Jon Krohn: 00:00 Pretty much every company has bought AI tools, but few of them are seeing a return. My guest today says the fix is a budget split that will make your CFO wince spend seven times more trading employees on the tools than on the tools themselves. Welcome to episode number 1019 of the Super Data Science Podcast. I’m your host, Jon Krohn. Today’s guest is Priyanka Vergadia, better known to her quarter million strong developer community as the cloud girl. Priyanka wrote the number one bestselling book, Visualizing Google Cloud. She led developer relations for North America at Google, drove enterprise go to market for GitHub Copilot at Microsoft, and now has gone all in on her own firm advising AI adoption. In this episode, Priyanka explains why AI raised the technical floor and made taste the new ceiling, how to structure clawed skills so your AI output stops being slop and her 10, 20, 70 rule for AI budgets.
00:57 Plus, she shares some breaking news you’ll hear on this show first. Enjoy. This episode of Super Data Science is made possible by Anthropic, Notion and Gurubi. Priyanka, welcome to the Super Data Science Podcast. Where are you calling in from today?
Priyanka V.: 01:12 I’m calling in from San Jose, California.
Jon Krohn: 01:15 Yes, yes. The heart of AI, the place to be. You must love living there. It seems like you’ve been there a while.
Priyanka V.: 01:22 Yes, since 2014.
Jon Krohn: 01:24 Our listeners are likely already familiar with you, Priyanka. You reach over a quarter million developers with your community, which is branded The Cloud Girl because of your cloud expertise, of course. You spent over a decade turning dense cloud and AI concepts into sketches and stories that developers actually remember. And so for example, something that you’re really well known for is a book called Visualizing Google Cloud. It was the number one bestseller. I can see on Amazon that it’s been reviewed a crazy number of times. And it refuses to dumb down even this complex Google Cloud ecosystem of 220 products. So it makes it easy for novices to hit the cloud, but it also has a lot for cloud fluent audiences. And you have a follow-up book, Visualizing Gen AI as well, which is obviously about visualizing another complex topic area. You’ve previously said that both of those books, Visualizing Google Cloud and Visualizing Gen AI, they emerged backwards.
02:23 Do you want to tell us about how those books came about?
Priyanka V.: 02:26 Such a great question. It’s because I’m a visual learner and I never intended to write a book because most technical books are full of text. And so if I’m not going to read something, I’m not going to be a promoter of writing something like that. So this is why I say it emerged backwards because I started putting some blogs out. This was during COVID. And these blogs had these images that I’ve made for some products in cloud early on in 2020, 2021, that timeframe. And they started to catch on, on GitHub. I have a repo on GitHub called GCP Sketch Notes. And it has thousands of stars. Very quickly it started to catch fire and attraction. And people came back to ask for PDFs. And I was getting a note every day on they’re preparing for certifications. They’re preparing to get promoted internally, externally. And it sounded like it’s making an impact that I didn’t quite fathom when I was writing and bringing these blogs to life.
03:41 And when PDFs were being requested, I was like, now I need to figure out how to turn this into a book instead of sending people all these PDFs or the versions that they have to download from GitHub. So that’s when I approached a few publishers. And honestly, I didn’t have to even sell the idea for the book because the blogs and the GitHub people was doing it all on its own. And Wiley had done some books with people who have done visuals before, they sent me a few samples of work that was done in the past and some inspirations and it quickly happened. And then the second book was really an inspiration from the first one where the idea of generative AI and the way it was evolving was so fast, but a lot of people were not able to join the party because everything was talked in research papers and the terms that was not accessible to normal learner.
04:42 And that was the idea behind that one where it’s like, okay, I’ve proved this idea once before. I think there’s a need for it. Let’s start writing it. And I collaborated with Lack on it because he’s done so many books before. And this time we did it with O’Reilly because they reached out and we were like, okay, let’s do it.
Jon Krohn: 05:00 So Lack is your co-author on that second book. You wrote the first one, Completely Solo, Visualizing Google Cloud was done all on your own, but yeah, you had a co-author on the second book. And both of those publishers, Wiley, outstanding. O’Reilly, of course, I think the publishing brand most associated with our industry. And so yeah, unsurprising that both books have done so well. Highly recommend people checking them out. I’ve been looking into them as I’ve been preparing to do this episode with you. And Google must have been happy that pretty much all the visualizations in visualizing Google Cloud are done in Google colors.
Priyanka V.: 05:36 Yes. And I think one of the other bits, which doesn’t quite. It gets hidden because the 220 products have 220 plus product managers. And so I had to go to every single one of them. And not that anybody asked me to, but it was more of like I wanted to build relationships with people who are building these products. So as I study them and represent their product on the sketch, they are part of the story. And so I made more than 220 plus friends and relationships as a part of it.
Jon Krohn: 06:15 That’s cool. How did you make the sketches? They’re really fun.
