 Hey there my name is David and today we're going to talk about the past present and future of Notion AI Thanks so much for tuning in today Little bit about me. My name is David. I joined Notion back in 2018 we were a 10-person company operating from a little garage in San Francisco But today I work on the product marketing team at Notion focused on our AI efforts I've always said that product marketing feels like one of the most fun jobs that you could have at Notion because we do have this really passionate community of users who are always giving us suggestions about how the product could be better or What changes they would like to see in Notion and I get to be the person that shows up on launch day and says Here's this new feature that you've been waiting for or here's something that's really going to change the way that you use Notion it always felt like being kind of like the Santa Claus of Notion So what are you going to learn in this webinar today? Back in February of this year. We shipped a generally available AI product before a lot of our competitors Which was Notion AI? That meant that we skipped a lot of steps So we cut some corners and there was a few learnings along the way So if you're building your own AI product There's probably something that you can learn from this webinar about some of the things that we took away from that lunch I'm also going to show you a live demo of our newest AI feature Which is AI autofill and Notion databases and explore some potential use cases and then given sort of the Successes and learnings from our AI launches this year. I want to share some predictions about where AI is headed next So first and foremost, what is Notion? Notion is always best articulated as its core use cases which is wiki docs and projects all in one place and The initial launch of Notion AI earlier this year was very much focused focused on this Document use case plugging AI into your documents to be able to generate text on your behalf or edit text on your behalf Whether it's in your meeting notes or blog posts, etc But of course there's lots more opportunities to plug an AI into our other use cases as well Which we'll get into in a bit But what is Notion AI? I want to start by talking about October of 2022 This is about a month before chat GPT had its big launch moment And AI really started to really capture the public imagination in a huge way that we hadn't seen before in October every single Notion employee was all busy at a company off-site We were busy with company planning and our two co-founders took advantage of this opportunity when everyone was busy in Occupying themselves and they started hacking away at something new So when we all came back from this company off-site We arrived on Monday morning with a mysterious message in our Notion Slack Workspace and it was this video posted by the two co-founders That they had hacked together Plugging an AI into Notion and immediately started to think about what are all the possibilities of what this can mean? And they assembled a team that worked really really quickly and one month later by November 2022 We had built this invite-only Private alpha for Notion AI There was a wait list you could sign up to get access and we started giving access to users Slowly but steadily so that they could start testing out what it would be like to use generative AI in Notion and Then a few months later in February of 2023 is when we really opened the floodgates And we made Notion AI generally available to our tens of millions of users no wait list no limited preview Now many of our competitors at this time had made AI announcements But it was a little unusual for a company of our size to have a generally available solution that anyone could go and test out Right away now there were some trade-offs with the shipping speed Which we'll get into in a little bit Now here's what Notion AI looks like You'll see that when I start a new line in Notion We see this command press space for AI and that brings up a whole many of options here Whether I want to summarize the contents of this page if I want to improve my writing fix my spelling and grammar Or choose from a number of prompts here, you know like blog post or you know write me an essay about something So just to give you a little sneak peek of what this looks like Now AI is looking through all the contents of this page and has generated this summary really really quickly But one of the questions that we get really frequently about Notion AI is how is this different from chat GPT? and it's a valid question and I want to share a strange analogy which is to the digital camera Now I imagine that many folks who are watching this webinar at some point in their lives have owned a digital camera Now how did we quantify the value of owning a digital camera? It would be in the number of photos that you took with the camera But before we all had a smartphone in our pocket We would use this camera separate from our other devices But I'd be willing to bet that now that you have a smartphone with a built-in camera always on your person That you're taking more photos today now that it's built into the device that you're already using all the time Versus the number of photos that you used to take when it was a separate device this digital camera that was separate from your phone And this is exactly what makes Notion AI different is that it's an assistant for your work in the same place where you do the work Now a really simple Example of this would be for summarizing meeting notes now chat GPT is great at summarizing meeting notes However, in order to do that you would need to be taking your meeting notes Whether it's in Notion or Google Docs, what have you you can copy those meeting notes Go over to chat GPT in a different browser tab Paste the meeting notes and then into your prompt to ask for a summary you get the summary you copy those meeting notes You go back to your Google Doc or Notion or wherever you have your meeting notes, and then you paste your summary That's a lot of steps just to get a summary of meeting notes versus a Notion You just pull up AI right inside the meeting note right in the place where you're already working and click the summarize button So just like a digital camera versus a camera that's built into your phone You're going to use it more often because it's built into the place where you're already spending so much of your time and Therefore get more value out of having access to something as powerful as generative AI So this presentation is called the past present and future