Productside Stories
The Great Compression: Will AI Replace Product Managers?
Featured Guest:
Summary
In this episode of Productside Stories, Rina Alexin talks with Karl Abbott, Senior Ecosystem Product Manager at Microsoft, who spent 16 years at Red Hat across technical support, training, technical account management, and product. Karl describes how he fell into product management, how his technical background lets him bridge the technical and business worlds, and how he built CAM, Karl’s Agenda Builder, and an offline installer script that made it into the product within a week.
They then dig into what AI changes: how far a non-technical PM can go with vibe coding, a three-step way to start, and Karl’s rule that the moment someone else depends on a vibe-coded tool, an engineer needs to review it. They discuss why AI output that is confidently wrong deserves a second look, why cutting the junior pipeline is a bad idea, and why the prediction that product managers disappear by 2030 mistakes compression for extinction. Karl closes with career advice for the question, “Where do you want to be in 10 years?”
Key Takeaways
A non-technical PM can get far with AI, but not all the way.
- AI can take you from a problem to a working solution. The risk is reaching a point where it works but you don’t understand why, which becomes a problem the moment it breaks.
When someone else depends on it, get an engineer.
- Personal tools and prototypes are fine. Anything going into production, or relied on by other people, needs engineering review. Karl runs a handful of vibe-coded tools for his own work, and hands anything others depend on to engineering.
AI output that sounds right still needs verification.
- AI invented a daughter in a blog post Karl asked it to help write. He trusts AI most where he can verify the result, such as a vibe-coded prototype he can run, and least where he would need deep research or an expert to check it.
The PM role is compressing, not disappearing.
- Karl expects the product manager title to still exist in 2030, with more design and engineering fluency required. One person can ship a first version, but supporting and growing a product at scale takes specialists, and cutting the junior pipeline today erodes the judgment teams need tomorrow.
Chapters
- 00:00:00 –Introduction and Transition into Product Management
- 00:00:53 – From Red Hat Support to Product Manager: Karl Abbott’s Path
- 00:04:06 – Why Technical PMs Can Spot a Business Problem Hiding in a Technical Error
- 00:05:27 – Evaluating Coding Capabilities
- 00:06:27 – Building as a PM: The Agenda Builder and the Installer Shipped in a Week
- 00:09:15 – Navigating Engineer Boundaries
- 00:10:10 – Building an Offline Installer Script
- 00:14:08 – How Far Can a Non-Technical Product Manager Go With AI?
- 00:15:42 – Three Steps to Start Vibe Coding Without a Coding Background
- 00:17:45 – The Line: When Someone Else Depends on It, Get an Engineer
- 00:19:25 – Vibe Coding vs. Vibe Prototyping
- 00:19:57 – Confidently Wrong: Verifying AI Output as a PM
- 00:23:00 – The Junior Talent Pipeline Problem
- 00:27:39 – Will Product Managers Exist in 2030? The Product Builder Debate
- 00:30:20 – Why One Person Can’t Scale a Product Alone
- 00:32:11 – “Where Do You Want to Be in 10 Years?” A Better Answer
- 00:35:03 – Closing Remarks and Where to Find Karl
Why Listen to This Episode?
In this episode, you’ll gain:
- A practical starting point for vibe coding when you don’t have a coding background
- A clear rule for deciding when a prototype needs engineering review
- A simple test for when to trust AI output and when to go find an expert
- A better answer to “Where do you see yourself in 10 years?” in a market that refuses to sit still
If you’ve ever shipped something you couldn’t fully explain, this episode is for you.
Introduction and Transition into Product Management
Rina Alexin | 00:00:00 – 00:00:50
Hi everyone, and welcome to Productside Stories, the podcast where we dig into the very real and raw lessons learned from product leaders and thinkers all over the world. I’m your host, Rina Alexin, CEO of Productside, and today I’m talking with Karl Abbott, a senior ecosystem product manager at Microsoft, working on Linux on Azure, with 16 years at Red Hat behind him. Across a lot of roles: technical support, training, product. And he’s also the host of his own podcast, Productly Speaking. We’ll be talking about the great compression of product, design, and engineering, and whether AI-powered prototyping actually makes non-technical product managers more capable, and what is going to happen to product engineering overall. It’s a pretty big episode. So welcome, Karl. Excited to talk to you.
