 Hello, my name is David Kosnick and I'm really excited to chat with you today. I'm product lead at Coda and for the last year I've been obsessed with AI. How is AI going to change productivity? How is it going to change my role as a product manager? And what does that mean in terms of what skills are valuable? And so I wanted to share with you today some of my own tips and tricks that I've learned over the past year building Coda AI and using AI all the time as well to make my job better. And so let's jump in. Thanks for joining. I've been a product manager for about 13 years now. I started my PM career at Google. I worked initially on search and then on YouTube. I did very data oriented types of product work on search lots of AB testing and very design oriented product work at YouTube. Actually in both places I worked on AI. AI to power search and results and ranking. And AI to power early creativity tools for uploaders on YouTube. And so I've been long, fascinated and obsessed with AI. Where is it going? What's it going to enable? What's the important thing for people? And so I left YouTube after a while and started my own company in the video game space trying to build a next generation Minecraft. Initially in virtual reality and then became really excited about early generative technology that could basically make deepfake. So tried to make a deepfake sandbox video game, which was a ton of fun. And then joined Coda about three years ago. Worked on a ton of amazing things here. And AI started to become more and more powerful. And so of course, given my background, couldn't resist getting really excited about the possibilities of AI with Coda as well. And maybe to step back a little bit. We're all product managers here, interested in product. A lot of people call PMs the mini CEO of the product. Or Jack of all trades, the human glue, the conductor. What does a PM actually spend their time doing? What does my experience spend? Is that is occasionally true, but the reality is much closer to being a human search engine, being a copy paster, being a content massager. You are the glue, but it's really messy glue that binds different people and different teams together in a project. And so a little bit of my story of what work has felt like for a long time is someone asks a question in Slack or an email and your whatever tools you're favorite on. No, how does this thing work or is this a bug? And you're not actually sure the answer. And so what do I do? I'll go pull up Jura or my bug tracking tool and see is that an open issue? Is this something that's known? I can't find it. I'll dig around the docs. Maybe this is actually working as intended. Maybe someone's thought about this. I'll see, okay, is this really significant or other people complaining about this? I don't see it anywhere. How much time should I be spending on this? And pretty soon you have about 100 browser tabs open to answer one small question. And this is the life of a PM. And this is part of why I got excited about Kota. So I'm curious if anyone here has used Kota, please chime in in the comments. But for those who aren't familiar, Kota sits kind of at the intersection between rigid purpose built tools like Salesforce, Jira, Product Board, and free for all flexible surfaces like dots, sheets, and PDFs. And it has the best of both. I was a long-time Kota user before joining here. And part of what sold me on Kota was how you could really bring all of these different 100 tabs that cross a PM's life into one place. And so today I as a PM, I use Kota as a team hub, which is the one link to send for how a project is doing, what we're working on, what's happening, as tracker all the way from company level trackers with OKRs to sprints, who's doing what, and for write-ups that have extra interactivity and structure where people can vote on things and give quick reactions as they go through write-up. And Kota for me has been amazing as a single source of truth, simplifying process, creating visibility across teams, and most of all, speeding things up, speeding up my time, speeding up the team's time. So I tell you all of this as background of a little bit what my baseline has been like since using Kota, normally without AI at all. And I think that the reason that context is helpful is AI takes a lot of these same trends for me around bringing things together, moving faster, reducing redundancy, making all the team easier to contribute and pushes it to the next level. Like they're a perfect fit. And so I got obsessed with this idea when ChatGPT came out around what is busy work? And as a PM, I spend a lot of time on busy work personally. There's so many times I'll spend hours every week writing up next steps from different meetings, turning them into issues in task trackers. I'll be summarizing how different projects are going to different audiences and stakeholders. And the reason we end up as PMs doing a lot of this busy work is it's valuable. People find it useful to know what's going on to have someone who's triaging all the bugs, who's looking at all the feedback, who's doing these very repetitive manual things that create a feedback loop and create connective tissue within the team. And so busy work I found is valuable, but it can be very tedious and can be very repetitive. And so got really interested in