Priyanka V.: 06:18 Early on, a few of the sketches were done by me with my own hands in Adobe Illustrator. And then I couldn’t scale because I also have other things I do. So we hired a designer who could take the idea and the way I draft and design with the colors and everything. And then they could do it much, much faster than me because they are professional designers. Yeah.
Jon Krohn: 06:42 They’re full-time job. We’re going to get into the vast career that you’ve had, the kind of day jobs that you’ve had on top of writing books and creating all this content. But to dig a little bit more into the book and kind of your way of visual storytelling, something that seems to be a common thread in the way that you talk to people about technical concepts is that you seem to create a really recognizable bridge between their experience and your explanation. So whether you’re dealing with a long legal document, a kitchen blueprint or a librarian, your best analogies start from something that’s familiar to all of us. How do you find that familiar entry point when you’re explaining a technical concept? Maybe our listeners will learn something. Maybe I’ll learn something from that.
Priyanka V.: 07:26 Every time I’m looking to explain a concept, like I’ll give you an example that I was doing last week, going from prompt engineering to context engineering, to harness, to loop engineering and then now to graph. The world is crazy and the internet is calling all of these are like the previous ones dead. And I was trying to, in my mind, I’m saying that that is an evolution and it’s how we’re learning how to do systems engineering for AI. But how do I represent that when I attach the word evolution to it? So my first process is, how would I describe this in one word? And that in this case is evolution. And then I’m going back and saying, how would I represent evolution visually in something that I already know and see as a normal human being every day? And I went back to a seed, a plant, a seedling and then a plant and then a tree with the roots.
08:31 And then I got to the graph with all of the branches.
08:36 That’s like an exemplified way of explaining how I go about that process. I go with the word, then I go with the representation of that word in real life. And then that real life would take me there. Sometimes it’s as easy as every things are layers. And I would think about an onion that has layers. And then you peel the layers and then you explain each layer separately. There are many ways to do this. And these analogies are in everyday life. You just have to think through them just a little bit deeper to find that explanation that would fit. And at some point that analogy would break. And you have to be honest about that too. And be like, okay, this is like stretching too far, but I need you to understand that this is where we draw that line. So I do that as well if I’m writing longer, more detailed things.
Jon Krohn: 09:28 Nice. Lots of great tips in there. It’s interesting that you talked about having to have, with your first book, visualizing Google Cloud coming out in 2022, you of course would have had to have had a human illustrator at that time. And even your second book, Visualizing Gen AI, it was published late last year. And so probably most of the time you were writing that text to image models weren’t that great. It’s really been this year with NanoBanana. We can do lots of image adjustments. Yeah. So you were dependent on your own illustrations or a human contractor, but now we’re at this time. You’ve actually said that AI raised the technical floor because of generative AI. And of course that couldn’t be even more true for folks like a lot of our listeners who are hands-on AI data science engineering practitioners, where now it’d be kind of wild for you to be typing every character of your own code anymore.
10:25 As you’ve said that while AI raised the technical floor, something called taste became the ceiling. And it’s a difficult to define word. I don’t think I’ve tried to define taste on air. Can you tell us what that means?
Priyanka V.: 10:39 I think taste is abstract to define. It means you have played with the tools enough. You have a perspective of your own. This is just my definition. There’s so many out there floating. If I’m interviewing someone, I’m not looking for specific skills to use a tool anymore. I’m looking for can they think? Do they have a perspective on. And the perspective comes from the lived experiences that they’ve had in the past. If they’re a designer, the way they think, the way they come up with ideas, how many of them can they crank out? Because as we say, not say, but it is real. All of these models are trained on average internet data. So everything that comes out of them is at best average. The people who we want on our teams now are people whose ideas and lived experiences are worth tapping into. And they are unique and they’re different.
11:46 They have their own style. And that is what I define as taste, which is just super abstract. But the elements that you would look for in an interview in the past are so different now. And that’s how I define it.
Jon Krohn: 12:00 Yeah. So it seems to be across whether you’re producing images or video or code, this taste aspect that humans still seem to have some edge on relative to machines at the time of recording this episode anyway. It does seem like it gives us some hope that AI won’t be able to do everything that we can do just yet. So for me, when I see something that is very obviously GenAI created, especially now, Claude, OpenAI, ChatGPT, these tools have become very adept at creating business documents. So PowerPoint slides, Word docs, these kinds of things. But you can usually tell right away, as soon as you open it, that like, okay, this is GenAI created. And so a friend of mine recently sent me a pitch deck for his startup. It’s an AI startup. And he said, “I’d love for you to look over this and let me know what you think of it.” And as soon as I opened it up, I was like, “I’m not reading this because this is just…” I called it AI slop to him.
13:13 I said, “I opened up what you sent me. I’m not going to review it because I feel like I’m wasting my time because I don’t know how much time or though you really put into this. And if you’re asking for feedback on your business, I’m not going to put in my time and attention unless it’s obvious to me that you have.” And so this is all leading to a question for you because how can we as creators now make sure that it’s clear and creators of anything, of code, of an analysis, of an image, of a document? How can we make clear to somebody that we’re trying to get it across to that even though we may have used gen AI or AI in general for parts of what we’ve done, that there’s still a lot of thought, a lot of taste that has gone into this.