Karl Abbott | 00:00:50
Yeah, thank you for having me, Rina. I’m excited for the conversation.
From Red Hat Support to Product Manager: Karl Abbott’s Path
Rina Alexin | 00:00:53
So, Karl, you’ve taken on a number of different roles at various companies. How did you end up in product management?
Karl Abbott | 00:01:00 – 00:02:50
Yeah. As I’ve talked to a bunch of different people who have also been product managers, there’s a common thread, and I fall into that common thread. I tripped and fell into it. I didn’t expect to become a product manager. I started in frontline technical support at Red Hat all the way back in 2005 and had a long career in support and training, and then a job called support readiness, and then technical account management, where I was the key technical point of contact for a handful of large accounts. I was sitting around as a principal TAM, pretty happy with the role, and Red Hat needed a product manager for the RHEL kernel, which I had no idea they needed. Somebody else on the product management team was talking to somebody that I had mentored a few years before in support, and they dropped my name. So there you go, little things, paying it forward, right? They dropped my name as a possibility, and the product management team reached out to me saying, “Hey, we’ve got this role. Would you be interested? We’d like to talk to you.”
I took the conversation thinking that this was going to be a quick no. It’s a big role, product manager for the RHEL kernel. This is huge. It was a lot more than I felt ready to take on or interested in taking on. “Thanks, but no thanks.” That’s what I thought I was going to be saying at the end of the call. But the more I talked to them, the more interested I got in the role, and then I talked to the hiring manager, and the more interested I got. Long story short, I interviewed, got the job, and now all of a sudden I’m a product manager. I have found that product management has literally brought everything that I’d learned prior to that, and keep learning as well, together into one role. So it’s been really exciting. I find that 80% of product managers, and maybe it’s more than that, have been anointed product managers at some point in their career.
Rina Alexin | 00:02:50 – 00:03:20
Yeah. It comes and finds you. It finds you like a cat, right? It’s like, now you’re a product manager. But I think it’s also a testament to the kind of role that it is, that so many different people from various backgrounds can really make it a career. So I’m going to ask this of you: maybe put yourself in the shoes of the person that threw your name in the ring. Why do you think they suggested you would be good at it?
Karl Abbott | 00:03:20 – 00:03:40
I had a lot of technical background. I’d been known for doing that. I’d been a team lead and helped mentor a lot of different people, and I’d had the opportunity to work with a lot of large clients in a lot of high-stress situations and learn how they were actually taking our software and using it to solve real problems. So I think all of that together was, “We should talk to that guy.”
Rina Alexin | 00:03:40 – 00:04:06
So you understood the big picture, you understood client needs, and then you also had that technical foundation. I get that. It gives you almost two sides of the coin that you need to be good at to be a great product manager. So you come from a technical background. Not everybody does. In fact, many product managers who get anointed come from marketing, operational, even support roles, sales roles. And that’s part of the conversation today.
Bridging Business and Technical Needs
Why Technical PMs Can Spot a Business Problem Hiding in a Technical Error
Rina Alexin | 00:04:06
How does that technical background really help? How do you leverage that in your current role?
Karl Abbott | 00:04:10 – 00:05:27
Yeah. The technical background helps quite a bit when you’re able to understand technical systems and how they work together. But if you can’t then understand the business side of it, you’re at a disadvantage. So being able to bring business knowledge and technical knowledge together, to understand how the technical systems connect to the business systems, is a huge advantage that I have as an ecosystem product manager. I can look out at a technical system and see all the little intricacies and how the parts fit together, and then turn around and have a conversation with finance people and business rules people.
Just this morning I was on a call, and it was: no, this is not a technical problem, it’s a business problem masquerading as a technical problem. You’re getting an error from this technical interface, but it’s really a business rules thing. You’re just in violation of this business rule. We’ve got to figure out what we can do about that, and then we can work toward a solution. It was one of those moments where, for that customer, at that time, they said, “Oh, we had no idea that this was a business issue. We thought it was some technical issue that you could solve quick.” And it’s like, “No, it’s more complicated.” So being able to bridge those worlds is very, very foundational to the role of product.
