:10:08: these things happened!

00:10:10: The funny thing is when those things i had plans for doesn't happen quite disappointed.

00:10:22: And as I've lived all these years, now when things don't work out... As I have planned, I'm always watching the corners and like something's gonna show up!

00:10:34: Now i better be watching-I'd better be open to

00:10:36: it.".

00:10:38: I liked that because.. When you then started Notion maybe some people also know a different time of startups than is right now.

00:10:50: I don't want to take away too many of the stories that are heard from outside, but just know how agile you change and steer and lead your company.

00:11:01: You're the COO and co-founder as a founder team!

00:11:09: You had legendary people on your team... You worked legendary in design and usability ...and became quite early tool for people started to adapt or play with.

00:11:20: And as I heard from the stories, you also stumbled into that.

00:11:24: You planted a little bit different in the beginning and it turned out to be something

00:11:29: else.".

00:11:29: I would be very curious to hear that because it sounds like again one of these stories didn't plan?

00:11:39: Yeah!

00:11:39: The story-of-notion is quite interesting too.

00:11:42: It was started by Ivan back in twenty thirteen... ...I wasn't an early investor yet but actually my first company I invested after Pulse was acquired by LinkedIn.

00:11:53: And the original idea why I was so excited about it, was essentially creating a platform where other people can build software and this is of course like you know pre AI...and uh..And The Idea Was Like If You Know Kind Of Like Can You Decompose All Enterprise Software Into It's Building Blocks?

00:12:14: Then You Just Provide That As A Platform So People Can Create Their Own Custom Software Like, they don't have to go a coding boot camp or hire programmers.

00:12:25: They will be able create this piece of software themselves.

00:12:29: and it felt like you know... Felt very compelling idea which is why I invested in that.

00:12:35: but actually we sort-of meandered for four years after that.

00:12:39: because turns out people don't wake up wanting the build software.

00:12:44: uh..they wakeup do their job.

00:12:48: One of the main things we've learned in our journey so far is that you really have to think about your company as like, I don't know it's a surfboard and you have to catch the right waves.

00:12:59: That come your way.

00:13:00: And then some ways.

00:13:02: for us The starting point For what really put notion on the map was that It Was great editor and A great knowledge base for Your Company.

00:13:15: After a year or so, we added Notion two point O which was the concept of these general purpose databases and I think for lot people.

00:13:25: People think of notion as like oh yeah it's Like A nice note taking tool Or notion is a great project management Tool.

00:13:33: And all These use cases are really what put us on The map but Underneath these Use Cases We always had the platform that we were trying to build right i mean in today People build all sorts of really compelling software with notion.

00:13:50: You know if you look at the marketplace, I think people built also lots of things with it and so It's taken us a while to eventually get to the core mission we had but it required us to Really surf a bunch of waves to get here.

00:14:07: I mean The last.

00:14:08: even in the last three or so years, you know surfing the AI wave has been So much fun, but also like so much change.

00:14:18: Right?

00:14:18: It's how do you evolve the product?

00:14:21: How to evolve that technology stack?

00:14:23: How do we involve people and culture?

00:14:27: it being quite hard To make evolution especially when your customers are sort of annoyed with all changes You put into a product But it felt necessary.

00:14:41: We were We needed to make all these changes, too.

00:14:47: Be the right product for this generation as a part of it came from defense.

00:14:52: It's like well somebody else will create this.

00:14:54: before that happens let's do it ourselves.

00:14:57: And then the defense became offense Right?

00:15:01: As we started to see more and more people leveraging The power of AI we realized This is actually the thing That will grow the business and grow the product and Make the product even More accessible To People who don't want to arrange the building blocks together, right?

00:15:16: And so it's actually opened up the product to a much larger audience.

00:15:20: Um and so anyways I think its been quite an evolution.

00:15:25: i think one of things we've tried do is keep team small because not like some profitability metric were trying hit but mostly becuase you know its easier move a small boat rather than large ship.

00:15:42: And I think one of the things we've tried to be is just stay nimble and sort of adapt.

00:15:47: To the ocean off, you know all these different Things that are happening in the world and make sure we can continue surfing well

00:15:55: as The COO.

00:15:56: oh i'm pretty sure You're the one.

00:15:59: try it?

00:15:59: You have a challenge to organize like decide Like where to put the effort into the team especially if your change and I mean From the early days I kind of knew how Notion would adapt as a tool.

00:16:13: Now you adapted AIY, so you adopt the teammate that you built with but... The people who need to build these tools are much different and still building great product is something totally different than just running in the background giving little bit glimpse into steering team in a phase from we built web product to the background runs AI.

00:16:41: And now, We need rethink UX for AI and at that same time but thats different question.

00:16:48: I heard you also have your own model research team because this is new story coming up But some people it's like totally new field.

00:16:58: i'm pretty sure guys see differently otherwise they wouldn't do one thing.

00:17:02: give me shine little bit of light into.

00:17:07: Yeah, I think it's a really great question because the lot of companies are trying to figure out like well Okay.

00:17:11: Well how do we evolve in this new era?

00:17:13: Like How Do We Drive It?

00:17:16: and maybe like what The way we did it in notion could be helpful for A Lot Of People.

00:17:21: And so i Think The Best Way To Describe It Or Think About This For Us Has Been This Evolution Inside Out Right.

00:17:32: And So What Do I Mean By That?

