AI Is Rewriting Product Development

Machine-generated transcript from the episode video, cleaned for punctuation and names. Speaker labels are not included.

[0:05] Hi there, it's Rick Koleta, your host from the GTM Vault Podcast, and in this episode we bring you Dmitry Stavisky, a seasoned full stack startup adviser with extensive experience in building and scaling tech companies internationally. Dmitry shares his journey from climate modeling and coding to leading product development at unicorn startups. He also discusses how AI is reshaping product development, the challenges and opportunities it presents, and what the future holds as AI goes mainstream. Additionally, Dmitry offers insights into his advisory services for companies post Series A on international expansion and early stage startup founders. Tune in for a deep dive into AI's impact on product development and marketing. Dmitry, thanks so much for joining us today. I thank you for the invitation and for the kind nutal. Good to see you starting the S and now broadcast. Yeah, we

[1:11] met before the S St, or before your s St, at the life de. It was a good event. Glad to hear. Yeah, it was a pleasure hosting the event and thanks again for being our speaker. Now I want to get right into it, Dmitry. With your extensive background in building and scaling tech companies, how have you seen the landscape of digital product development change over the last decade, and what do you think are the most significant differences between developing digital products in the past decade compared to now? All right, well, this is what happens when you grow great, you get asked about for last de. Okay, let's recollect what the industry was like 2014. Look it up. 2014, Apple launched iPod, iPhone 6 with the larger screen, Apple Watch and Apple Pay. Microsoft launched Windows 10. Remember that? That's was

[2:16] launched was with Alexa. That's what the majors were doing. Things were changing rate and then suddenly this D. So what happens since? I think, from the user perspective, the biggest change was high expectations for UX, performance, aesthetics for product design, and greater emphasis in design are thinking. We fully switched to mobile first. It wasn't the case in 10 years ago. Product teams got into the good habit of continuous customer engagement. Again, this is F believe, but it wasn't a common practice years ago. We are much more reliant on data analytics and testing sense

[3:20] datais nor. It's a big deal in technical terms. Ifed to computing, it was just starting years ago. Now it's there. Whole API ecosystem flourish. These were all good things. Now, not so good things. 2014, you could get much more traction of app stores. The last 10 years, app stores got much more proud for user acquisition. That meant high. It's my me to. Yeah, the noise. What else? Continuous integration, CI/CD in divorced profession, standard protic practice. Regulation is much more serious these days, starting with GDPR and then Californian CCPA, which affects

[4:25] basically everyone now. That was the gradual change. In 2017, Google published P further, Attention Is All You Need, Bally suggesting the transformer architecture for machine learning. That's when things started to accelerate. So over the last, I would say, two, three years, it was all about this is and disruptive, and this is the major change which is still happening. Well, the also has changed the Ling of the whole cor C was transition to remote and work. Well, this is a change for, but this is also in change the. And then, well, the whole macroeconomics and geopolitics. Real life economy is back. People are buing robots. We are

[5:28] industrial in. People got serious about F spe. Yes, it how was that for the. That's great. That's a lot. It wasn't decade, totally. And when you look at how AI is changing the way we build digital products, can you talk to us a little bit about that, Dmitry? And if you want to share some specific examples from your experience building companies like Ed win or buddy. a or even Evernote. Yeah, sure. So this transformation is still in progress and it's happening very fast. Our conversation will be on this topic, at least we have a historic interest in the for now. Here's what I see. There are two aspects to that: how the AI/ML affects internal processes and the tools compiled by

[6:30] and how does it affect product functionality and user experience run time. So on product functionality, this means return of the conversation interfaces. Actually, remember 2014 was the year when Amazon launched that was the first attempt to transition from buttons, windows and Ice negation, the faes. It was Prat, the underline technology supp. So now, so you get, you asked about, we were building, we started 2016 and we were building a tutor of English as a foreign language, and the Harvest langage still learn teach is speaking. So we were very excited about home system and try to build conversational exercises on that stack. Built it, it worked. That got us some interest and inves from Google, but it

[7:36] was very expensive to produce content for this exercises. So it was 2018, 20. Fast forward, B, this is the company that b is also English as a foreign language, early childhood education for young kids. For them, building andal exercises much easy and much more Prof. Evernote, well, Evernote, this external exal bra taking in fire. 2014, mature technology and product. We had search, we had grammar phys, so they were ninjas, not searches. Today, new team at build and then face basically what it looks, what you would expect from cting on top of your D, no selection, and I sure that