“This is not a technical problem. It’s a business problem masquerading as a technical problem.”
Evaluating Coding Capabilities
Rina Alexin | 00:05:27 – 00:05:45
And I think it’s becoming even more so. We hear it all the time that product managers are expected to be builders. So I can imagine there’s definitely some value from being able to prototype, and actually look at the underlying code yourself and judge whether it’s even good. Grade yourself on that ability. How good are you at doing that?
Karl Abbott | 00:05:45 – 00:06:27
I’m at least good enough to prototype things and to build code. But I don’t know if engineers would say, “Oh, your code is great. You wrote it to our standards. We love what you’ve done. You kept it really nice and succinct.” My standard has always been to get it to work, and then kind of whittle away at it and work on it. I’ve never followed a coding convention. I do have a computer science degree, and I’ve done a lot of coding, but I’ve always been in and around the code and never in that engineering role directly. So I’ve had to use code to understand code, to see what people have done and see how these things come together, but I’m not necessarily somebody who’s had to generate code to get the job done.
Building as a PM: The Agenda Builder and the Installer Shipped in a Week
Rina Alexin | 00:06:27
But have you built something yourself?
Karl Abbott | 00:06:30 – 00:09:15
Yes, I have. In fact, I built what became a big part of how Red Hat technical account managers do part of their job. I built a piece of software called CAM, Karl’s Agenda Builder. I built it a long time ago now. I was trying to think about how old it was just the other day, because I was talking to somebody at Red Hat who said it’s actually still in existence. It’s still being used and still being maintained, and it’s been something like 15 years or so since that project got started. It came out of, you can call it laziness, but it came out of a desire to solve a problem. The problem I had was that I had all these customers that I was now responsible for, maybe seven customers or so at that time, and I had to meet with these customers on either a monthly or every-other-week basis. I had to prepare agendas for each of these meetings. I don’t want to spend time writing agendas. I hate the idea of writing out agendas that say we’re going to talk about every open ticket and what the status is and all this stuff. That’s a lot of work just to write out, and to be writing out constantly.
So what I did was, we had just moved ticketing systems and got access to the Salesforce API, because that’s what we were using at the time, and I was able to take the ticket data out of Salesforce, take a common report template that another TAM had built, and merge the two together into a Python script: we’re going to pull the data, build the report, and out of that comes the report. At first, you had to have a special developer license token to even call the Salesforce API. Over time, we matured our practices around it, which meant I could actually allow anybody else in our organization to use it, and I was able to release it to the organization. Over time:
- It went from being a command-line tool to one that was wrapped by a web application, so even the non-technical people in our organization were able to sign in and generate these reports
- We added a QBR presentation feature, where you could just log in and click a button, and your QBR was roughly 80 to 90% ready to go, at least from a presentation standpoint
These were very valuable to us because we had to do these things all the time, and to take the hours it would take to generate them and cut it down to, say, half an hour apiece was a massive time saving. So being able to have that builder knowledge, but being very close to the problem as well, like, “This is my problem and I want to solve it because I don’t want to have to be doing this work,” that’s where it came in handy.
Collaborating with Engineers and Prototyping
Navigating Engineer Boundaries
Rina Alexin | 00:09:15 – 00:09:40
And where were you in Office Space with TPS reports? I feel like that movie would have had a much different ending. But has being able to build helped you when you’re engaging with your engineers? Does the conversation you have with engineers get better when you can bring them something more concrete, or is there a balance where you shouldn’t be assuming too much?
Karl Abbott | 00:09:40 – 00:10:10
Yeah, it depends on the situation. Sometimes bringing something ahead of time is a valuable thing, and it gets you a lot further in the conversation. Other times there’s a desire from the engineers that, “We are the masters of how, and we don’t want a product manager starting to dip their toes into the how. So thank you for your idea, Mr. Product Manager, but please go away for now.” And I’ve had engineers come back a few days later and say, “So what were you saying again?” And I’ll say, “Okay, yeah, let’s bring that idea in now.”