00:17:33: Like Instead of us saying like, oh top down here's the charter or here is a different way we're operating.

00:17:40: The way it has played out actually started four years ago when we put our team of five people to work on AI features and they shipped their first product called the AI writer And in those five people essentially were able take whatever GPT-IV and essentially integrate into existing products.

00:18:01: I think that worked really well.

00:18:03: And so then that team of five became twenty.

00:18:07: The next product we worked on was enterprise search where you can essentially instead of asking searching for pages, You Can Ask Questions and So We went from one feature to maybe like three different features or Three Different Products?

00:18:22: um...and That Team Became Forty People.

00:18:28: Forty became a little bit larger, and at some point we become like all four hundred people in the engineering org is essentially leveraging AI or building on AI products.

00:18:40: At which point you don't even need to word AI because this just new way of operating right?

00:18:45: And so I think aggressively growing like small nucleus So that it sort almost absorbs rest.

00:18:54: organization work really well for us.

00:18:58: And in that process, I think several people had a hard learning curve and they understood what we were trying to do.

00:19:09: They sort of went on the learning curve became part of team today.

00:19:14: There are also some who weren't believers or felt like doing wrong thing eventually ended up leaving company.

00:19:28: You know, that's just sort of what you have to do in terms of sort of really investing into the future.

00:19:34: I think if we had hesitated... ...I think we would've ended up with two different companies.

00:19:40: It was like half the company is working on AI and other half are working something else.

00:19:46: So anyways for us being evolution has basically gone from four-to-forty to four.

00:19:51: hundred engineers all leaned onto same thing on the course of roughly two years or so.

00:20:00: Like it took two years for that full evolution to play out?

00:20:03: I mean, at this point you don't have to convince someone to work on AI Or You don't like having them dream about what's possible.

00:20:10: people are already seeing big gains happening But if your a company with ten thousand people trying to figure how am i gonna transform my talent To this new world?

00:20:25: Hopefully this is like an example for folks to think about.

00:20:31: I have one million questions related to that, but the ones that really struck my mind... For some people it's very hard to foresee how.

00:20:44: for them AI has more text completion and a little bit of chatbot doing something in the background.

00:20:49: On the one hand you've customers that are used to notion rebuilding their own stuff, so you have the entry point of people taking them along for a new kind of experience.

00:21:00: To work with it tool one hand at the same time use scale and leverage the company an I mean US provider at the sametime need to figure out how much tokens How much compute our own models do we buy these models?

00:21:17: I don't know how we bet on the compute side.

00:21:19: So you're, your own both of these sites which must be quite exciting.

00:21:24: let's start with a UX site and How end users interact With AI tools because most likely i guess You guys have much more evidence On that since people use These tools with you Compared to other companies And you can build on That.

00:21:40: so what are you seeing in terms Of future user interfaces, future user experiences?

00:21:49: that goes beyond?

00:21:50: oh make me an email or write me a text.

00:21:54: Yeah I mean it's great question.

00:21:57: i would say like maybe if i take a step back its probably the evolution has been from AI feature to AI product.

00:22:15: The feature was probably like an AI writer.

00:22:17: Essentially, if you're writing a doc, AI can... You could call that a GPT-IV wrapper of sorts.

00:22:24: AI product were something we released a year ago where we had meeting notes enterprise search and research mode available.

00:22:32: I mean those who are full end to end products That you use notion as your enterprise search engine connect all the tools And get one place

00:22:45: Which is not an easy product to build.

00:22:48: I mean, this something a lot of companies try to solve?

00:22:51: They're huge company.

00:22:52: just focus on that.

00:22:55: Yeah

00:22:58: exactly!

00:22:59: I would be curious hear what you can share about

00:23:03: it.

00:23:03: And then i think what do consider notion today?

00:23:07: very much like the best platform for agents and humans collaborate right.

00:23:12: so if what's going to make humans and agents more productive.

00:23:17: I think it's gonna be like us capturing the right context for the right work, And Notion is essentially a platform that delivers on there Like It could be things you've put in Notion.

00:23:27: but we can also use workers To create A Context Graph That You Can Give to Agents Effortlessly Right?

00:23:34: Like thats sort of The Platform We Provide Now.

00:23:37: Okay so So that evolution?

00:23:39: i Think that Evolution Only Happened Because I think there was a lot of founder conviction in doing that evolution.

00:23:49: Because if you read the Reddit, like Notion's Reddit which is pretty active every time we would introduce anything with AI people were like here we go again some AI feature.

00:24:02: what are you doing?

00:24:03: Why are you messing up the product?

00:24:05: just oh Here We Go another AI release.

00:24:08: why will they not solve the underlying thing?

00:24:10: and then Fast forward?

00:24:13: two years later, I think there's now posts where people are like.

00:24:18: well.

00:24:19: I'm not sure about the old notion Like.

00:24:22: i only use the agent to run The work and I am using Claude And Notion as my infrastructure.

00:24:29: It is completely changed right?

00:24:31: People find so much value with it.

00:24:36: So its been quite a journey.

00:24:39: I think Ivan, my co-founder CEO like he has a good way of explaining how to build with AI products.

00:24:45: He sort of calls it that's similar to brewing beer and we're in the seventh season of Brewing Beer... As

00:24:55: German i can relate to that.

00:24:57: Yeah!

00:24:58: And I think its great analogy because you are working on this non deterministic technology right?