[8:42] it took them less effort to ask to develop that verar based powerful search. Now some of that TR in the cloud, what much of the conation fun actually can cap the device, but is SC mostly on the device, for a number of reasons. So it's pretty amazing that you are running speech recognition and speeches and smartphones comp that. Why they extract knowledge from conference calls and one-on-one calls. The, for me, the killer feature of taxic is summarization. So people TR do transcription of their call and really f with taxic civilization. You get very high quality bullet points sum of your

[9:48] call. Again, this is something that was all impossible 10 years ago. Anyway, this is what you see. Incred well piles to write code, people use and create content, product like, and the marketing product teams use tools like tactics to create artifacts, meeting notes, brainstorming sessions, like that. Market research, ChatGPT ing PR tool to keep for what research. For any research, people use it as a storing P good, well, based on. So it's a great tool for buttons. Yeah, sounds like the acceleration, the adoption is accelerating. What do you think will happen as AI becomes more mainstream in

[10:52] product development? Yeah, so that's, this is something that all of us are trying to figure out. Yeah, my theory, but I have to preamble it. What will be talking about, that soon intelligence, because if it does, it's a totally different world we are into, solve world ofand, and but I personally think that to be from planning. So I think we'll see continuation of the existing trends, same but more B and more. Why? I suspect that me and probably you, I hit in a bubble. So these internal processes and tools that product teams are using, the, I'm not sure that this is the case

[11:55] everywhere. It's a standard practice here. It's pretty, we actually amazingly. But I think there will be a big change for in how we write code and how we develop software. Where is this joke, the hardest programming language today, English, and there is a lot of trust to it. With gpg and pilots, you code in English. A product manager who can think structurally and is good at communication can do a lot of work that used to require developers, which is very ening. It could be the time for humanity, humanity measures to sh from, I don't want to be stereotypical, but from my observation, people are

[12:59] better communic and people we will learn. This will probably result in smaller teams being super productive. Again, just be is a, it's a, this musical 10x developer will become 100x Cod developer. For experienced engineers, this will mean faster development cycles. This will mean much higher revenue for employee. It's not a common me, but I kind of like it. How much revenue your team that uses. Yeah, tubing is obviously changing. I think we are at the very beginning of unbundling of enterprise software. Today, a lot of yeahp whatever systems have a lot of okay functionality, and now startups come in, we have F,

[14:05] they fast and they develop much better solutions for vertical slices solutions back, and these are examples. Do you see any factors influencing the rate of adoption here to accelerate AI's mainstream adoption? Geographical and mental distance from Silicon Valley, greater Bay Area in China, Station 42 in Paris is probably the biggest. Got it, yeah, that makes a lot of sense. Yeah, that's where the builders are located. Yeah, I think we are at this stage when word of mouth and the podcasts that you listen to met how quickly again adoption myew. This is disruption comparable

[15:10] in, I think so. It's a similar scale disruption, but this time is much, much FAS in because the hardware is already in place. Right, people, right, right. And when this does gain mainstream adoption, what challenges do you foresee emerging for product developers and engineers? Yeah, so I think the main constraint is still the functionality of the FR models. Every major release of GPT is an extinction event for B if want fies, both people, startups, because functional subsumed by the model. But many tasks that people are trying to do are still challenging. The different is moving very fast. Handling this rate of scale is a challenge. Much of this

[16:12] functionality is moving to the edge from the cloud, and that's really beneficial for the product economics, so that's helping. Compute capacity is still a constraint, all the in Med CHS and general. And then access to data, this is becoming a limiting factor. Resolution of interaction for Biz issues that is fair and doesn't create extra friction is a big deal, industry, for the industry, I think. And as companies get smaller, Dmitry, as you were alluding to, that you don't need as many people as engineers upskill and increase their capacity, in your opinion, what should engineers, product developers, or even marketers for that matter, focus on to develop their skills and stay relevant

