# Run Your Product From Chat

Thomas Hartmann, Just Product 2026, 13:00. Automatic transcript of the livestream (Parakeet ASR, unedited, expect misheard names). Timestamps are positions in the talk video.

[0:02] All right, so the next speaker, and please mind I'm terribly biased uh with the next one.
[0:09] Um, is my partner in crime for now five years.
[0:13] Um, we have been running this conference together for the last uh four years, and uh went through our own like kind of crazy transformation um in in the last year, and he's gonna tell you about it and also what that means for your personal product workflows, right?
[0:32] So the talk is called run your product from chat, and please welcome with a crazy clapping, stomping, shouting, applause.
[0:46] Thomas Hartmann.
[0:56] Ah, he's he's doing the long way for the dramatic entrance.
[1:00] I love that.
[1:02] Thank you.
[1:03] Have fun.
[1:04] Thank you very much.
[1:05] So I need the slides.
[1:09] Or should we do it without slides?
[1:11] Ah, okay.
[1:12] Run your company, run your product org, run your product from chat.
[1:17] Basically, it's all the same.
[1:20] Ah, shit.
[1:22] I sorry, I forgot something.
[1:24] I need to do some work.
[1:25] I hope you don't mind.
[1:26] It will only take about two minutes.
[1:29] I have an email which I need to answer, and I wanted to update the button on our website.
[1:35] Let's see.
[1:38] I got an email from Julian.
[1:40] He asked for a team ticket, and he can only he needs an invoice or an offer to kick off the purchase process.
[1:50] Would you please create the offer?
[1:54] Create an email.
[1:56] And what else do we need?
[1:57] Go in event write and create the voucher codes so he can register.
[2:02] We are kind of um late, so please don't ask questions.
[2:10] So let's see.
[2:12] Okay, let's do that.
[2:13] And then the other thing was Sebastian told me that our website is terribly boring.
[2:28] Should should be something playable, should be should be fun, but not too I don't know.
[2:33] Um just come up with two or three suggestions.
[2:36] Uh show me the prototype, please.
[2:38] Um, and open it when you're once you're ready.
[2:43] All right.
[2:44] Um back to the talk.
[2:46] Sorry about that.
[2:47] Uh while the AI is working, uh, I will uh talk you a little bit through how we what we experienced in the last year, right?
[2:55] What that what that means for AI and product management, and uh last but not least, I will talk a little bit about the impact in um of AI on product management, right?
[3:07] So the last year, I think there was like two moments where we really which was like pivotal moments.
[3:15] Yeah, the first one was October 2025.
[3:19] I was at a workshop at a client, Hackathon for product management, and actually that guy in the middle, he really blew my mind.
[3:28] What he did is he's an engineer/slash product manager.
[3:33] So he interviewed another product manager um site and extracted all the knowledge he got from his interviews with the end clients, right?
[3:42] With that knowledge, because he had access to the uh GitHub repo, he pulled the repo and he looked at the algorithm they used to select you know photos for photo books, and he looked at the algorithm, he brainstormed with the AI how they could actually potentially change the uh sorry about that, um, how he could potentially change the the algorithm, right?
[4:09] And after two hours, he had basically a prototype.
[4:14] And I was holy cow, that used to take me four weeks, and he did it in two weeks, uh two two hours.
[4:25] It's like boom.
[4:27] And I was really that that really blew my mind.
[4:30] And I really, I came home and I thought, we're all going to hell.
[4:34] AI will replace all of us.
[4:37] And I had sleepless nights, and I told Sbastion about it, and really I thought, okay, this is like it's unbelievable.
[4:45] It's so fast.
[4:46] It's so, it's I I felt that I saw the future.
[4:51] Then there was a second point in time which really struck us.
[4:56] And that was end of January.
[4:58] And over the weekend, I built an app myself for a site business of mine.
[5:04] Um, and all which you can see is really working, you know.
[5:07] I had filters in there, there was a map, you know, you can go to the listing pages and stuff like that, right?
[5:14] And proud as hell on Monday, I turned into office and said, Sebastian, I built an app on my uh myself on the weekend.
[5:20] And he was like, like developer, right?
[5:24] Looking to a non-developer, yeah, okay, show me.
[5:27] And I showed him the app, and he was like, okay, and you know, like I could tell, okay, he gets more and more impressed, right?
[5:35] And at some point he said, like, okay, how long did it take you?