of Notion AI And I wanted to start with some launch learnings We talked about how Notion shipped this generally available generative AI solution to all of our users Before a lot of our competitors made their solutions generally available and now that shipping speed came with a number of learnings Because of how quickly we introduced the solution to market now one of the assumptions that we made in the initial launch You could see in our original marketing materials now we thought you know There's so many possibilities of what people can use generative AI for people can use it to generate a blog post generate a sales email and When you look at our original landing page You can see that we weren't very imaginative about the use cases here You can write a cute haiku about robots burritos or your quarterly okay ours and in the initial alpha Version of Notion AI every single time that you created a new page in Notion Which is a screen that our users look at really frequently We had this really noisy prompt that put all of these different ideas in front of you every single time you create a new Notion page Do you want to create a blog post with AI? Do you want to create a social media post with AI? We sort of assumed that people would use this purely generative use case similar to chat GPT what we were really missing as part of this launch is The people really want for AI to be an editor of their content not always an author of their content Now the big learning from this launch We assumed that most people would use those prompts on the new page screen that I just showed you or that they would enter their custom prompt here We were really really surprised at how frequently users Interactive with AI from this button here in the formatting toolbar now This actually has a really important learning here, which is that people prefer to write the content themselves I'm the human. I'm in the driver's seat. I have all the context about what I need to write I would prefer to write the first draft myself But then I do know the quality bar can be a little higher Maybe I wrote something that's too long and it needs to be more concise Perhaps I want to change the tone to be more confident We were really surprised at how frequently people prefer to write their own content first and then lean on AI as an editor and not an author So that's the first learning The second is that people really needed Extremely specific and actionable use cases for this brand new technology that they hadn't interacted with before We got so overwhelmed at thinking about all of the possibilities of what you could use AI for that I developed what I thought was a really clever marketing framework of work faster write better Think bigger because there's so many different things that you could do with AI I wanted to sort of categorize them into these three Sort of key value props of notion AI you can work faster by generating a summary of content you can write better by Calling up AI from your formatting toolbar and asking it to make your writing more confident We can think bigger by treating AI as a brainstorming partner. I was just thinking about all the different use cases However, what we learned is that this was not nearly specific or actionable enough for people What we really needed was really specific prompts that people could copy paste Really specific use cases and the number one feedback that we heard after this launch is Wait, what is this thing? What do I use it for? It's like customers wanted Sort of a Lego set. They wanted to see what's the end product that I'm going to achieve What are all of the specific possibilities of what I could build and instead? What we shipped was sort of a bucket of loose Lego bricks and left it up to the user to figure out how they wanted to interact with AI So if I had the ability to do this launch all over again Rather than having this really high-level marketing framework of work faster write better think bigger I would have focused on more specific and actionable use cases Even perhaps giving copy pasteable AI prompts to users in the marketing materials The learning number three from the initial launch of the notion AI in February of this year is that security and privacy is really important And there's a really important decision that we made when we are developing notion AI And it wound up getting buried in our terms of service Which is that we won't use any customer data to train any AI models It turns out this is actually really important both to consumers and businesses But unless you were reading through the fine print in our terms of service You probably wouldn't have realized that notion made this decision So in hindsight, we really should have elevated this commitment to privacy and security in our marketing materials And now we're sort of playing catch-up So that summarizes some of the learnings from our launch of notion AI back in February of this year But I do want to call out that one of the most important limitations of that initial launch is that AI only worked on One page at a time But when you talk to someone who's been using notion for a long time They will always tell you that the most special part of notion when they really had their light bulb moment about notions Customizability and notions power is in its database feature And so we thought that the next logical step for notion AI would be not only using AI on one document at a time But what if he could use AI across dozens or even hundreds of pages all the same time using notion databases? And so here we have an example of a database that a product manager might be really familiar with which is a launch roadmap Now each one of the pages in this roadmap here contains So much detail about each one of these items on the roadmap Now the product manager who wrote this content is probably pretty comfortable reading through all of this But imagine if you're a cross-functional partner and you're coming to this launch roadmap and you're really just wondering What are we shipping next? What's on the roadmap? Are we going to ship the feature that my customers asking about in the support ticket or what have you? And it's gonna take a long time to read through all of that content And so we can leverage AI to provide a summary for us of all of the content inside each one of these pages So we're going to call up this AI summary auto fill property and In just a few seconds we see how quickly Generative AI was able to search through all of the content inside each one of those pages and just give us the TLDR Now we have this really quick