14:06 All this invisible learning behind the scenes that creates something novel. How can we express that effectively to executives or readers or whoever?
Priyanka V.: 14:16 There are these tools out there that are doing like, was this written by AI or not? I don’t know if I believe in that. This is where I think a much deeper, the question that you’ve asked goes way deeper. I want to be differentiated as a creator to the person who’s reading my stuff. There’s parts of it that are, and so there’s transparency and there’s parts of it that all of us are using AI to either flesh out, to do research. Well, I put on all my blogs that I take help for research for my blogs. And the lived experiences are added on top of those researches. And then I draft the bullets for my paragraphs. And then that goes into my clawed writing skill that actually does the blog itself. And so the output, when you put through these tools, it could look like that it was AI generated, but the experience in there, the example that I’m putting in there is all based on the conversation I’ve had with the CTO or a CXO.
15:25 And I think that is where I feel that people want to read my blogs because they know that what they’re getting is deeply ingrained into the experiences that we’ve had. So that’s one aspect of it. The second aspect, because I make a lot of images with, and my Gemini nano banana is my base model. And I’ve sort of fine tuned it with all of the 300 some images of my hand designed sketches to have my character and my style into the model now. So I have a tool that draws for me. The base model is Gemini and it’s fine tuned. Now that tool, if I just put initial prompt into it without much thought, it would generate slop. But if I put my storytelling thought process into it, which is super custom where I defined one of my criteria in our previous question on what is that word?
16:26 How do I map that word to the world? And then take that concept. So I have this whole skill around my storytelling concept. And then that skill runs first before it draws anything. And then it gets my approval on what it has come up with. And then I would say tweak this or that. And then it creates the sketch. The sketch that comes out has the CR on it, which means that we have these watermark tags and the model has created this, but it cannot create what it created without all of this taste that I’ve put into the skill. So the reason why everybody still wants these sketches to be downloaded, they know they’re AI generated. It’s very clear. But the material and the information on it, again, is off of those lived experiences, off of that storytelling framework, which is why they find it easy to learn a concept.
17:21 It’s a long answer to a question, but it is that twofold answer where you, who you are, what you bring from the experiences and what is unique about you that is bringing all of that. Everybody can teach you what AI is. There’s lots of information out there which will tell you what a token is. But why would somebody want to read it for me and my piece versus somebody else’s? And that is because of who I am and what I’ve done and the credibility that you have, all of those things. So what you need to do as creators, I feel to be differentiated is bring all of that credibility as front and center into your content and bring your personality into the content. Because there’s so many times that people have come to me and said, “I’ve seen so much content on YouTube. I paused on yours.” And I don’t even have many subscribers compared to the world out there.
18:22 I don’t think that I do. But it is because they connect at the level that is deeper than content. Just you’ve lived what I’ve lived and I want to be part of that. So it’s all about expressing that, I think.
Jon Krohn: 18:36 Yeah. I guess it’s hard to imagine that we will ever have. I don’t know. I could end up being wrong, but it’s hard for me to imagine that machine generated art, for example, will be hung in art galleries to the same extent as human art. Because what makes human art valuable is often the stories behind it, the people behind it. Van Gogh being a little bit zany, lopping his ear off. It makes headlines. A big part of why the Mona Lisa is so famous is because it was stolen from the loof. Yeah. These human aspects of things I think make content more interesting. And I think the advice that you gave, it applies not just to creating content for the public, making GitHub repos, writing a book, creating a blog post, but everything that you said I think is valuable to anybody who wants to create useful material internally at their work or what have you in this GenI era.
19:36 So beyond the content creation that you’ve done, you have had really impressive day jobs. So you’ve helped shape major AI launches at Google where you were for many years. And until recently, this is the hot off the press news. When I booked you to have you on the show of last week, you were still on LinkedIn at least. It showed that you were at Microsoft, but now you are full-time on the Cloud Girl. The Cloud Girl is not just a brand, it’s not just your content creation brand, but that is also a consulting business. So I have some follow up questions for you around your commercial experience, but can you walk us through kind of the key steps, what you were doing at Google while you were there, the Microsoft journey, which I believe had a lot to do with GitHub Copilot, if I’m not mistaken, which is something that is really important to a lot of our listeners.
20:31 And then yeah, after you’ve done that, I’d love to hear why you decided now is the right time to go out on your own with the Cloud Girl business and tell us about what the business does.
Priyanka V.: 20:41 Yeah. Okay. So I’ll start a little bit more back before Google because – Sure.