Building an Offline Installer Script
Karl Abbott | 00:10:10 – 00:14:08
One thing that did happen when I was working at a company called Element: I worked on the Element Server Suite installer product. Before we had an installer, it was just a disparate set of projects that had to be installed separately and configured to work together, and it was a bespoke thing. You had to have a consultant come in and install it at scale. So my job was to come in and productize the installation of all these things into one consistent single plane of easiness, so that a customer can just go run the installer and get what they would have previously had to do with a consultant.
One of the environments that we had to do a lot of installations in was what we call disconnected environments, where there’s no internet and no outside network. If the installer needs a resource from the internet, it can’t go pull it. And our SREs, who we had working on the project, by the way, I love working with SREs, because they have an operational mindset at the start of building something. They’re already thinking, “How would somebody actually use this in production?” when they start writing code, which is phenomenal. So if you’re not already using SREs to help you write your software, in B2B especially, that’s a great place to start. But they were overwhelmed. We had so much work on the plate. We also needed to be able to install into these disconnected environments to start opening up opportunities, because one of the key places where our software was used was in these places where you can’t get to the internet: messaging software. Messaging software that could work for people within a certain disconnected network but not out there on the internet. That was a key use case for us.
I remember thinking, well, it can’t be that hard. We’ve got most of the parts. I’ll just write a document on how to put the existing parts together and deliver a document that people can follow, and then they’ll be able to at least install in disconnected environments, and eventually we’ll get this functionality into our installer, because the answer I had gotten was, “We don’t have the time to do that right now.” So I went and made probably the most complicated document I’ve ever made in my life. It was easy to follow, but if you missed even one letter in one step, you were probably going to end up in a really weird spot, and that’s going to take you a lot of time to unpack. You hear a lot of people joke about telling the AI, “Make no mistakes.” This was kind of like that: “Okay, customer, follow this document. Make no mistakes.” It’s just not possible. People are human. Somebody’s going to make mistakes somewhere. And I went, well, this is not sustainable. I can’t give this to customers, because they’re going to shoot themselves in the foot left and right with this document. I’ll turn it into some scripts, because it’s easy enough to tell a customer, “Go run the script in this environment, and then go run the script over here, and it’ll work.” And I got it to work.
And then I realized, uh-oh, I’ve actually built something. I’ve built an installer that works for these environments. And I went, I can’t own this. I’m a product manager. I cannot be the person that owns this. I can’t release it. So I took it to my SREs and I said, “Hey, guys, I can’t own this, but here’s what I did.” And they went, “This is amazing.” By the end of the week we had it in the product. It went from “We don’t have the time to even consider how to get that in” at the start of the week, which was Monday, to by Friday we had it in the product.
Sometimes that is actually very good, because at least they then saw how it could be done. But that was because I’d spent a lot of time with systems and doing this type of systems engineering, and that’s a lot of what these scripts were doing: basically using systems engineering tactics to get things installed and make sure that everything was in the right place. So I was able to do that, but then they were able to read what I did, and instead of using my hack-job version of it, they were able to take it and get it into the official product.
How Far Can a Non-Technical Product Manager Go With AI?
Rina Alexin | 00:14:08 – 00:15:05
So what you’re describing right now, I think, is very relevant to our conversation, because my reaction is, that’s incredible. We’re also talking about building in the past. The types of builds that are happening right now, in terms of the speed, it is more and more common for me to hear a story of “I built it in a week,” right? That wasn’t as common back then. And to your point, the reason it was enabled that way is because you’re close to the problem. You understand both the build and the need, the requirements. That’s what people are trying to find now that we have AI: can it be one person who is close to the problem and also knows how to build? But as I mentioned, a lot of product managers do come from non-technical backgrounds. So how far, in your opinion, can a non-technical product manager actually go with AI?
Karl Abbott | 00:15:05 – 00:15:42
You can get pretty far, but you can also get into a spot where you don’t know what you’re doing anymore, even though you keep making progress, which is a scary thought to me, at least, because the AI can get you very, very far. You can go from, “Hey, I’ve got this problem, I need a solution,” to having a solution, and it works. And you can say it works, but you don’t really understand why it works. If it ever breaks, or if there’s ever any question about what’s going on under the hood, that’s where it can start to become problematic.