[17:14] in an AI driven landscape where companies just don't need as many people as they used to? Well, is the pral bicycle for the mind. You really want to know how to ride this bicycle. So from beneficial tech, if you're technical, not technical, this context, what is going on, that's AIG deal. Communication skills, right? We are now to the ey, right? We talking to Copilot, talking topt. Historically, it was not a for for many technical people. He, it's worth learning. There's a lot of excited prot prot J J. It's evolving with models. The skills that you develop in trting become solitary quickly, but it's still worth doing because it makes you produ now and you make it easier for you to learn the

[18:19] next termal, this spiral. Yeah, and of course it's worth keeping eye on these new tools as they come out. Now, the way basics don't, the basics for the product people don't change. Customer feedback is still all important. Trility to absorb it, any distribution is still cre. Ben of overestimating technology and underestimating distribution still there. Yeah, I think that's what I will be thinking about, some less on the skill side. It's more about things to worry about. AI is a human identifier, so it will, among other things, it will amplify by that, and it's responsibility of the product teams to prevent that, and it's easy. Well, AGI super intelligence that you develop is

[19:22] application V being used forious purpose. That's iability. Yeah, it really sounds like we're gliding into a brave new world. Dmitry, can you share more about your advisory services for startups and post Series A companies on international expansion, including getting global ready and selecting the next market? Yeah, this is the knowledge that I developed first at Evernote and then other VES, is help launch new market. Basically, I'm helping teams that have proven themselves in their domestic heart and want to increase their use of b f me launch new geographies. This involves everything from selecting new markets, getting all really internally, tooling, processes, loing, everything, changing your, to very people if you need to market,

[20:29] going with market strategy. And these are usually Prett involving products, yeah, transition out, and this is a pretty high AR activity for invest larger. Can you talk us through a little bit about your process? I know you have that great blog post on your Substack about exploring new market opportunities or defining product and positioning for many days. Do we have, but so the process always starts with company l s. They decide to revisit it. I think the right way to do it is to ask a question, well, are we ready and is it worth our

[21:33] focus? That stange, it's still super important for companies to prioritize. So I like getting engaged stage and helping company, which give you where they are and the size theying. There are several to, okay, how big is the opportunity, how hard is it, and how much, and I'm ready. And then, of course, they have to wait gu otheres. Well, next step is choose your. Okay, yeah, we want to do that, where do we start? In most cases, you don't want to start in zal markets at the very beginning. If you're new, the first time, it's much better focus on one market just to tune your processes. It's your learning exercise. Maybe in the future you'll bize

[22:40] it. You do research, you look, among other things, you look comp. The problem we soling doesn't even exist in the mark, doesn't. Okay, it does exist, so how are people solving this problem now? Are they using some approach or are there some? If you get a check mark on this one, hear yoursite in more and review your read. And becoming global really actually is very disrupting for the companies because it affects all company functions. Yeah, first thing the C to mind is optimization of user interface. It's more than that. So let's say that you're using AI. Hey, does your, does the model that you sitting to support the language your D? Yeah, and then finance. Finance needs to learn how to deal with

[23:45] foreign currency. Legal needs to learn how to operate in this new regulatory environment and jurisdiction. QA, they need to learn how to do legation. For analytics, most companies have do a pretty major refactoring project, adding new dimensions to the database, language, and so on and so on. So you all that, in parallel, you work on your go to market strategy. I advise people to start by reviewing their zero to one. How did they go to Z from zero to one in theing market? Is this approach applicable to new market? In most cases it is, but youment new realities. And in general, every market, they have to think about every market this back to zero to one stage with all the risks. Okay, you with, can you have your

[24:52] your rolling, you exec. One last question I have, Dmitry, is there a specific market that US based companies should enter after dominating the US market and evaluating where is it? Like South America, Europe, or is it really based on the use case? It's different for different countries. There is no single reci. Size of the market is super important. The biggest one is China, but for many companies, Chinese or tiets for different reasons. Sometimes is don't want to go, sometimes the competition. Yeah, America, it's now heav spere. Yeah, Spanish speaking Latin America is big, but it's not no. Europe has very high power, but again, this EU is. Japan is great market, but it has a

[25:59] very, very developed local system that not very about, and they have very high quality standards, so it's not get into single. South Korea, Bally, it's different from every. Well, thanks so much, Dmitry. Really appreciate the insights. Learn more about Dmitry Stavisky and his innovative approach to AI startup advice and international expansion. Visit his profile on intro. cdri steisy or follow his blog on world. subs.com. Thanks so much. Thank you. Follow.