[5:39] And I was oh, the whole weekend, you know, it really captured me.
[5:43] I was the whole weekend, you know, doing this stuff, and he said, like, what?
[5:48] And I was like, huh?
[5:49] I didn't understand if he meant it was little time or a lot of time, right?
[5:54] And I said, how long would you take to build that thing?
[5:58] And he said, like, two to three months.
[6:01] And I was like, what?
[6:04] How is that possible?
[6:06] How can I?
[6:07] I've never written a line of code, create an app which a professional software developer would say, I need two, two, three months.
[6:19] This is like mind-blowing.
[6:24] Yeah, and from that point in time on, we've been aligned.
[6:29] On Houston, we have a problem.
[6:33] There is something seriously changing in this world.
[6:37] And our first question was: will anybody come to the product masterclass if AI can explain everything?
[6:45] Why would somebody come to us to get knowledge about product management if AI can do everything, right?
[6:52] And there was like a lot of you know other questions like, okay, what's left for human?
[6:56] Is our job secure?
[6:58] Is I don't know, like how whom of you felt that way?
[7:02] I I I bet many of you, you know, experienced that deep down and say, like, okay, but I feel super unsecure about the future.
[7:11] I don't know where this where this is heading to.
[7:14] And that was really that moment.
[7:16] That's like really, okay, no idea.
[7:19] Yeah.
[7:21] And from there, you know, like after a while, we looked at a situation and said, okay, on the one hand, you have this thing and it feels superhuman, right?
[7:31] This thing you can code an app on the weekend.
[7:34] But on the other hand, side, if somebody writes me an email that they need a voucher, uh an invoice for the just product, I'll go to Eventbrite, I look that look up this bloody feature, you know, voucher code, which I used to never can find, right?
[7:51] I open Fastbill, I need to check if this person is already in the system.
[7:55] I create the account name, you know, in in between, you know, somebody calls or something distracts me.
[8:01] I come back, right?
[8:02] I download the PDF, I draft the email in chat, then I put it into Gmail and I send it out.
[8:08] Takes me at least 20 minutes, 30 minutes, or sometimes I have it in the back of my head for two days because I just don't want to do that.
[8:19] Yeah, that's the situation.
[8:21] On one hand, super intelligence.
[8:23] On the other hand, side, super, you know, like stuff, which is annoying, right?
[8:30] So we made the decision.
[8:32] We want to use AI to automate our stuff, right?
[8:36] Our workflows.
[8:38] So whatever workflows we had, we started to automate, right?
[8:43] Publish a blog post.
[8:45] Does Webflow have an API?
[8:46] Yes, oh great.
[8:47] So let's build a skill, embed the process, and perfect.
[8:52] Next day, you know, build an uh write an invoice.
[8:55] Does uh fastbill have an API?
[8:57] Yes, connected to Claude, wonderful, it's working, we have a skill for that, right?
[9:02] And we did that over and over.
[9:04] So all the tools actually are connected to Claude now, right?
[9:08] And then at some points, we really got annoyed with the tools which don't have full API access, right?
[9:16] Writing a newsletter, oh Sebastian, can do you can you do that?
[9:19] Because you need to go in there manually, you know, pasting stuff, really annoying stuff, right?
[9:26] Or stape here, you know, N at N, you should just say chat what they do, upload it, right?
[9:30] Webflow, you can upload into the CRM, but you if you change the structure of the page, it doesn't work with right.
[9:39] So we replaced all these tools with full API tools.
[9:44] And at some point, we built our own applications on top of that, right?
[9:50] And that actually created the product masterclass operating system step by step.
[9:58] And here is how it looks like.
[10:32] It has full access to our uh GitHub repos.
[10:35] So Claude can access these tools and the context.
[10:40] And on top of that, we've built the skills with which embed our processes and workflows.
[10:47] And that allows us to get amazing outputs.
[10:53] That's the operating system we built for ourselves.
[10:58] So connections, context, you give that to an agent, you build workflows on top of that, and you will get amazing output.
[11:05] That's an operating system.
[11:11] And let's see how that works in operation, right?
[11:15] So we just tried to get a just product offer, and you can see highlighted which you know connections with which context and which skills it used.
[11:27] Yeah.
[11:28] And in the background, hopefully, yeah, we'll see.
[11:31] 5050 chance.