20:47 So 15 years of my experience in the tech world, first few startups, and then Akamai, and then Google. And the reason I mentioned that is because most of the time before Google was working directly with customers. I was in roles like solutions architect, technical consultant, frontline with the sales teams, but owned the technical business decision. And then when I got to Google, it was similar role, but then I explored and found this new world of developer relations, which was instead of going one-to-one to a customer and engaging with them and helping them on their business problem, architecting their solution with cloud, now you’re getting to do that one to many. And that happened through a 20% project that I got interested in making some videos and the rest is history. And I was like, “This is actually really cool. I’m great at convincing people on this is what you should use for the problem that you have.
21:47 I’m also great at building demos and prototypes and walking through the art of the possible with a specific technology.” So DevRel world started to look appealing. I interviewed, moved into that. So then comes the developer relations experience from 19, 2019 to 2024. Had a blast. And
Jon Krohn: 22:11 That was for our listeners in case this wasn’t obvious. All of that was at Google. So you’d been doing this sales engineering solutions engineering at Intel, at Akamai Technologies, as you were saying, in the San Francisco Bay Area. And you continued to do that at Google for almost two years as a customer engineer. So when you were saying interviewing, you were internally interviewing for this developer relation role.
Priyanka V.: 22:34 Exactly. So internally, you have to interview internally to move into other teams. And that’s kind of what happened. And so I worked with customers, with partners, but customers for a year, partners for another year. And then I moved into developer relations by internal movement. And that led me to exploring this world of developer relations. Now in that, I’m not just working with customers. There’s one aspect of engaging with developers, building a community, which was translating exactly from the technical consultant solutions architect world. But there’s these additional aspects of when a product or a new feature comes in, how are we the customer zero? When it goes to the customer, what are they going to feel as an experience? Are they going to like this? What is the friction that they will face? And then interface with engineering and product to make sure that those kinks are taken care of while we also figure out, I led the entire content team for a few years and content meant everything that went on Google Cloud tech YouTube channel, the blogs for developers, the training material that we would put out for specific products or a set of products or use cases that went on Coursera and external platforms, also internal platforms.
23:58 And all of that is in service of the developer and making sure that their life is easy when they touch this product. And also when we launched Gemini Code Assist, the goal was turned into this whole idea of storytelling. When you launch a product, and this has become so much more important now that with AI, you can literally just create products, create features. People are shipping every day. Are your customers able to digest that level of information with the same frequency that you are pushing them out? No, because you need to tell a story around what this product does or feature does for them. How does it increase revenue for them or solve X problem for them? And that is what is the skill that I feel more people are able to do if they have seen different parts of the business and have not been in only one area or have had lateral movements so that they can see that part of the story being told.
25:05 So anyway, so I got to do a lot of that for Vertex AI when it was launched. Gemini Code Assist when it was launched. Tell
Jon Krohn: 25:13 Us more about Vertex AI.
Priyanka V.: 25:14 Yeah. Vertex AI is Google’s platform for everything AI. So you can get a custom model, you can get open source model and deploy it in cloud. You can fine tune that model and deploy it. And now you have your own API endpoint and you can manage all of that infrastructure. You can deploy your own GPUs with it. They have got a lot of custom stuff in there as well. So it’s easy. So you don’t have to write code for a lot of these things and you can just throw in a few examples and fine tune a model very easily. So that is the platform that was launched in 2021, 2022, that timeframe. And now it’s being used for all of these things to get all these updated models in there. So that’s Vertex. Gemini Code Assist was this first, very first launch for AI assisted coding.
26:10 GitHub Copilot was already in the market before then. And this was Google’s first attempt to go out with an AI coding agent in the market. And so it was a very fun time because it was getting to write the story for the first few demos that we would go do and get the team to work on the first few hackathons that we are going to launch and get feedback from the developers. And then from then, one of my mentors reached out at Microsoft and cursor was hitting very hard at the time. They’d just done this relationship with OpenAI. So the idea was you come in, figure out how we want to do go-to market for GitHub Copilot in this heavily competitive situation. And it felt very interesting because I had done. And go-to-market meaning it sat between the thought leadership, which I was doing at Google from the developer relations perspective, but also brought me back to, we got to go talk to the CTOs, the CXOs of the Fortune 100s of the Microsoft customers, which was a very good opportunity to sort of combine those two pieces together.
27:24 And a lot of this, I believe, with AI has nothing to do with products. Everything is to do with trust, relationships. Can you get somebody? And it was never to do with products, honestly. But with AI now you can actually clearly see that it is all about relationships. It is all about who do you trust the most? Can they be with me on the journey to get me through this transformation? And whichever company is able to put the right people in front, like the FDE roles that we are talking about and which are becoming so hard right now, that’s all the idea is. You’re bringing your best engineers and the best communicators, putting them out there to be talking to the biggest stakeholders. And so I ended up building, before this term existed for deployed engineering, I ended up building a team of outcomes-based engineers for GitHub Copilot, for Microsoft in that intersection.
28:27 And we were talking to Fortune 100 CTOs to help them do AI transformation for their software development life cycles.