00:25:08: You're trying to get a feel for like, well what is the right experience?

00:25:12: For the customer.

00:25:14: And I don't think we got it right in the first couple of seasons.

00:25:19: In fact seven Is The Number Of Times We Have Had To Rebuild The Harness To have the agent do What We Want It To Do Right and each time.

00:25:29: So you

00:25:30: mean from scratch up not just Adjusting system prompts, but like really going to which model to use how it calls the tools and so on.

00:25:41: Yeah rebuild the harness from scratch seven times now.

00:25:44: And I think you have to do that because The new model capabilities that come out every six months.

00:25:49: if You'd really want a leverage all of it here?

00:25:52: You can't just like tweak the previous harness right uh-and for us Like we support all the different models Right and so so we've had to do So anyways, I guess.

00:26:02: long story short like i think the beer was not so good in The first couple seasons but it's gotten quite Good now and i think we know We're continuing to invest In it.

00:26:11: And the whole company is very much focused on It?

00:26:13: uh...and so um..so i Think that's That's a hard part about building products today.

00:26:20: Its Like A little bit like you're playing with a Little Bit of the unknown like Nobody can explain Why LLMs do what they Do.

00:26:29: rewire your product to what the users expect, right?

00:26:36: Like I think just to give you a simple example like... I think SaaS used to be this piece of software where we click bunch of buttons for things happen.

00:26:47: Uh..I think human behavior now has no tolerance for that.

00:26:52: Right!

00:26:53: I want to explain what i wanna do and the software should able it evolve your user experience to what the human sort of expectations are, right?

00:27:08: And that's a hard transformation for existing companies.

00:27:12: We had an interesting discussion today and here is a thought on UX over the future.

00:27:19: I mean i'm running along agent tasks today looking at my code burn.

00:27:24: let me check.

00:27:25: it's three hundred sixty nine forty five.

00:27:29: So that's quite cheap for the day still, but it's still running.

00:27:33: And we had a discussion of like what on the to-do list?

00:27:37: For their task?

00:27:38: and I said...I don't know because through everything away from the original plan months ago i used the plan mode a lot in cursor and planned me this and then so on..and I resetted It just gave us some of the subagents tools to use relentlessly towards the goal.

00:28:01: And then we had this discussion here because my chief of staff said, yeah but like I want to keep that two lists even for The Agent.

00:28:08: go back to the list and check?

00:28:09: And i said ,I don't care where The Agent writes it.

00:28:12: how about a Goal List?

00:28:13: But How Would You Write A Goal Lists?

00:28:14: What's The UX?

00:28:15: Because a lot Of Companies At The Moment Try To Have This Kanban Board Experience.

00:28:19: The Agent Is Working .But I Don'T Care Too Much If It'S Pursuing My Goal.

00:28:24: The Only Thing I Need To Do Is Sometimes Adjust give the frame and so on, but that's a totally new user interaction.

00:28:32: Totally new!

00:28:34: What is your thought of it?

00:28:37: Well I think what makes this work today like... So fun and dynamic And also frustrating at the same time.

00:28:46: because you build these things We got the current expectations or behavior dialed in.

00:28:55: Then just another thing comes up and people are sort of moved on to this new behavior.

00:29:02: And then, you know the other thing is like... The top-of-the class is so far ahead than your median.

00:29:10: or yeah it's just a difference between how you may be using it versus your median customer in the world?

00:29:17: Like the Median Customer In The World things have AI as better search engine.

00:29:22: essentially they're not even aware of agents plans or goals, or loops.

00:29:29: Or graphs and all these things are just like foreign concepts for them right?

00:29:33: And so I think when you're building a product that serves tens of millions of people i think you sort have to figure out okay well how much do you lean into this leading edge?

00:29:44: and then How Do You Bring The Median Closer To The Leading Edge?

00:29:49: It's A Tough Balance But I think what you're saying is essentially sort of has been my experience.

00:29:56: And part of the reason we have to redo the harness, it's because you just cannot leverage the latest model capabilities a different way right?

00:30:06: Uh...I remember at that recent release from Claude like everybody was saying delete all your previous instructions, delete all the previous stuff cause'like the models.

00:30:16: just do it now and if make the most of Opus five, you have to delete the previous instructions so that you can just have them model go at it.

00:30:26: Which is fascinating.

00:30:29: and then there's a thing like changes happening every three months now right?

00:30:32: And say okay we'll do that maybe they will be another one or other set-up things you need.

00:30:40: So anyways its very dynamic not only product but also pricing in packaging.

00:30:50: Like if I tell you like, hey Kristoff.

00:30:52: This is how our pricing packaging works.

00:30:55: It'll work for a quarter and then like depending on your usage like it could bankrupt my company and have to keep evolving.

00:31:03: okay well what?

00:31:05: What are the usage patterns and what does The cost of these things...it's not cheap!

00:31:12: We built software where we're running at ninety percent margins man Running software businesses.

00:31:18: now there There's a real raw material cost now.

00:31:22: Like if you get thirty, forty then your on the good side already.

00:31:27: but I mean exactly in that topic The difficult thing i see user wise and you said at the gap between the capabilities of the users huge And At the end the user says give me button I don't freaking care.

00:31:40: I want to button Give it To Me.

00:31:43: But what I experienced is there's a difference in terms of, hey AI makes me more efficient or faster.

00:31:48: Anything forget that.