[11:33] We'll see if Julian has now, or if I have now a draft email for Julian with an offer.
[11:40] Because hopefully the AI went to Eventbrite, created a code, yeah, went to PipeDrive to see who's it who is it, went to Fastbill, created an offer, and went to Gmail to draft an email.
[11:55] So let's see if that worked out.
[11:58] Oh, sorry.
[12:01] Oh, done.
[12:03] Voucher.
[12:04] Okay, I cannot read so much.
[12:07] Where is the offer?
[12:09] Open it, please.
[12:13] Show me the email.
[12:16] Okay.
[12:18] Ah, here.
[12:20] Power smile, happy Julian, good mood street.
[12:22] That looks really good.
[12:25] And oh, we don't want to do that.
[12:28] Uh, and the invoice, five tickets, wonderful, two thousand euros, and I can just go back and say in please send the email.
[12:45] The email.
[12:47] Wonderful.
[12:48] Task one seems to be seems to be done.
[12:51] Wonderful.
[12:52] So that saved me.
[12:50] I don't know.
[12:57] If it took me 15 minutes, it's five times faster.
[13:01] If it took me 30 minutes, it's 10 times faster.
[13:04] And this is this is, you know, 10x.
[13:07] Yes, of course.
[13:08] It's even more, I would guess.
[13:11] And we didn't just do it for this workflow, we did it for all workflows we have.
[13:16] So we have, I think, 46 skills, uh, maybe even a little bit more, and each of these workflows saves us a ton of time.
[13:26] And what's the impact of that is not just that we are amazingly insane fast, but it frees up space.
[13:35] And you know, like it even enables us to do things we couldn't do beforehand.
[13:41] After somebody is in the product masterclass, we always do a review call uh call for half an hour.
[13:47] And after the review call, we chill which you just give the video to the to an AI, and the AI creates this page, you know, it says, okay, which angle we should use.
[13:58] It also tells us okay which parts of the video we should actually use.
[14:03] It uh it creates subtitles, right?
[14:06] It sh it it cuts the videos and it creates the whole blog post, and that takes you two hours on the side.
[14:14] If I would do that myself, I would at least need three days straight focus focus focused work for that.
[14:21] So this is this is literally it's insane.
[14:25] We couldn't do that without AI.
[14:27] It's not possible.
[14:29] And you know, like on top of that, we have way better decisions because we have all data at hand, right?
[14:35] We can just ask, okay, who was visiting a website?
[14:38] How many people, you know, like uh how many people go to conference?
[14:41] Is this person already registered at a conference?
[14:43] All these questions are at our fingertips.
[14:46] Yeah, and the pure goal is if you have all your operation, you know, automated, you have really free capacity and space for the strategic questions.
[14:59] Yeah.
[15:00] And that's that's really that's gold.
[15:04] So and I think what works for us will also work for you.
[15:08] You can do that yourself as well, right?
[15:11] So let's see how you can you know take that system and bring it to product management.
[15:16] Because that seems to be complex, right?
[15:18] It it looks overwhelming, but if you work in product management, I think you can start really small.
[15:25] You connect your favorite AI with Cheera Confluence in GitHub, and you have your first product operating system.
[15:35] And with that, you can already do crazy things.
[15:40] Yeah, let's see.
[15:41] Let's just try with a feature, how that could actually work, right?
[15:46] So we started to develop a feature at the beginning of the talk, right?
[15:51] And it's a really small one, I know, but just for you know, see how that works.
[15:58] So, Product Masterclass website.
[16:00] Hopefully, it went to GitHub, pulled a code, hopefully it brainstormed a little bit, playful buttons, right?
[16:07] And uh hopefully it did a prototype which we already somehow saw.
[16:14] Let's see.
[16:16] Okay, enroll now, enroll now, and we have enroll now.
[16:22] So let's see.
[16:19] Let's let's hear.
[16:24] Do you like this one?
[16:26] No?
[16:27] Make some noise.
[16:28] We are I don't see you.
[16:30] Okay.
[16:31] Is that one better?
[16:32] Would you like that one?
[16:34] Okay.
[16:35] Same.
[16:35] Third one.
[16:38] Third one is the best one.
[16:39] Okay.
[16:41] Alright.
[16:41] So then let's see whether we can.
[16:45] Okay.
[16:47] Enroll button.
[16:49] So I really love the third button.
[16:53] Um please uh write a ticket, uh, develop it and push it directly to main.