Jon Krohn: 28:35 Yeah. For Deployed Engineers, definitely a hot job title these days for our listeners to consider. Yeah, if you’re looking to kind of be moving laterally into something that has a lot of growth potential. FDE for Deployed Engineer is something to think about, which as Priyanka described is going and working alongside the client often at their office and making sure that your product like GitHub Copilot is being effective there. It’s interesting how you said while you were at Google and Google was launching, was it called Gemini Code Assist was the name of the part? Yeah. And so that Code Assist product, which I guess now kind of evolved into the Gemini CLI.
Priyanka V.: 29:14 CLI and then anti-gravity and then it’s like
Jon Krohn: 29:17 A few
Priyanka V.: 29:17 Other branches. Yeah. So
Jon Krohn: 29:19 Many product managers, so many products. So it’s interesting how you said when that product was coming out and you were involved in that, how the existing incumbent in the space was GitHub Copilot, whom you then later joined and you made a big impact in them still. I guess there’s a big market, but you still played such a big role in GitHub Copilot rolling out to enterprises. There’s something that you said earlier, you were talking about how quick it is now to be just rolling out features, rolling out products, shipping every day, how quick things change and how important it is to have these kinds of roles like your senior director of AI Transformation role at Microsoft or your head of North America developer relations role at Google, how critical it is to have those roles to be able to communicate changes because I run into constantly now abrupt changes in my product experience pretty much every day I see the little.
30:15 There’s this friendly little leaf in the bottom left corner of my Claude Mac OS that says update, ready to reinstall, click here. And I’m like, okay. And then things happen where a couple weeks ago in the Claude MacOS experience, you could easily navigate in the top left hand corner between the chatbot experience, cowork and cloud code. And then now somehow they’ve just combined cowork and chat I guess into one panel and now I can only click between code. And obviously there’s reasons why they do that and they think it’s ultimately going to be better for me, but it’s frustrating to have those changes happen when you’re used to something working some way. It seemed to work really well and you’re like, how come I can’t find cowork? It’s gone.
31:02 I’ve been on a long tangent here, but there’s a specific thing now that we’re back on Claude. Way back in the content creation section that we were in like 15, 20 minutes ago, you were talking about how a big part of your success with using GenAI tools with Claude specifically is having skills set up. And so that’s something that I meant to at that time, we ended up going off and talking about something else, but I’d love to come back because I feel like that’s something really practical and technical and useful for our listeners. Could you explain for our listeners who don’t already know about skills, what those are and how they can make the most of them in Claude?
Priyanka V.: 31:39 Skill is something that you would define your task to be. So break down your task into small sub-tasks and you define how you do that. When I write the blog, I do a research. This is how I would do it as a human, right? So think about this. When you’re writing a skill, think about it like a human. Now, the way Claude helps you do it is amazing, but before you even get into it and start setting up a skill, think about your task explicitly. I’ll give you an example because it’ll be more material that way. I’ll do a research first if I’m about to write a blog. And on this topic I would. And the research prompt has to be really good as well where I want only high quality content written by researchers and scientific research from schools and universities and organizations like this.
32:40 Only look for that stuff around this topic and then create a report. Let’s say that is the prompt, but that becomes part of my skill as step one. Then the next part of that skill is after I do the research. So the prompt is the how, right? So I’ve put the how in there as well. Research is the task. How you do it is that prompt. Then the next step would be to synthesize that research. I usually do a human in the loop thing in there. I don’t trust it to make decisions beyond that. So I would do a human in the loop. It sends me a text message. I’ve got all this set up in Hermes. So it sends me a text message saying, I’ve done the research, here’s the doc. I like to read because I’m trying to do that intentionally so that I don’t lose the skill with AI.
33:33 But you can also, I’ve also done cases where it would send the audio to me and I can just hear what the research was while I’m running or while I’m on a workout, which is super handy. And then I would give it instructions on, okay, I want to change this or that. And then the next step is like kick off writing the first outline.
33:59 Most people would just go in, write a blog on this topic. You’re going to get slop. The whole idea here is how would you approach a blog? I approach with research, then I would go in and do my analogy and storytelling on top of it. That’s my next step. Then I would synthesize after the synthesis, the storytelling, and then putting it into a format that I usually used to write blogs when I was doing it all on my own, which is I need to have three images in this blog and they need to be developer focused, which usually talk about the flow of movement of tokens or query. And I have some of these examples in there. And I need to have one practical example in there. And that can come from, there’s prompts in there where Claude would ask me of a practical example from my experiences.
34:56 And so that it can take those and do that. So I’m going into too much detail, but the idea of a skill is how do you do the task? Define it into sub-task. Those are your bullets that go into the skill and also some example prompts that go in there. That way you will get a much more personalized, the type of outcome you would. It’s never perfect, but at least 80% there and now you can start editing it from
Jon Krohn: 35:26 There. Perfect. Yeah, that did have a lot of examples, a lot of detail, but hopefully it helps us understand just like a lot of your cartoons, your illustrations go into practical examples. And so we got lots of examples there of how to build effective skills in Claude. So thank you for indulging us with that. Back to the enterprise stuff with your Google Cloud experience, your GitHub Copilot experience at Microsoft. With your work bridging the gap between high level boardroom strategy and real world enterprise AI execution, you have something that you call the 10 – 20-70 framework to help guide budgets for AI success. Could you tell us about that 10, 20, 70 framework?