How Non-Technical PMs Can Leverage AI
Three Steps to Start Vibe Coding Without a Coding Background
Rina Alexin | 00:15:42 – 00:15:55
Do you have any advice for people who do not have a coding background, and they know that there may be these kinds of expectations of them? Where do they start? How do they even learn about it?
Karl Abbott | 00:15:55 – 00:17:45
Yeah. You can go the traditional way: there are plenty of learning sites on the internet to learn how to code. That’s a very well-established path, but that takes a long time. You can also pull up an AI agent and start in three steps:
- Start small. Take a small problem that you understand. That could be something like a report you repeat every week, a small internal workflow, or a way to organize customer feedback. Don’t start by saying, “Hey, agent, build me an app,” because that’s exciting and interesting, and you will get somewhere really fast, and you’ll feel amazing because you’ll be like, “I just built this thing. I don’t know how I did it, but I built it.” Start small and think about what better would actually look like. Then ask the AI to build the smallest version of what better looks like.
- Make the AI explain itself, and edit the code yourself. Start prompting it to explain itself. Things like, “What does this code do? Why did you choose this approach? What assumptions did you make? What might fail with the design?” And then go into the code, and if you don’t know how to do this, ask the AI, “How would I go into the code and make changes?” Start making changes yourself, and start testing and seeing what happens. The goal isn’t to learn everything and become an engineer in an afternoon, but to start to build an understanding of what’s actually going on under the hood.
- Take it to an engineer. Once you’ve got a little bit of an understanding of how you’ve built it, and you may have made a few adjustments or modifications yourself, take it to somebody at your company who has a deeper understanding of the coding aspect and show them what you built. Ask what they think would need to be changed before that type of code could be trusted, or just get their insights on it.
The Line: When Someone Else Depends on It, Get an Engineer
A product manager should bring in an engineer the moment someone else depends on what they built. Personal tools and prototypes are fine, but anything going into production, or that other people rely on, needs engineering review before it can be trusted.
Rina Alexin | 00:17:45 – 00:18:15
I love that last step that you mentioned, because I think what we’re talking about, in terms of your background, is you’ve had so much experience that even you would know where there’s nuance, where there is an issue potentially, where there is a part of the flow or a security issue that could be happening. But if somebody is non-technical, it’s harder to know where to even look. Is there a line for you personally where you’d say, “This is it for me. I actually need an engineer in the mix”?
Karl Abbott | 00:18:15 – 00:19:25
Yes. The moment it actually has to touch somebody else. If somebody else starts to depend on it, that’s the line right there. I have a handful of vibe-coded things that I actually use on a regular basis to help me do my job a lot better. But I’m watching the processes, monitoring what’s going on. It’s me, so if something fails, I’m going to catch that failure and I’m going to see it. But if somebody else has to depend on that, this is where I’d really want an engineer to start reviewing what I’ve done and say, “Hey, please go easy on me. Take this. It’s working. It certainly works for me here, here, and here.”
“The moment it actually has to touch somebody else. If somebody else starts to depend on it, that’s the line right there.”
In my opinion, one of the places where vibe coding, especially with a non-technical person, starts to fall apart is when other people have to depend on the actual output. If you’re just doing little projects for yourself, it can be immensely helpful. If you’re just doing prototypes, again, immensely helpful. But if you’re going to put it out into production or need to have somebody depend on it, this is where engineering is needed.
Vibe Coding vs. Vibe Prototyping
Rina Alexin | 00:19:25 – 00:19:57
Love it. It’s a very clear line, and it does make sense. What I’m hearing from other product leaders is that they do have an expectation that their product managers know how to vibe prototype. And I like saying vibe prototype versus vibe code, because it’s really about getting more information about some of the assumptions you’re making. What speeds up discovery is that you can now vibe prototype multiple versions of something and get a lot more feedback faster. But it’s a clear line: when somebody else needs to touch it, get an engineer involved.
Risks of AI Hallucinations and Job Market Trends
Confidently Wrong: Verifying AI Output as a PM
Rina Alexin | 00:19:57 – 00:20:25
There’s definitely also a danger, and I’ve heard this not just in product but across the board: AI has this capability of looking like it’s right. It sounds right, but it’s not. It’s like the most confident and capable person sometimes, right? And it’s not always clear if you don’t know what to ask. If you don’t have that context or experience yourself, you don’t always know. Has that happened to you? Has it burned you or someone on your team before?