00:31:49: and What i see?

00:31:50: Is i do stuff That i never done before because it was on the list And There Was no time not The capability whatsoever and Me as user i need to be capable To price It.

00:32:02: i Need to Be Capable to know what its worth and What I saw always with notion As you guys were very Very good In.

00:32:11: I remember, i saw a meeting.

00:32:13: I don't know what it was YouTube or something where you designed the button and made sure how much you invest into that button when people click so on?

00:32:21: So your very in to details of that.

00:32:24: do u see this pricing topic becoming something as part user experience with good feeling What model to choose and so on?

00:32:35: because I mean even open AI and entropic have the model picker.

00:32:40: Uh, i Have this tool here To choose the intelligence.

00:32:45: people listening?

00:32:45: I have this freaky uh Quotex thingy you to change the intelligence of the model.

00:32:51: but How do People get that?

00:32:54: in order to have a gut feeling for oh You know what?

00:33:04: user experience now or is there still pricing and packaging?

00:33:07: I don't know.

00:33:10: Yeah, we are currently... A friend of mine said that in the MS-DOS era it's like the experiences were pretty elementary and slightly complicated.

00:33:25: so i imagine a world where you do not have to pick models.

00:33:35: You know companies can optimize the best intelligence they can get for every dollar.

00:33:39: They spend.

00:33:42: I also imagine a world where you Know in a company of a hundred people there might be two people who knew how to use that really, really well and i would love For The Two People To Craft The Right Setup Then Click A Button And then Have The Other Ninety Eight People Just Get On That Train right?

00:34:06: we are very focused on is, we're extremely focused on this like multiplayer collaborative secure AI platform.

00:34:17: And we think that it's an extremely important piece of infrastructure to build because I think its probably the fastest way where you can diffuse power inside organizations right?

00:34:28: One person builds a factory and other people in their company just get onto.

00:34:35: We see this over and over again at every company.

00:34:39: Um, we also see This is a big opportunity for I think you all have this concept which i love like middle stand.

00:34:48: Yeah

00:34:48: It's right it's

00:34:50: small businesses

00:34:52: notion as A Small Businesses Company.

00:34:54: Like You know vast majority of our revenues come from small business And We See Is An Important Responsibility For Us To Bring Small Business into this AI era, right?

00:35:07: We want to make sure small businesses can leverage the power of AI just as much as like a Fortune five hundred company.

00:35:14: And when different people can create the right setups and put that setup in a marketplace or get that from a consultant more we can get the Small Businesses onto the right platform.

00:35:26: Right then.

00:35:27: so I think our responsibility is a platform Is to continue be on the bleeding edge So at top-of-the class Is able to build the right set of things and then provide a platform so that the rest of the people can actually like get on it Versus having to build at themselves, right?

00:35:45: Like this whole idea that I have to remember which model to use Which prompt with structure all that stuff is too complicated.

00:35:52: It's not gonna help other People like leverage the power of it.

00:35:55: Instead we want more people To be able to click button Get the right setup.

00:36:00: by the way That basically what companies in our customers are saying Every time I go there and explain the platform in their capabilities, they're like, actually...I get it.

00:36:10: Your platform has all these capabilities but what i need now from you is to tell me how should I work here?

00:36:18: How do I transform my company?

00:36:20: Just give it to me!

00:36:22: Tell me about its needs to be done.

00:36:25: It's an interesting responsibility for us because Historically, the platform and the malleability is what people appreciated about Notion.

00:36:34: But now they're really increasingly looking at Notion to tell them like the blueprint of how they can transform their companies?

00:36:42: And we are taking that responsibility quite seriously!

00:36:45: Which was an interesting position for us because I had a conversation years ago with Stuart Butterfield, founder of Slack And I mean, he's a philosopher by education.

00:37:00: We really deep dived into this concept of what they were building.

00:37:04: and yet the famous article in two thousand thirteen when they launched in The Beta close one it was called we don't sell saddles here.

00:37:15: love that because you're really described how workplace would change with them on end.

00:37:21: What I see with you, do know these Moleskine books?

00:37:25: The black notebooks.

00:37:27: And there was this phase when people took pictures on the internet of their Moleski and how they were getting things done in those little stickers and all that stuff.

00:37:38: And Notion has this ability yourself... ...and then you share or someone else does it!

00:37:44: And interesting question is – Is the tool needed to help true transform the work?

00:37:52: because what I always try to do is teach people new ways of working in a way, but.

00:37:58: A tool itself never changes by.

00:38:00: the tool can be the accelerator can really leverage on that?

00:38:07: Is that what you currently in and What glimpses Do You see of changes at work?

00:38:13: there are much different like four years ago for example.

00:38:19: Well, it's changed so much.

00:38:20: I think if you just sort of think about how people work today... Maybe like five years ago the simplest unit was taking notes right?

00:38:34: Like everybody does that and open a piece of paper or write... The simplest unit has probably shifted from taking notes to chatting with your agent.

00:38:48: I can quickly tell the agent like, oh here's an idea have you know.

00:38:53: Here is how it could work and i Can point that to a few different things And actually be like okay?

00:38:57: You just write a draft Like That other thing at The time you did right.

00:39:01: It's easier for me To say that rather than like Write it out.

00:39:05: um and so I think if you Think about like Okay Well what position What?

00:39:12: what's the value or what's The position of notion in this broader Thing It's a little bit like, okay if you imagine each professional worker.