Priyanka V.: 36:11 So 10% on tools, 20% on execution with those tools, and 70% on education and skilling and upskilling. And I know this sounds crazy, but I have worked with enterprises that have large number of large teams and have bought the tools and don’t see ROI. This is exactly the Stanford report that just came out, the impact report is a great example. It has like eight or 9% of the actual AI use cases in production, in real production use cases are about eight to 9%. Everything else is just like experimentation. And this is exactly the reason you’re not going to see ROI, you’re not going to see real use cases that are leading to revenue or cost reduction or savings because you’ve bought the tools. AI is a habit.
37:18 And habits don’t form in days. They form in an extended period of time. So this whole concept of token maxing and all of these things are just natural evolutions as well. Yes, the idea of token maxing is super weird because we went from, okay, if you’ve got a tool, now the AI officer in the company is like, nobody’s using this. We got to make them use this. So you get into this whole problem of now everybody’s using it for writing emails or like the dumbest tasks. And now you’re in this token maxing situation, which you never though of where it’s like, oh my God, now we are spending so much money on this tool and we are not seeing ROI. So you got them to use it, not effectively, and you’re in a different problem. But I think it’s all a good problem because you at least got them to touch it.
38:18 When you look at 10, 20, 17 year old, if you spend that 70% of your budget and time on upskilling your employees, which means showing them effective ways of using the tool, not just telling them use it. Showing them effective ways of using the tool, not just giving them training, but actually giving them real use cases of the thing they can do in their job. And that requires time and effort and energy. My DevRel hat on, that requires building a community and saying, I tried this thing today. Let me share it with you all. If you’re a testing team, if you’re a coding team, if you’re a team that’s product managers, got to share those experiences with each other. And you have to bring space for that as a leadership team to allow people to build and form these communities. And after this is a big J-curve, right?
39:19 And after six, eight months, you start to see effective use of tools actually helping them be productive. And then you get to a point where it’s like, now we can write test cases with this. That’s looking really good. How do I write them faster? Improve my prompts a little bit more. How do I take an entire process and make that an agent? Now you have agents in each of the different business units and you can form that into a repository of agents and now an entire company is becoming efficient. But this is a trajectory and the curve, you have to see the vision for a year or so and pour into it, which is why I say if you spend 70% on some of this stuff, which is going to be very costly and hard for a CFO to agree to, but that’s the only way to build a habit and an effective habit.
Jon Krohn: 40:17 I recently read in an economist article about how it was basically the economist trying to back into what kinds of money enterprises across the world would need to be spending on tools, on Cloudcode, on Google Gemini, in order to be able to make good on the hundreds of billions of dollars that are being spent on AI data centers in 2026. And it was that same article brought up, I don’t know if they were citing you or citing Stanford Research or what, but they brought up how, okay, if hypothetically organizations were going to spend this much on the tools, they would need to spend an order of magnitude more. So similar to your seven to one ratio on transformation, on getting people educated in order for any of this to be effective, for any of the tool purchases to be effective. And so that all of a sudden when you’re like, whoa, when we’re nowhere near even that 10% being spent in order to realize the AI data centers are being worth it, it starts to make it seem like, I guess we’re going to see what happens.
41:35 It depends on some of these players are banking on super intelligence or something to make good on these huge AI data center investments. Yeah. Anyway, that’s kind of a side story, but that kind of economic story ties into the same kind of benchmark that you’re talking about here.
Priyanka V.: 41:54 Yeah, absolutely. And it’s hard to buy into, right? Because you’re traditionally used to thinking as organizations like, what is going to be my spend this quarter, this next quarter? And then just return on that spend is what you are looking at. But this is a longer term play, which is hard to digest and understand. And it’s easy for people like me to understand because we’ve lived in the developer relations world where building a community does not have near term benefits. It has long-term benefits. You build a community of developers that are working on open source for something. That’s yours when you start to see actual benefits of that. When Kubernetes came out, it was a large amount of effort for large period of time before Kubernetes became the thing that people wanted to use. So I cite this because it’s hard for people in economics and also CFOs to kind of wrap their head around this.
Jon Krohn: 42:58 But yeah, to be getting an ROI on these tools, we’ve got to be spending a lot on organizational transformation on education for sure. Speaking of which, half an hour ago, maybe started asking you to talk about your journey from Google to Microsoft to what you’re doing now with a Cloud Girl. And so you’ve just in recent weeks seemingly decided that it’s the right time to jump off on your own and be doing the Cloud Girl. Why is that? And does it relate to us being in this unique time in history, maybe in this unique time in history where we need to be spending lots of money on training people?