this idea of like what if AI can help with this problem? Now, there's been a lot of AIs that have been coming over the last year that are writing assistants, and those are great. Writing assistants make content. But to really get rid of busy work became obsessed with this idea of like what is a work assistant? And a work assistant makes progress, not just content. And progress can take many different forms. And so as we got to working on building COTA AI here, we thought about flexibility and power of basically having building blocks that are AI integrations that can power really incredible solutions. And so we have one on a canvas that will help you write AI in a column that will do all sorts of tasks for you in what we call a block, which can synthesize context. And so wanted to show you a little bit of how I've been using these AI tools that we've been building at COTA as a PM, eating my own dog food while working on them to save time and have more impact. So here we are, actually, we were in a COTA talk this whole time. This was an embed into a Canva presentation, and I'm just popping out the navigation here, and you can see this is one page among many. So this is actually very similar to the team hubs that I'll have for different projects and different teams of things I'm working on. So usually these have, you know, a place for information about the team, what we're working on, and went to walk through a bit of a story of a scenario of how, as a PM, touching many different parts of a project, I've been using AI to save time. And so we'll begin the story with users, with feedback. And so feedback and feedback groups are the life cycle, the lifeblood of really productive teams, in my opinion. It's all validate where you're going. And so a very common thing I do is join customer calls all the time, you know, many days, multiple. And COTA has hundreds of different integrations, with Zoom transcripts, with Gong recordings. And so I started basically taking customer calls I was in, and integrating them into COTA as a table. And so you could see each row is a call, and here is basically the transcript of a call. And what I used to do is basically every call I'm on, I don't really get credit for the insights, unless a lot of the team can understand where those insights came from and why. And so for me to say, oh, I was in 20 calls last week, we really need to do that feature, is way less compelling to the whole team, than saying, here are 10 quotes that exactly explain why this is really important. And created that level of direct feedback creates so much buy-in from the broader team, and why are we doing what we're doing? Who is this helpful for? And so in the old world, pre-AI, I would spend a lot of time each week writing out my conclusions from talking to customers, copying other's key quotes, which is really helpful, builds my credibility, justifies our roadmap, is the sort of core insights we'll build our loop around, but it's a lot of time. And so one of the things I started using credit AI for is we have all these different interviews here, is I would go ahead and ask AI to summarize the transcript. And we have these nice drop-downs, say in one sentence, and I'm gonna go ahead and ask it to fill it in. You can see AI here starting to work, and every single one of these long transcripts now got summarized by AI. And the really cool thing about this is I can ask for a slightly different version. Let's update this one. And we've designed Code AI to be an assistant. It's a co-pilot, it's not an autopilot. And so each of these different responses are editable. So actually, I remember this call with Oliver, I was in on it, and I thought there was one other really interesting part, which is Oliver is currently looking at 20 tools to solve his problem. And so we've designed all of these to be editable. And so I found this type of AI to get me to a first draft, the 80% of effort to get most of the bones of something in literally seconds. And I'll often go through an interview and make sure these things look great, but it just saves me so much time. And here's another example. We have a whole taxonomy built out for future requests on what are the tags. We can look at trends over time. Really helpful to look at this type of aggregate data. Again, really painful to annotate. So I say, all right, AI go ahead and tag any product feedback from this transcript by the feature type. And the feature type is this table. This is a column in Coda that is a lookup into a whole schema of a bunch of options. So I'll go ahead and hit fill. And wow, it basically categorized instantly the type of feature requests that came up in each of these different calls. And there's tons and tons of different things here. And so one of the things I've loved about using AI is just how quickly I can scale it up for these types of repetitive, valuable tasks. And then I can spend my time reviewing, tweaking and telling the broader story of here's why these things are helpful or being in more customer calls instead of writing down all of the summaries afterwards. So I've been really, really finding this to be super helpful. So I have all these feature requests. I've got a whole bunch of insights. I need to go and write a practical requirement document. This is many people consider like one of the core jobs to