00:39:22: Each company will have to accommodate people wanting to chat with different agents right?

00:39:29: I think some may have a preference to use Claude and other people may have the preference to used ChatGPD.

00:39:36: Other People May Want To Use Open Weight Models Or Whatever.

00:39:39: Right And There Is Going To Be Agents And Notion Has Its Own Agent.

00:39:45: But The primary value that we provide to companies is, you can choose to use NotionAgent but also continue with your current agents.

00:39:56: What we bring in the context?

00:39:58: infrastructure?

00:40:01: How will the agents get better?

00:40:04: if they have a right context it'll be sharper.

00:40:07: That's our first thing We provide.

00:40:10: Second thing Is You Can Go From one person building it and having to share those prompts with another person, you can essentially build the right setup by one person.

00:40:24: And then other company basically just gets access to it.

00:40:28: A great example of this is like in a one data scientist at Notion created an agent called DataScout which essentially as good as senior data scientists that notion He just shared with the whole company and now everybody has their own personal data scientist.

00:40:47: And his job, he's basically transformed from him having to run the queries himself To him running this system.

00:40:55: Basically It is like extremely high leverage Where all that he was doing Is making sure the DataScout System continues to deliver To everyone inside of the company.

00:41:06: So thats the other thing.

00:41:08: Finally I think especially for companies.

00:41:11: I think companies have to be careful not to depend too much on a singular lab.

00:41:20: To give away their IP, right?

00:41:23: And so we've built essentially a harness where you can choose whichever models or agents and tokens that want to use because these businesses should have full control over it as opposed.

00:41:43: And so to your point, it's like I don't know.

00:41:47: Some people may want start with a doc.

00:41:49: some people might wanna start the workflow.

00:41:51: A lot of we're gonna just wanna started agents.

00:41:55: and in that setup i think company has make sure infrastructure is ready support those natural human expectations.

00:42:06: where notion comes in is that we provide that infrastructure.

00:42:09: We think it will be critical infrastructure to provide both humans and agents the right context, the right model support and the right multiplayer collaboration support.

00:42:24: That's our current sort of thinking-and-platform building

00:42:27: From from Building internally my own being or one Our own engineers here.

00:42:33: I know how hard it is, especially with the memory in the contacts and

00:42:36: databases.".

00:42:38: And then a stumble upon an article of you guys from The Information on the Weekend that even not like... Like as many other companies saying, okay we don't need that.

00:42:48: Many people you're actually hiring.

00:42:50: so your planning to hire much more people and maybe you shine a little bit light of that into companies who still think they'd still thing.

00:42:58: hey this is for being more efficient than I need less people but you are actually growing with more people.

00:43:05: why?

00:43:05: Is that?

00:43:07: And how was that bet on AI from me?

00:43:09: it's more like the strategy under not about.

00:43:14: The biggest mistake people are making with AI is I think they're starting to think of AI as like an efficiency play.

00:43:19: It's like oh, I can do what i'm doing right now But I can't.

00:43:21: it would have the people and I think?

00:43:25: They write play with a eyes too.

00:43:27: Just ten exude ambition.

00:43:28: Right And so Why Are we growing the company?

00:43:35: We were going in the company because we are not trying To Do the same goal that we had two years ago With half the People.

00:43:43: We see a path for us to have like the ten X, The Ambition from two years ago.

00:43:49: Two years ago I was selling you software.

00:43:51: today i can sell your work.

00:43:53: that is ten times the bigger opportunity For me To go after right and when You start to raise Your ambition?

00:44:03: If you look at a lot of companies that are leaning very heavily on AI, they're continuing to grow.

00:44:12: I think Ramp put out a study on this and it's happening because people are realizing like wow!

00:44:37: you have to operate knowing that anything's possible.

00:44:42: I heard a crazy story recently, I was catching up with designer who used to work at Cursor and he basically said like Grockbach was built by ten-fifteen people in two or three months right?

00:45:01: That is remarkable!

00:45:04: The fact that you could build a new product like that, that is so compelling.

00:45:10: Leveraging the infrastructure they had built over the years for cursor Is just outstanding right?

00:45:17: And So You have to take a pause and be okay.

00:45:20: if your are thousand person company it might Be easier to spin out new products.

00:45:26: That gets on these exponentials.

00:45:30: If you're not doing Then you have to ask whether your like playing the efficiency game or raising the ambition.

00:45:39: If you have someone from the middle stand listening which is very likely today

00:45:45: and

00:45:45: they're like okay but we are not a tech company, and where?

00:45:50: in different kind of market how would want them to see the world in different way except for off course that this should use notion with most likelihood already.

00:46:00: but Beyond that, let's say founder CEO someone being responsible and.

00:46:06: Be tough in clear on them also when it comes to understanding technology or anything else that's needed.

00:46:11: what would be your current recommendation for them?

00:46:16: Great question.

00:46:18: so lets see like you're maybe a small business with I don't know.

00:46:24: Let's say you are operating four restaurants.

00:46:27: right You have an opportunity build the fifth restaurant in a completely different way, right?

00:46:37: Instead of you doing it the same where you've done for previous four restaurants maybe on this one.

00:46:43: You set up the right digital infrastructure.

00:46:49: It's like okay.

00:46:49: well how is customer service happening?

00:46:52: How are we getting new customers?

00:47:01: how are my different tools connected, right?