[16:59] I know we usually don't do that, but I have 250 people uh from QA sitting in front of me.
[17:06] They will make share that everything works.
[17:08] Uh please don't ask questions, uh directly push it to main.
[17:16] All right, let's do the uh let's let the AI do the work.
[17:21] So we did that one.
[17:24] And it will create a beautiful ticket in the background.
[17:28] Let's see, maybe it's already there.
[17:30] No, it's not yet there, but I show you what happens in the in the in the background because it will create a really long story.
[17:37] First human readable, and then down there, there will be you know like a lot of implementation stuff, like which uh which stuff it needs to touch and how to implement and and stuff like that, right?
[17:48] I honestly I never written uh such a long uh ticket myself.
[17:52] Uh the AA does it.
[17:53] Uh I'm really happy for that, and that's why it actually is um you know possible for humans to develop it or the agent itself, which we which is my preferred uh solution, actually, right?
[18:05] And yeah, of course, if you look at this again, right?
[18:09] It would have taken me one sprint for preparation and one sprint for for implementation, I don't know, two to four weeks, and now we've done it hopefully within this talk.
[18:20] So AI is enabling you to be crazy fast, yeah.
[18:26] Iterating a prototype in a meeting with your client or you know, like this is like crazy, crazy speed.
[18:35] So what is at the end of the day the impact for you and your role?
[18:42] Will AI replace PMs?
[18:45] I think no, not at least not for now.
[18:49] Yeah, AI won't replace PMs.
[18:52] PMs who use AI as I showed you, will replace those who don't.
[18:58] I think that's the truth for now, right?
[19:03] Will is AI better than humans in every task?
[19:08] No, not at all.
[19:09] Yeah.
[19:10] I think at the moment you should look at your at your workload and divide it into protect what you as a human are better than the AI.
[19:19] And uh what the AI is better, right?
[19:23] You want to master the first one.
[19:20] You want the protect part.
[19:27] This is what you want to master for yourself, right?
[19:30] You want to understand your customer, you want to make the strategic decisions, yeah.
[19:36] You want to have that product sense, yeah, which which helps you to create value at the end customer, right?
[19:43] And everything else, you try to push as far on the right side as possible.
[19:49] The more you can automate automate, the better it is, the more free space you have.
[19:54] And I think that's what you saw with Verena this morning, right?
[19:57] She pushed a lot of stuff into the automate things, right?
[19:59] And that gives you free space to think about more strategic decision about the the important stuff.
[20:07] So, do we need product fundamentals in the product masterclass?
[20:12] That was the our question.
[20:14] And yes, hell, we need it more than ever.
[20:20] Think about it as, or we think about it, AI is a lever, right?
[20:25] A fool with a tool is still a fool.
[20:29] Yeah, I think you would agree if you see that, right?
[20:33] And a bad product manager with AI is even a worse product manager, because he or she will create on the work on top work, right?
[20:44] Because it's like it's just a disaster, right?
[20:47] But on the other hand, side, if you have a great product manager.
[20:51] With AI, it's a freaking machine.
[20:55] Yeah, and he or she will feel superhuman.
[21:00] Yeah, I think I really it's a lever, right?
[21:03] Good is even better, bad is even worse.
[21:08] So, how will product management change?
[21:10] Yeah, and I think um most of you know the term builder by now, yeah.
[21:15] So nobody stays in their in their lane.
[21:17] You know, everything will merge together.
[21:20] So you you you you're more of a uh PM slash engineer slash designer, right?
[21:26] And it everything merges together.
[21:28] But there is a second picture, which I actually find even better.
[21:32] Yeah, imagine you are a three-star Michel star chef cook, right?
[21:39] And you still stand in the pass, right?
[21:42] Your AI agents, they cook something, and everything passes by, and you look for the hair in the soup.
[21:49] And if you find it, you send it back to the kitchen.
[21:53] That's your job now.
[21:54] There is a lot of stuff coming, and you need to review and find the hair in the soup, and that's really really hard, and it's really drain energy draining because you get pages and you know, like a lot of stuff thrown at you, and you need to judge whether this is good enough or not, right?
[22:12] And there is a second picture which I picked up at another conference.
[22:16] Engineers are the babysitters of the AI.
[22:18] Basically, that's describing the same thing, right?
[22:20] So AI does the thing, and you are the babysitter for for them.