be done of the PM is like, why should we work on? What should we work on? Like what are we building? Why for whom? And so for me, I'll say often product requirement documents, PRDs, I end up doing them, you know, at the last minute when I really need to do them to get a bunch of stakeholders bought in or to finalize some key decisions for engineering. And some of that for me, honestly, is like writer's block. Like it's kind of daunting knowing this is a core thing for PMs to get right to go and look at, you know, a blank canvas that looks like this and be like, man, I have a review coming up on this approach. And like, I have so much stuff in my head and knowing where to start and to get it out can be really challenging. And so I've been loving using AI to get again to that 80% starting line that first draft that I can react to really, really quickly. And so I'll go ahead here and just open a little slash command and say, I'll just put in a prompt here I was using earlier, write a two page product requirement document about a new image editor based on feedback from this feature request page. So all I do here is just go ahead and use an at reference to a page. Could have been a table, could have been a range of other things. And please include a section on the problem statement, a table of target audience and proposed features. And so I'm giving AI here context about all of these different transcripts on that other page, all the different future requests that customers mentioned. And let's see how it does. And so one of the really cool things here is AI can help with structure in addition to content. And so you'll notice it created this table of target audiences. It has an audience type, it has a description. And then we have this problem statement here and some proposed features. And maybe I'll say, you know, this problem statement, I'm going to go ahead and ask AI to refine it. I'd say, please make this twice as long and focus on why this problem matters. And I'll go ahead and ask AI to help. And here we go, I've got a couple of more helpful sentences. And again, it's just making that feedback group much, much easier and tighter to work with. And so one of the things I love about write-ups in Coda is they can be super interactive. And so one of my favorite Coda features in general is adding these little reaction buttons like thumbs up or thumbs down throughout and saying, hey, do you agree with this problem statement? And so when I send this doc to some stakeholders, I'm not wondering what they thought of it. I know at each point if they're bought in or not. And you'll see this pattern show up a bunch in the rest of our story. And so let's say I finished my PRD here and I'm going to take it to my weekly meeting. And so the way I run meetings in Coda is I have a table of topics and people can vote on them. And so I can get a sense of what people are most interested in. I can see, oh, two people are interested in this. Maybe I'll just grab them offline instead of using the whole group. And as the person running the meeting, I'll make the final choice, but use this as an input. And so people will vote. And so I have here a link to my products brief. And we'll have discussions and usually we'll do that right here. But as much like the transcript you saw earlier, I'll take notes live during a meeting about what people said. And again, at the end of my day, I'll be like, oh man, what are actually all the next steps based on this meeting? Like what needs to happen now? Who do I need to ping? I need to send a Slack reminder to people. And so I've been using AI for this, which has helped a lot. So I'll go ahead and add a new column here and I'll say find the action items from this notes column. And let's show it as checkboxes and let's keep it to a sentence per checkbox. And we'll go ahead and have AI come up with it. And so these are our next steps. And this is awesome. This is like a first draft of all the next steps that I would have had to copy paste by hand, slightly reword based on the discussions. And I can go ahead and paste this into a bunch of places or link to this in other parts of Coda. And so I've been, this has been saving me a lot of time. So let's say we had our weekly meeting here and everyone's really excited about the product requirement document. We go ahead and build it and we're ready to test it. Testing is a really important part of our feedback group. And so we have a bunch of trusted testers. We have a whole program of real users and I'd like to reach out to them. And one of the really cool things Coda has is these really deep integrations. And so I can go ahead and integrate my Gmail account, which I authenticated and click a button to basically send content from this draft thing from my email. So I can send emails from Coda, which is great. And so what I'd like to do is have a bunch of really personalized emails to teach these people that I can go and send out. So I could say, create a email, 50 words about a new image editor feature. Here's the contact, here's their title. Invite them to be a trusted tester. So I'll go ahead and generate these with AI and get a nice first draft out here. And maybe I want something that's like a little bit more personal. And so I'll say something like key points. Jess, I remember she has a dog. She has a dog. And this