Karl Abbott | 00:20:25 – 00:22:20
I’ve not been burned by it, but I’ve also learned not to trust AI output wholly, especially in an area I don’t know. If it’s an area I don’t know, I kind of have to trust it until I find somebody to help me better understand that. But I had AI help me write a blog post, and it invented a daughter for me, and I don’t have a daughter. It’s like, “Oh yeah, and your daughter.” And I’m like, I don’t have a daughter. I never said anything about having a daughter, and now there’s this line in here about having a daughter. That’s easy enough to catch and say, “Yeah, that’s clearly wrong.” But if it’s confidently wrong about that, what else is it confidently wrong about?
“If it’s confidently wrong about that, what else is it confidently wrong about?”
I’ve used it for trip planning and things like that. Going to an area I’ve never been to before, I want to cut down the research time, and it’ll produce a beautiful-looking itinerary, and I’ll think, “Oh, this is great. This sounds great.” And then you start actually researching the different things in the itinerary, and you’ll find, well, that restaurant closed down five years ago. That hike is 16 miles instead of five miles. All these little inaccuracies just start surfacing and coming out.
On some level, there are certain things AI output is really good at, like a vibe-coded prototype. If I ask it to make a change, I can verify the change on the other end. I can run the app, and it’s either doing what I want or it’s not. I can verify that. So where I can verify AI, I find it to be the most useful. Where I can’t verify AI without doing a lot of deep research or going and getting an expert, that’s where I have a lot of trouble actually trusting that output, because it can be so confidently wrong. And yet a lot of times it’s right. It’s right enough that you want to trust it, but there is that wrong, confident edge to it.
Rina Alexin | 00:22:20 – 00:23:00
I can definitely share a personal experience here. My husband and I went to Mexico, and we used AI also to research restaurants. The restaurants luckily were definitely available, but there was one that was several miles farther away than [unclear] in our itinerary. So you’re right. You really do have to verify it, and at Productside we definitely take that seriously, because it is good at a lot of things. It’s good at writing very long documents that a lot of people also don’t want to read thoroughly, but then when you’re presenting this extremely long document, some of it’s going to be wrong.
The Junior Talent Pipeline Problem
Rina Alexin | 00:23:00 – 00:23:35
And so I think a lot of companies are going to have to figure out, especially, we’re hearing a lot about, okay, AI is not cutting jobs yet, but people are not hiring, so it’s essentially cutting jobs. And when you look at the kinds of jobs, a lot of them are more junior. As we’re talking about, the junior pipeline is actually going to be a problem down the line, right? Because you need experience to know when something is right versus sounds right. Do you have any opinions about what is happening in terms of the overall job market and job talent pipeline for product?
Karl Abbott | 00:23:35 – 00:27:10
On the job market, I do think that in general we’re able to get more done with fewer people. I don’t think you have to eliminate everybody, but to do the same amount of work, you can get more done with fewer people. I also look back in history and think that every time we’ve hit a point where we’ve been able to do the same with fewer people, we’ve come up with more things to do. So I kind of hope that the current job market is just in a corrective state, that we’ve overcorrected, and we will bounce back at some point.
But the junior issue is very interesting, because there are some short- and medium-term impacts that are hitting us today. A senior with AI can easily keep you from needing to hire an extra junior to help out and do that work. I think it’s a really bad idea to not hire juniors and to cut that entire pipeline off. And I think that longer term the market’s going to correct on this. Five to ten years down the road, or whenever, we’re going to look back and say, “Oh yeah, we never said that AI could replace junior talent.” It’s going to become such a bad idea at some point, and people are going to wake up and go, “We can’t do that. We’ve got to keep hiring juniors.” So hopefully this is just a short-term issue as well.
“I think it’s a really bad idea to not hire juniors and to cut that entire pipeline off.”