00:47:05: And if you just set it up in a way where your fifth restaurant is forget the number of people.

00:47:12: but I think like you have a dashboard where can run that restaurant and see the business or grow the business.

00:47:19: You may realize wow actually get much faster payback maybe nine months restaurant here, which means that I can actually like instead of me opening two restaurants a year.

00:47:35: I could be opening like ten because of this new infrastructure i've built right?

00:47:41: One of the one of the craziest stories I heard was somebody who's using Notion recently with Chick-fil-A franchise owner.

00:47:53: Right classic middle stand.

00:47:55: it is small business running few franchises.

00:47:59: He used Notion workers to connect all the different data sources where his data is.

00:48:05: On top of that, he built a bunch of custom agents and His setup was so good That he was invited by The company To give a presentation about just like okay well how How?

00:48:16: Is it running this franchise Like So Well?

00:48:20: And again It's not Just About Efficiency.

00:48:21: Its Like The Top Line Is Growing As Well.

00:48:25: And now I'm convinced that every small business can do that, right?

00:48:29: It's like if you could speak your local language whether it is English or German.

00:48:34: You can tell the AI hey here what am trying to do!

00:48:39: Like make sure i have this infrastructure which gives me access all my data and on top of my data ,i want an agent who does these things everyday morning & evening.

00:48:49: there are a good chance for them.

00:48:54: So I think even if you keep your existing business status quo, You can take the new business and see if you can find something.

00:49:01: And If you've seen that is working then you will back solve it for your existing businesses too.

00:49:07: Hopefully That means not just that you are more efficient than more profitable It means that its given you time to actually like do a lot of stuff right?

00:49:19: Infrastructure wise You guys have to scale up the infrastructure in a much different way.

00:49:24: And I could make that decision as if, hey you know what?

00:49:41: I throw all my data as middle-stand into Notion.

00:49:53: And, i can have it behave like a database scale in different sources connectable to the outside compared other ones.

00:50:02: because In future would say The limit of speed is not the model Because models will become much faster.

00:50:09: but how tools connect?

00:50:12: and How do you guys make sure your infrastructure As fast or even faster in order of evolving, because I guess you made figure it out before I do or anyone else building.

00:50:24: But that's not an easy one?

00:50:26: You know what i mean.

00:50:27: like the model is fast but

00:50:30: yeah... Yeah.. I think to your point ,the models already has so much capability today even if there was no more progress .I think would probably take us a decade to fully leverage.

00:50:46: So I think on the infrastructure, you know this is an area we're investing a lot.

00:50:49: We find ourselves slightly in a fortunate position that The Infrastructure we built for human collaboration turns out to be pretty good infrastructure For agents as well right.

00:51:01: so like i have A piece of document or database where u and i can collaborate today?

00:51:07: is just another person that shows up in terms of the work it's doing and you have full attribution, so forth.

00:51:16: The surface we've built—the canvas we've build —is actually extremely good for humans and agents to collaborate with each other.

00:51:25: I think where this is getting tricky for us Is that human scale... ...is different from agent-scale.

00:51:31: So they give a simple example.

00:51:33: Like the Human Scale databases was like A hundred thousand rows is good enough.

00:51:41: Like, a hundred-thousand rows means like I think have a hundred thousand tasks or customer list.

00:51:46: that's good enough... Agent scale as wild right?

00:51:49: So if you're trying to connect the notion workers where an ocean database gets synced with your Salesforce records and some snowflake data or something You need maybe three orders of magnitude more scale!

00:52:02: You need like one hundred million records supported in a database And this is exactly what we're working on right now.

00:52:08: It's like, We want to support a hundred million records in our database and for you to be able to sync worker between the two and be able connect that data on Notionsend Right?

00:52:20: So so That's actually Like A big investment.

00:52:23: were making it just Making sure Our knowledge base are databases or workflows.

00:52:29: Are You know all This stuff Is ready For agent scale.

00:52:34: And then there's a question of like, okay well does this infrastructure work for small business?

00:52:42: Can a Chick-fil-A franchise owner find the way to think their legacy system data living there versus logistic systems or payment system and can all that be synced on the notion side.

00:53:04: And it's actually like super fun and deep work we have to do to support those use cases, which gives us the confidence that we are going deep on that side.

00:53:14: Right?

00:53:15: I think there is companies Like The Labs That Are Going Deep On Agents.

00:53:19: We Are Going Deep On The Infraside So That We Could Be Completely Complimentary And Completely Useful Infrastructure For Agents To Go Do Their

00:53:28: Work.

00:53:33: This is something I stumbled upon because you need to work with so much context, especially when it comes to a team.

00:53:38: If your working on your own its quite easy.

00:53:41: one tool that's fine.

00:53:42: but if want leverage into the whole team You really need make sure like what kind of data going where who using and can become very chaotic.

00:53:54: That's good cause in some ways we talked about.

00:53:59: every company has couple people you know, who knows how they can set this part up.

00:54:06: And once you create that right setup that's secure has the right permissions as a write data access.

00:54:13: it is such a huge unlock for the rest of the organization Right?

00:54:17: Like if he just said that up.

00:54:18: well now You've unlocked the power of agents or like one hundred people rather than Just For Yourself

00:54:26: People wise in terms off since I mean This also podcast about How Work Is Changing.

00:54:33: What Are The Profiles you're hiring at the moment, how hard are on?