[22:25] So, my three takeaways to be relevant in the future, you need to master AI in your field.
[22:34] Yeah, if you want to be employable in five years, you need to three to five years, doesn't matter.
[22:40] You need to master AI in your field, definitely.
[22:44] There is no way around, yeah, and you better start today than uh than tomorrow.
[22:50] The second thing is we've been we've gone through a really you know depressing phase, so to say, right?
[22:57] Like, where is this all going?
[22:58] You know, like uh, you know, like nobody will be employed anymore, and so it's it's so much fun to work this way.
[23:05] It's unbelievable.
[23:06] I don't know like how much software I built without you know writing a single line of code, and it's super fun because you know, like you can be so creative, it really really is.
[23:19] I don't know, it's its best, yeah.
[23:21] And if you ask Sebastian, he he is glowing eyes, and you know, like because he's a builder himself all um all the time.
[23:30] Yeah, and the third one is with a tiny team, you can have huge impact.
[23:36] Because you can, you know, the lever is working for you if you re uh use it correctly.
[23:43] But I know you know, like the times are really sometimes scary, and there is a lot of news, and it's hard to get started, and you know, like know stay on top of the news, right?
[23:56] And it's it's hard to if you if you're not in there, it's hard to find a way in or a starting point.
[24:04] Yeah.
[24:05] Well, we have a starting point, the AI product masterclass just does that.
[24:10] You know, we help um people adopt the AI systems for themselves.
[24:16] Doesn't matter if it's personal for you or you for your team.
[24:20] We have um two starting dates coming up.
[24:22] If you're fast, we see each other in the next weeks.
[24:27] Um please check out check out our alumnus stories.
[24:31] Um really proud of the of this page.
[24:34] Not only because AI created all the the detailed pages, but I think we had an impact in all of these lives or all the, you know, like in for all of these people.
[24:45] Verina is in there too.
[24:47] So yeah, just check it out.
[24:50] Thank you very much.
[24:58] Thank you so much, Thomas.
[25:00] Um, and the the big question is now do you have questions for Thomas?
[25:07] And the second question is where's the cube?
[25:12] Ah, the cube is over there.
[25:14] So do you have there is a question over there?
[25:19] All right.
[25:20] Throw it over.
[25:21] Like look that way, please look that way.
[25:23] Throw it, throw it, throw it, throw it.
[25:25] Throw it, throw it, throw it.
[25:27] Yay!
[25:28] Alright, so now you see why you should always look that way.
[25:31] You okay?
[25:32] All good?
[25:33] Okay, good.
[25:34] Fantastic.
[25:35] I I have a quick question.
[25:37] The you talked about protecting, augmenting, and automating your work.
[25:41] And you talked a lot about the protecting and the automating.
[25:45] How do you know that line between what you augment versus automate?
[25:51] Um I think everything which you are sure that the AI can do without interacting with them, just automate it.
[26:00] Yeah, but um, for example, for me, augmenting is like sending an email, I want to review it.
[26:06] Yeah, but there is certain stuff which you can just, you know, like um conveniently or confidently let the AI the AI do uh themselves.
[26:16] So that's I think that's the that's the border.
[26:21] Yeah, front.
[26:25] Here?
[26:19] Yeah.
[26:26] Hi.
[26:27] Um I think the whole prototyping part as a product manager, of course, super fun.
[26:38] But I also have a designer.
[26:41] Yeah.
[26:42] That could I'm trying to figure out how to work with him and with designers in general.
[26:49] When we have an idea, then I do whatever wireframes or mock-ups.
[26:54] But how do you work with your designer?
[26:57] Or how do you see this type of we have a design system and the AI pulls the design system in a bigger company would have a designer, and hopefully the designer is responsible for the design system and the AI uses the design system.
[27:11] I think that's the way forward.
[27:13] Designer is responsible to maintain the design system, engineer is responsible to keep the
[27:20] Code base is maintainable, and the product manager, what does he do?
[27:24] Nothing.
[27:29] Thank you.
[27:32] Alright.
[27:34] One more question.
[27:36] Then sadly move on.
[27:39] Virina.
[27:45] Oh.
[27:46] Oh, I said one more question.
[27:48] Maybe you can ask it later.
[27:50] Okay.
[27:51] Thank you so much, Thomas.
[27:54] Thank you.
[27:58] And also, although you bought it yourself, get to take one home.