Lauren Hanford, who’s a product person as well, who I like a lot, wrote a great piece earlier this year that helped sharpen my thoughts on this. She argues, and I agree, that AI can give junior engineers access to bigger problems earlier in their careers. It can allow them to:
- Work across systems
- Participate in reviewing agent-generated pull requests
- Develop the communication and judgment skills that used to come later in your career
So you can actually ramp up faster with AI. But the difference between where we’re at today and the world that Lauren envisions, where this is just how juniors learn, is how we actually train them to do that and what our expectations are around that role. You can’t just hand somebody an agent and say, “Okay, welcome to the job,” and expect that they’ve now got all the judgment they need and can go in and do the job. They’re going to need:
- Mentoring
- Real work that they can own, with clear boundaries
- Feedback on their decisions
AI can actually accelerate that process as well. If we keep not hiring juniors, what’s going to happen is we’re going to have a lot of experienced people with AI starting to review more generated work than they can actually review. You’re losing the amount of judgment you previously had, and the longer you let that go on, the worse that problem’s going to get. You’re going to get to a point where you can’t judge the output of the AI, and now you’re hoping that AI itself, as a technology, gets significantly better. It gets better all the time, but you’re hoping it makes leaps in terms of being able to verify its accuracy. Because what we have to remember at the end of the day, and it’s really hard to go back to this point, is that AI takes your text, converts it into a series of numerical tokens, and runs it through a model that just predicts what the next numbers are going to be, based on the numbers it’s been fed. Then those numbers get reconverted back into words. So it doesn’t understand.
Rina Alexin | 00:27:10 – 00:27:39
I am also definitely agreeing with you on what you’re saying, in terms of AI can actually accelerate a junior person. I’m also seeing it when I talk to more junior people: they are way more AI-native as well. So I think they also have an edge, and if anyone listening is earlier on in their career, you absolutely have an edge, because you’re open to these ideas, and you’re open to testing out a number of different tools and learning that much faster.
Future of Product Management and Career Advice
Will Product Managers Exist in 2030? The Product Builder Debate
Rina Alexin | 00:27:39 – 00:28:05
Now I want to, you actually shared this with me, so I want to get your take on it: there’s this prediction going around that by 2030 there won’t be product managers left. I tried to say it in my doom and gloom voice. It didn’t really work. So this prediction is saying that there’s now going to be this single product builder role. This is the great compression of product design and engineering. What was your gut reaction when you first saw this prediction?
Karl Abbott | 00:28:05 – 00:30:05
So I first saw that prediction, and the same report says it’s by 2030, no more product managers. And they also say that by 2030, 80% of organizations will have a CPO. My first reaction was a very sarcastic, “Well, if there are no product managers left, what are all those CPOs going to do?” If you have nobody to do the work that you’ve now decided you need to do, what’s really going to happen? But if we get down to the seriousness of it, they made the statement to be provocative, but they’re basically citing role disappearance as opposed to role compression. It’s like, “Oh yeah, they’re just gone.” And it’s like, well, no, things are compressing, things are changing.
“I don’t think we’re going to end up in a world where product managers don’t exist. The title is still going to exist in 2030. It’s going to look different than it does today, absolutely.”
The title itself will definitely still be there, but a product manager is going to need to know more about design. They’re going to need to know more about engineering. They’re going to be able to cross these streams a little more. You’re not going to get rid of the need to have somebody who is able to dive deep into engineering. You’re not going to get rid of the need to have somebody who’s got a very, I’m going to call it, very human sense of design, because a computer can design, but a computer is just copying design based on what it’s been fed. Being able to tweak it in a way that lots of people are going to like, that’s still a very human skill. So you’re still going to need that type of skill in the design world to bring to the table. You’re still going to need all these different people. I don’t think you’re going to be able to just go to this one product builder role where, “Yep, you’re doing everything now.” If you want to be successful, you can start that way, but when you go to scale, that’s when you start to need the extra expertise.
Rina Alexin | 00:30:05 – 00:30:20
Yeah. That was my thought: if it’s just going to be a product builder role, that means it’s not just no product managers, there are no designers. What’s going to happen to engineers, right?
Why One Person Can’t Scale a Product Alone
Rina Alexin | 00:30:20
So are there other reasons why you think this one person mapped to a product may not actually work?