summary so that any cross-functional partner can come here and learn about what is each one of these features But in my opinion, I think these summaries could be a little bit more readable a little bit more scannable They're a little verbose. They're a little long And so we have the ability not only to do a really quick one-click summary or extract the key info Or do a translation we can actually input any custom prompt that we want and so what I actually want to see here is Explain this feature to me like I'm a Beginner in three bullet points no more than 15 words per bullet and Now we see that this is able to take all that content again produce the summary in this more readable format and I Know that many of our customers. I know that certainly I have colleagues across the world. There's a lot of distributed teams I have colleagues in our Japan office and our Korea office and You know, it would be great to produce a summary of this content and non-English languages And so we can even ask it in the custom prompt here For the same explain this feature to me but output in Japanese And you can imagine how useful this would be to cross functional partners and other offices across the globe Who are able to come to this database learn about what's on the product roadmap and get the really really quick summary of what's coming up next in Their own language. It's a really powerful features available in AI autofill Now giving you the ability to leverage the time-saving efficiencies of generative AI Not just on one document at a time, but across hundreds of documents at a time So that's AI autofill But where are we headed next? I wanted to share a few sort of predictions based on what we've learned and a lot of the opportunities that came out of those previous launches and Take a few guesses as to where we're headed from here. And first I really want to ask this question of who actually gets a value out of AI I think for the past year We've heard lots of lofty promises from lots of software companies about how AI is going to completely transform your day-to-day work And I think that's true of some people I think it's true certainly of AI enthusiasts and people who have really mastered the art of prompting the AI and understanding where it fits into their workflow, but I don't really believe that AI Certainly generative AI has been completely transformational to the average Joe's day-to-day work Now Today it sort of feels like a stuffy concierge desk at a hotel Now if you've ever stayed at a hotel with a concierge desk, you know that there's so much that they can do for you They can plan a birthday party for you. They can recommend local experiences. They can even set up a You know a whole local experience for you and your family. There's so many different services that they can offer restaurant reservations, etc But I don't think anybody I certainly have not approached one of these stuffy looking concierge desks Even though this value is included with the cost of your hotel stay Even though there's so much that they can do for you at the end of the day You have to walk up to the desk and you have to express intent. You have to ask for help And I don't think this describes the average person's experience with AI today I don't think the average person is going about their day-to-day work and constantly pulling up chat GPT or notion AI Or any other generative AI tool and feeding it specific prompts about how it can speed up their workflow And I think the thing that will make the biggest difference here is again this idea of intent in the future I believe that AI will be more helpful when it becomes passive Ambient and anticipatory as long as we're keeping the human in the loop not making any decisions on behalf of the human Always asking the human to approve or reject any changes that may be made more like a concierge who actually leaves their desk Goes to any hotel guest and anticipates their needs before they have them I think there's a huge opportunity here for AI to become more passive more ambience and start to anticipate Users needs before they have them Next I really want to see over the next year sort of a transition from AI feeling just like another software tool To more of a teammate that actually understands how the business operates a lot of software tools today feel like sort of a generic Out-of-the-box large language model But and this is even true of a notion AI today if I ask notion AI to Recommend Q4 okay ours for my team based on Q3 performance It's not going to give me a very good output because it doesn't have all of the context of how the business operates This is also true of any other generative AI tool like Jack GPT or a standard Google search Google has been using AI for years, but the next huge opportunity here is for an AI That does understand the context of your business how a business operates all the previous decisions that have been made every task That's ever been assigned to an engineer so that it can actually become more of a helpful thought partner And you can lean on it for advice throughout your workday rather than just as a brainstorming partner or a tool that Improves your your documents and how they're written and I want to end today with a few thoughts on what is the real Job to be done for AI now at notion we've always been really interested in the origins and the history of computing and I think we can all imagine how scary it must have been back in the 60s back in the 70s when computers were this brand new thing and I imagine there must have been a lot of people thinking Back in the 60s and in the 70s our computers going to replace us now one of the computing pioneers we talk about the most frequently at notion is Douglas Engelbart and He his most important paper was not about replacing human intellect rather His paper his most important paper was titled augmenting human intellect And I think this will be true for AI as well We certainly certainly believe this at notion that the role of AI is not to replace but to augment human capabilities What if AI was able to take over all of the you know really menial Tasks the work about work that prevents you from being able to really Focus on what your best at the thing that you were hired for So if we get to a point where we can truly delegate all of our tedium To an AI that's able to handle those things on our behalf It really allows us to rise to our full potential Augment our own intellect be more creative and do the best work that we were hired for in the first place Thanks so much for watching. Thanks for tuning in today