Priyanka V.: 43:38 Yes. Oh my God. I have been always passionate about two things which drive me every day. And all the roles that I’ve had kind of fed all of that. So I’m very, very thankful to have all these roles at these amazing companies. I love learning new things and I love teaching those things. And that’s my cycle of like going from learning to teaching to solidify my understanding and then keep doing that. And I’ve been doing that for years as CloudGirl on content on my YouTube, Instagram and LinkedIn and X. But in the last six, eight months, it just seemed like the opportunity for what I could be doing with that learning and teaching passion for companies or startups to do product storytelling. This whole idea of we can ship so much faster, but we don’t know how to tell an enterprise why we are important and why you should be using us, that product storytelling muscle for startups and enterprises.
44:50 And then the second, which is career storytelling muscle for people who are moving into FDE roles or want to make those moves, becoming AI engineers and don’t know how to position themselves because the resumes are all looking same now. And once you are past that resume state, how do you stand out? Which is all about storytelling. You’ve got great stories. How do you represent them and you in that process so that they end up picking you? So there’s all these, we’re always storytelling by the way, but they are now becoming so much more important because everybody else is sounding very similar to you. Products that you’re building are very similar to the product somebody else is building. So product storytelling will differentiate you in the market. Career storytelling will differentiate you as an individual, as a professional, so that you can grow faster. And then my third thing that I’ve been passionate about is content.
45:47 So learning and teaching and growing that community of CloudGirl. And so that is what I am focused on. It gives me so much energy to be helping a few startups and companies bringing their products to life, helping a few individuals get their careers into places where they want to be. It just brings me joy to do that. So I want to focus on that and I think it’s perfect time to do that.
Jon Krohn: 46:14 I heard a rumor that you might be working on a next book. Are you able to tell us anything about that given how popular your first books have been?
Priyanka V.: 46:21 Oh my God. Yes. So I’m working on storytelling book, tech storytelling. So this is in the works. I still haven’t decided what outlet it’s going to be, but to those first two points, like product and career storytelling, I think in tech there’s a huge need for this. And I do fill that gap in a lot of ways. And I want to bring more of that knowledge to the world.
Jon Krohn: 46:49 Fantastic. Well, your huge following must be dying for another book release. So that is exciting news indeed. Maybe they’re even hearing it here first, which is exciting for us here at the show. I
Priyanka V.: 47:03 Don’t think I’ve shared it anywhere, so this will be the first.
Jon Krohn: 47:06 There you go. Breaking news. Yeah. We’ll be looking forward to that for sure. Given all of your experience, such rich experience at organizations that are at the cutting edge of AI deployment, where do you see things going in terms of, especially now that with the Cloud Girl business, you are advising individuals and enterprises on what they should be doing with their careers, with their companies. What guidance do you have for my audience in terms of how as individuals we can be positioning ourselves yet best for the decades to come? And then as a second question, how can organizations get themselves best set up for the decades to come?
Priyanka V.: 47:52 We talked a little bit about taste. And to me, this is synonymous in a lot of ways. So storytelling or being unique or being differentiated. The only way to do that is finding your USB. Whether you’re a business or an individual, you have something unique that you bring to the table. You just haven’t spent the time to either figure that out or it got lost in the journey. So you have to do that homework. So the workshops that I’ve done in the past for this exercise is sit down for a weekend. And I have some questions on my. We can put some in show notes if there’s an option to put some links for this. I have a list of questions that I would ask myself. And these are not easy questions to answer. These are questions like how do I want to change the world?
48:49 What wakes me up every day? And this may take you some time to actually get to or figure out, or you might have a business today as a product. What does my product solve for? And it’s not the features and the functions. It’s the core of like, when would someone be dying to give you money for this? Or when you’re thinking about the product. But when you’re thinking about your career, why would someone hire you? Not because of the skills, because everybody’s got now because of AI. Anybody can write code. That’s not a skill I’m interviewing for or anybody’s interviewing for. What is unique about me? What do I bring to the table?
Jon Krohn: 49:33 Well, thanks for sharing those invaluable tips on what we can be doing as individuals and organizations to prepare for the years to come. I think there’s a lot of anxiety when you see things like you mentioned there how nobody’s interviewing for coding skills anymore. And that’s something that a lot of us, myself, a lot of our listeners have spent years, decades developing that skill. And now a clawed skill is doing it. Yeah. So it’s nice to hear your perspective that it seems like because of things like taste, because of these big questions, direction that people can give, that we can still provide a lot of value, hopefully for years and years to come. And we will have those questions that you talked about for our listeners in the show notes. Thank you, Priyanka. Before I let you go, I ask all of my guests the same final two questions.
50:31 The first one is, do you have a book recommendation for us?