Karl Abbott | 00:30:25 – 00:32:00
Scale. They had this great thing in that same report. They said that 12 people today can do one product, and in the future, 12 people will be able to build 12 products. And it’s kind of like, well, do we need 12 products? Does a company need to sell 12 products? They’ve got one great product today. Do we need to pivot into having 12 products, or are you saying that since we have one product, we may as well fire 11 product people and now we’re down to one product person and we’re good to go? That’s a question of how you interpret that particular thing. But once you have a successful product, or once you have a product that starts getting adoption, you’ve got customers that come along with that, and you’ve got to take care of those customers. If a customer has a problem, you’ve got to be able to support them. As they want and need new features, you’ve got to figure out whether that fits your product direction and whether you want to invest the time in it. Then you’re going to go do that.
And let’s say it becomes wildly successful: you go from 10 customers to 100 customers to 1,000 to 10,000 to a million customers. This is one person handling all of this? No. You’re going to burn them out very fast. You’ll burn them out before you get to 100 customers. And you’re probably not even going to get to 100 customers, because one person comes with one bias and one point of view and may not know all the right questions to ask, no matter how good AI is.
Rina Alexin | 00:32:00 – 00:32:11
Yeah. I mean, that person can get very far. They can get a lot further with AI than one person without AI is going to get, but they’re not going to get the whole way.
“Where Do You Want to Be in 10 Years?” A Better Answer
Rina Alexin | 00:32:11 – 00:32:40
And I think that’s the truth. So we’ve talked a lot about how you yourself have had a variety of backgrounds, and we’ve talked about how product managers will need to reinvent themselves as well. So what advice do you have, given this job market, where people are actively interviewing and they’re no longer sure how to answer this common question: where do you want to be in 10 years? I feel like that’s a really tough question now. What’s your advice for people who get asked that in interviews?
Karl Abbott | 00:32:40 – 00:35:03
My advice today is going to be a lot different than it would have been 10 years ago. I spent 16 years at Red Hat. That was how my career started, and then I left of my own accord and went and spent two exciting but rocky years at a startup. We went through several layoffs, and I got caught up in the last layoff there. The job market in tech, as a lot of people know, this is no surprise, hasn’t been all that great. I spent six months unemployed, which I know is a lot less than some people spend, but four months is the bare minimum that I’ve heard in this market condition. Then, as if that weren’t enough, five weeks after being laid off, I decided to go try ice skating for the first time, and I broke my leg. So now I had a broken leg and I was laid off. After recovering enough, I even had to relearn how to walk. Navigating this period of joblessness and everything really changed how I personally see the world. I’m an ambitious person who really enjoys doing interesting work, but I don’t take stable employment and a reliable income for granted anymore. It’s not like, “Oh, I’ve got my college degree, and I’m going to find a company and be set for life.” I don’t think that’s ever the case at this point.
So for me personally, the answer to what I want to be doing in 10 years is to still be working on problems that I care about, with people that I respect, in a role that leaves room for a life outside of it. If you’re interviewing today and you’re getting asked this question, where do you want to be in 10 years? I don’t think you actually have to have your next 10 years entirely mapped out. That’s too big of a question. It’s too long a time frame to accurately map, and the industry itself changes too quickly. In 10 years, who knows what it’s actually going to look like. But I think it would be great to have an idea of the kind of:
- Work
- Learning
- Stability
that you want from now to 10 years out. So focus more on that than on trying to have all the answers.
“I don’t think you actually have to have your next 10 years entirely mapped out.”
Closing Remarks and Where to Find Karl
Rina Alexin | 00:35:03 – 00:35:25
Karl, I think that’s a really practical piece of advice. I almost think that if you’re getting asked this 10-year question, the answer is, “I don’t know, and I shouldn’t, but here’s how I learn.” And so I like your answer a lot.
Well, Karl, thank you so much for this amazing conversation that we’ve had. How can our listeners follow you after they listen to this episode?
Karl Abbott | 00:35:25 – 00:36:00
You can find me on LinkedIn, and as Rina mentioned earlier, I also run a product management podcast called Productly Speaking, which you can find at productlyspeaking.com or on any of the podcast platforms. So feel free to reach out through the podcast or reach out on LinkedIn.