00:54:37: How capable people when they join notion in terms of using eye or does it really depend on The job.

00:54:45: Or do say You know what you really need to be capable and a certain point.

00:54:49: that's What I expect.

00:54:53: so we've made a couple different.

00:54:57: I guess We've made A couple Different Bets Here But it is a constantly evolving space, right?

00:55:04: It's like okay well what's the right bet.

00:55:05: Uh uh...it does seem that you're looking for slightly different.

00:55:11: maybe needs Right um.

00:55:13: and so I guess The big hiring bet we've made Is That We'Ve Gone A Little Bit More.

00:55:17: Barbell Shaped Hiring.

00:55:19: So were hiding alot of early career starters People coming out Of College.

00:55:25: Now These people don't know any other way Like they only Know The AI Way because they only learned it that specific way.

00:55:35: And I think this, they bring a lot of energy and drive sort of intensity to go build things.

00:55:44: so that's one bet.

00:55:45: on the other side we are hiding extremely senior architects.

00:55:49: cause when you think about like okay well how do we gonna support database have hundred million records?

00:55:54: How're going?

00:55:54: make sure is available all platforms supporting stuff experienced architects trying to do that, right?

00:56:03: And so essentially our hiding has sort of gone to the edges and it also helps.

00:56:12: Our company today has been more middle heavy.

00:56:16: we've hired a lot of mid-level experienced people.

00:56:20: in some ways I think It helps for notion kind where were going but it also help us see like new talent coming out as well as continue to invest in the core infrastructure side.

00:56:35: I do think, some of it probably is just basic expectation now right?

00:56:45: If you're interviewing at Notion we are not expecting you train on how leverage AI like expected for all two have early understanding.

00:57:01: This is not the case today, but a year or two ago there were also like a lot of people just skeptical with what AI can do.

00:57:07: I don't see that anymore.

00:57:10: now it's a question of how deep you go?

00:57:14: How much leverage can you get from AI right?

00:57:16: It's like have you gone from one agent to five agents too having a long agent working while sleeping all these different things and still requires human judgment in terms of like what the right design is and with the right ideas are, so forth.

00:57:36: But a good motivated person can do a lot interesting building.

00:57:42: now that's possible.

00:57:45: You have more agents working at Notion or people who work at Notions?

00:57:51: We have way more agents here than Notion.

00:57:52: I think even our custom agent platform probably has thousands of agents because these agents to do very specific things, right?

00:58:02: So as an example like I have an agent between me and my EA.

00:58:08: And every time we're just chatting on Slack it just picks up the right thing then puts that into our to-do list.

00:58:14: It doesn't require any change in our behavior because i'm still talking through her but it just kicks out what some of the to-dos is.

00:58:20: when you say its done it's just marks it down.

00:58:22: so we removed this extra work.

00:58:28: I think it's a good model for thinking how work should evolve.

00:58:32: Like, its not saying that you replace people.

00:58:34: It just says like hey all this busywork You had to do.

00:58:38: get rid of that and make time For more off your life's work right?

00:58:44: And i think That is the good way To get leverage Of Your Time.

00:58:49: If we talk in three years from now Does Notion have More Agent Customers Or Still More Human Customers?

00:58:58: This is one of the most interesting things.

00:59:06: I don't know what shape it takes, so this kind of like giving you a peek into our brainstorms right now... Well first off all i sort feel that death-of-seat based pricing has greatly exaggerated.

00:59:26: I sort of feel like there's something nice about seat pricing.

00:59:30: You know, it is predictable and easy for finance people to think about.

00:59:35: And so the question is does everything we do in software all become usage-based pricing or are they also going to be seats?

00:59:48: The question would that be human seats?

00:59:50: then will their also be agent seats?

00:59:52: So i can give you a case.

00:59:55: Notion is like critical infrastructure to give you context.

00:59:59: For every agent that you use, it doesn't matter which service you're using You should give a notion agent seat That gives the right context and tools And so forth.

01:00:11: In that world there's reasonable chance.

01:00:14: in three years There are more agents than human seats.

01:00:20: Even if we take usage path Is there going be content produced in notion from agents versus humans.

01:00:30: I actually think that agents will trump that pretty quickly, we don't have to wait three years either and so it's kind of like an interesting time in the world because i feel our jobs is going evolve form doing-the thing to evaluating and approving the things...to dreaming next thing right?

01:00:50: And this is not just an engineer thing, right?

01:00:52: This says everyone has to do this.

01:00:54: Even if you're a marketer your value alpha on dreaming the things and evaluating them.

01:00:59: with doing that thing as gone it's like easier to leverage agents for.

01:01:07: so I think its such incredibly dynamic exciting time.

01:01:12: thinking about future.

01:01:16: what gives me calm energy on notion side we're building the infrastructure, right?

01:01:22: And it doesn't matter if it's like humans using or agents.

01:01:26: We are here to provide you context and flexibility of different models in Agents so that you can evaluate what your agents did whether they want to approve them.

01:01:42: I've been telling people today is GitHub for knowledge work Like, is this beautiful infrastructure that programmers got for managing the work and working together?

01:01:58: We are basically providing that infrastructure for everybody else.

01:02:03: That's something I can relate to very much which you gave me into And to round it up.

01:02:11: give you the last question... ...I would be curious since you talked about how you stumbled upon these opportunities.

01:02:18: Where do want go from here?