is context that doesn't really fit a clean template. I'm going to ask the AI in a second to incorporate these key points if it can. But it's not going to feel like mad libs where it's just throwing in a part of the email. Alan, I remember really cares about tech debt. So I can find a way to mention that. That's great. I'll go ahead and revise my prompt. So I'll generate an image. This is the contact. This is their title. They work at this company. Invite them to be a trusted tester of it if you can find a natural way to work in key points. So I'm just going to use at. And then the name of that column, please do the emails from David. And I'm going to go ahead and fill it out. Again, so here we go. Let's see. This one didn't mention the dog. This one, we said, given your passion for tech debt and dedication to addressing the tech debt, we believe your feedback would be invaluable. So one of the things I've been really enjoying is making this finding a pattern here that works and then scaling it up to, again, a really large group of people where it used to be. I would take a lot of time to write each of these by hand to get our trusted testers who are doing us a big favor, knowing how much we value and care about them. And I'm still able to do that, but do it a bit faster, which has been great. All right. So one last practical tip about code AI. We also have this thing we call AI block. And it was designed to help synthesize a lot of context. And so here we have a very typical project tracker. There's a timeline. There's a table of projects. And actually one really cool thing is these are actually the same table. And so this is a cool code, a feature called views, where you can basically have... Let me move my face here. So you can basically have data represented visually in a few different ways. So I could go ahead and change this here. And this changed as well. You can see these are basically the same data. And so I could say, all right, there's a whole bunch of context on this page. I have this block here and I like to summarize what the team is working on in the roadmap. And this is literally, you know, one of the most commonly asked questions of me as a PM. It'll come up at the start of executive reviews. What's the team working on? What's the status? It'll happen in Slack from dependent teams and like, hey, how's this tracking? And so it'd be really nice to just have a natural language response that's super easy to update as the team makes progress. So I'll go ahead and click create. And this has a summary of what's happening. And a cool thing is I can go ahead and refresh this. And so one of the things I have now is basically at the top of all of my trackers, all the little block that's just like, here's a summary of what's going on. And after our stand-up, I'll go ahead and update it. And if anyone asks what's happening, I'll just send them a link to that page. And I'll say, so you can check it out. You can see our roadmap and you can see a description of it in natural language, which has been, again, really helpful. In engineering, there's this concept of dry, don't repeat yourself. I feel like as a PM, there's a lot of repeating yourself in slightly different ways to slightly different audiences. And that's a really important part of the job. Communication, creating by in speaking different people's language is a lot of what makes great product managers really effective. And I think AI is making that a lot more accessible and a lot faster. And so I can ask AI to speak to a slightly more technical audience or a slightly less technical audience and not repeat myself. I can write one thing and ask AI to translate it, which has been really, really great. All right. So I gave you a few different examples of over the course of a project from taking customer requests, summarizing them, annotating them, writing with AI, running a weekly meeting where we use AI to summarize next steps, personalizing outreach to testers and summarizing what's happening in our team. I've been literally doing all of these every single week and it has saved me hours across each week and has been incredibly helpful. So these are a few of my favorite ways to leverage AI as a PM. And I would also say for folks who aren't currently PMs but are excited about product and are kind of curious about it, I think it's making a lot more accessible to do things that might have taken a lot more energy before as well, which I think is really awesome. So if you're excited to try out some of these tricks yourself, you can get started with some really great templates. So if you go over to coda.io. slash gallery slash AI, there's a ton of awesome AI tools and interactive documents for writing product requirement documents, having trackers, running customer research with AI that will help generate questions and summarize it, running brainstorms with AI where it can come up with other provocative questions and ideas. So yeah, I definitely encourage folks to check it out. And let me know what you think. You know, we're always working on code AI and as a PM, I'm also always looking for hacks to be more productive, save time. And so I'd love to hear what you all are doing with AI to move faster with your teams together. So thanks so much for tuning in today and yeah, let me know what you think. Go ahead. Goodbye.