Priyanka V.: 50:34 I’ve been reading this book, which has nothing to do with tech, by Michael Singer, Untethered Soul. And I’ve read this book twice in the last two years. And this is the first book that has ever. I picked it up reluctantly by recommendation from a friend and has been the first ever book that I’ve read on mindfulness that I actually resonated with. So if somebody out there is like, “I’m intrigued. I want to know a little bit more.” I think this book just approaches it so scientifically that my engineer brain was like, “Okay, I can be with you through this to understand what mindfulness is all about.” So if you’re skeptical like me, I think this is something that I would recommend. It opened up my mind to it.
Jon Krohn: 51:31 It’s a cool recommendation. I will have a link to it in the show notes. And it is interesting because the people that gave the testimonials for the book, just at a glance here that I’m looking at, it’s a pretty wild mix of people. You’ve got a rabbi, you’ve got a priest. It sounds like a beginning of a joke. You’ve got a couple of yogis and you’ve got Ray Kurtzweil, the big AI thinker and author of The Singularity is Near and The Singularity is Nearer. And so quite an interesting mix of people walk into a bar and find mindfulness. Yes, exactly.
Priyanka V.: 52:08 It’s a very interesting take. And it does draw people especially from zeros and ones type of backgrounds like ours to welcome into that world, which if anybody wants to get into it, I think this is a good start.
Jon Krohn: 52:26 Cool. Thank you for the recommendation. And it’s sold millions and millions of copies. So it is a different world from selling techbooks, isn’t it, Priyanka, when you see the kinds of numbers that a trade publication.
Priyanka V.: 52:38 It’s not thousands. It’s millions.
Jon Krohn: 52:40 Exactly. But yeah, this is our. I don’t know if you knew this, but I wrote a bestselling book just before the pandemic called Deep Learning Illustrated, which was kind of a similar vein to your visualizing books.
Priyanka V.: 52:54 I checked it out. I am going to get a copy. I checked it out. I went to the website. I was like, “This is very similar to how I think as well. So I’m going to have to check it out.”
Jon Krohn: 53:07 Yeah. Looking forward to hearing what you think. Anyway, before I let you go, the very final question that we ask is how can people be following you? Of course, you have over a hundred thousand followers on LinkedIn, so that seems like a place. Where else should people be following you?
Priyanka V.: 53:21 In all of the platforms on X and LinkedIn and Instagram and Substack or my website, it’s everywhere the cloud goes. So if you just look for that and follow in all of these places, also on YouTube, that’s where you can find me. On YouTube, I do long form content. I actually do explainers, teach stuff. And on Instagram and stuff, I do roadmaps and what to do to build your career in super fast formats. But yeah, the CloudGo should take you in all of those platforms.
Jon Krohn: 53:55 And you recently started a podcast too, right?
Priyanka V.: 53:58 I did. I just started a podcast. If you go to my website, you’ll see the episodes there. And yeah, if you want to be on the podcast, reach out on any of these platforms and we can have a chat.
Jon Krohn: 54:11 Fantastic. Quite an opportunity there to get in early with somebody with such an enormous following and respect in this industry. Priyanka, thank you so much for taking the time out of your busy day to talk to us. Really appreciate it and hope to have you on the show again in the future.
Priyanka V.: 54:26 Thank you so much for having me.
Jon Krohn: 54:28 Great episode today with Priyanka Vergadia. In it, she covered her 10 – 20-70 framework for AI budgets, 10% spent on tools, 20% on execution, and 70% on education and upskilling. She explained why AI adoption follows a J-curve. It takes six to eight months of showing people real use cases in their actual jobs, not generic training before productivity gains appear. Then teams graduate from better prompts to full agents across business units, full agents across business units. She talked about how she builds skills in Claude by breaking a task like blog writing into explicit sub-tasks and how AI raised the technical floor while taste became the ceiling. So she now interviews for perspective and lived experience rather than tool skills like coding. As always, you can get all the show notes, including the transcript for this episode, the video recording, any materials mentioned on the show, the URLs for Priyanka’s social media profiles, as well as my own at superdatascience.com/1019.
55:31 Yes, that is the episode number 1019. Thanks to everyone on the SuperdataScience podcast team, our podcast manager, Sonja Brajovic, media editor, Mario Pombo, our partnerships team Natalie Ziajski, our researcher, Serg Masís and our founder Kirill Eremenko. Thanks to all of them for producing another excellent episode for us today for Naval, that excellent team. To create this free podcast for you, we are deeply grateful to our sponsors. You, yes, you can support this show by checking out our sponsor’s links, which are in the show notes. And if you’re ever interested in sponsoring an episode yourself, you can find out how at jonkhrohn.com/podcast. Otherwise, please do help us out by sharing this podcast with folks that would love to hear from Priyanka’s Cloud Girl Insights. Review the podcast on whatever podcasting platform you use or on YouTube. Subscribe if you’re not already a subscriber, but most importantly, I hope you’ll just keep on tuning in.
56:29 I’m so grateful to have you listening, and I hope I can continue to make episodes you love for years and years to come. Till next time, keep on rocking it out there, and I’m looking forward to enjoying another round of the SuperDataScience Podcast with you very soon.