01:02:21: As a company and maybe even for myself, I imagine the future of just exponentially more creations.

01:02:35: Exponentially more creativity being unleashed in the world And hopefully also exponentially small businesses.

01:02:49: I mean, small businesses is really the heart of notion.

01:02:52: Really what drives our business and everything is much more doable now like starting a company building a product selling a product scaling in the business.

01:03:04: all this as much more do-able.

01:03:07: maybe to end on fun note here roughly two weekends ago you know i was having hanging out with my daughter, who's nine years old.

01:03:20: And for the last year or so we've been reading a lot of print newspaper and magazines over breakfast.

01:03:28: So we subscribe to The Financial Times.

01:03:29: We read that every morning.

01:03:33: Then she also has a few like you know explorers and make magazine.

01:03:36: I read The Economist...so we have been reading it for a year now..and i thought wouldn't be fun if you & me did a podcast.

01:03:45: maybe once every week We just discussed the stories we've read, right?

01:03:51: So you went from this idea to Saturday night basically like roughly two weekends ago.

01:03:58: I just got off my phone and started a voice memo on my phone.

01:04:03: one take twenty minutes...we discussed five stories..We had all our newspapers line out on table And said okay You pick three stories i'll pick three story which is sort of when back-and-forth And the thing I love about this is like, we're not looking at a computer.

01:04:18: We just literally did it over our print newspaper and voice memo.

01:04:21: Uh...I use Riverside to clean up The Voice.

01:04:30: I used Spotify for creators To launch the podcast.

01:04:34: I use ChatGBD For coming with cover art.

01:04:37: And like, ninety minutes later..we have a podcast!

01:04:41: We did out second episode couple of days ago.

01:04:43: Nice.

01:04:43: Lovely idea.

01:04:45: I love this activity.

01:04:46: The fact that her and i could create a podcast, you know leveraging all these different tools we have is just so gives me so much optimism for what people can do around the

01:05:01: world.

01:05:03: That's lovely idea!

01:05:04: And very fun note to end on.

01:05:06: actually pleasure meeting you in getting to know Also about your story, but very well connected to what you guys are building.

01:05:16: Thank you so much!

01:05:17: This is a company that

01:05:25: really changes the perception of scientific work and can be followed by the good man, four hundred thousand people on LinkedIn.

01:05:45: I was actually excited about five things.

01:05:48: at first The best careers are rarely planned.

01:05:52: Actually you wanted to go into this investment bank.

01:05:53: then you didn't get a job visa, went to Stanford, founded the news from Pulse, sold them to LinkedIn, worked there And then came Notion.

01:06:02: His great learning, if he has a plan, is worth it especially to be aware of.

01:06:06: Because often the next big chance arises there.

01:06:11: Second, KI is for him an growth strategy and that's really exciting because as he says, the biggest mistake is to look at KI exclusively as an efficiency program.

01:06:20: so do the same work with half-as many people.

01:06:23: Your opponent?

01:06:23: Just tell your ambition tenfold.

01:06:26: Ask yourself what you can achieve with KI.

01:06:28: until now was completely unthinkable and even further in, because the Emit-KI has made a lot of bigger goals.

01:06:35: The third transformation begins at the small one.

01:06:38: Notch started with five people who worked for MKI from twenty to forty, then finally about four hundred engineers integrated the KI into their daily work.

01:06:49: I find that pretty fascinating!

01:06:51: Transformation comes together like this rather through practical experience here by learning and by teaming up others.

01:06:58: Fourths we are becoming less self-employed.

01:07:00: more often Actualy describes an development that, as a special occupation of our work, shifts itself from doing it on its own to evaluating decisions and then developing new ideas.

01:07:13: And at iNotion we already have more AI agents today than people.

01:07:16: A data scientist is for example an agent built by everyone working as a personal data expert.

01:07:22: What does this actually mean?

01:07:24: For collaboration and skills in the future?

01:07:27: Fifth technology can improve our creativity.

01:07:30: My favorite story comes at the end, when he told us how his nine-year daughter started a podcast that discussed exciting stories and then opened up her conversations with some digital tools on the first podcast after only ninety minutes.

01:07:46: This is really nice picture for the future of work.

01:07:49: people have ideas.

01:07:51: Technology helps them to let these ideas become reality.

01:07:54: One question I would have liked to ask him very much.

01:07:58: If a AI agent takes over more tasks and maybe some people only make decisions or evaluate the results, how do people learn then actually make good decisions?

01:08:06: And it also adds that experience is missing which we collected through our own work in the past .

01:08:11: I think one of the biggest challenges for leading learning and future work lies there.

01:08:18: My conclusion follows a lot of desire towards the future.

01:08:20: she shows her radical work has just changed and how much we can win if our imagination develops as fast as the technology.

01:08:27: We also have to talk about the short-term side, but there is probably another conversation partner or a different conversation partner for sure!

01:08:37: If you liked this episode like me then please give us five stars, subscribe to us, recommend us more, feel free to leave a comment on one of your platforms.

01:08:49: And when you want to work with us at Blackboard, especially when you really want to deepen yourself with KI or are interested in Cloud.

01:09:02: Then you're right there.

01:09:03: if your leadership training, education self-reflection and self-management etc.

01:09:08: is then interested in our NWMS GmbH at newworkmasterskills.com.

01:09:14: We look forward to seeing each other again next week!