Category: Career

Building digital products is changing. 

If you were able to take anything away from my somewhat meandering and self-aggrandizing Becoming a Product Builder essay, it was probably this. 

It’s easier than ever to build good software. 

After spending the past 90 days doing it, I can confirm this is true. But that really wasn’t the point I was trying to make, nor is it really insightful.

And while I’d be one step closer to becoming a proper influencer if all I did was pound my chest, I do actually like writing things that at least attempt to have a nugget of insight within them.

So, here is Becoming a Product Builder, Part 2: The Actual Insight.

AI Builds the Average

Strauss Zelnick, the CEO of Take-Two Interactive, has a really thoughtful take on this in his interview on David Senra’s podcast.

He points out that AI is inherently backward-looking. It is trained on vast amounts of existing information and then creates new information by remixing it. 

I think he’s right. AI is exceptionally good at giving you a derivative product, a product that is like another product but slightly different. It produces roughly the average of everything it’s seen. 

When building with AI, it’s easy to think that its initial response to a prompt is good enough.

I can state confidently, it is not. AI nearly always produces “good” output; it rarely produces “great” output.

This concept is why so many of the real influencers are harping on the importance of “taste”. In a world that is going to become increasingly saturated by the average as more and more people leverage AI as their primary means of creating output, the experiences and businesses that stand out are going to be the ones where the people leveraging AI have taste. 

The people who can compile all of the average output of AI and combine it into something greater than the sum of its parts are the ones who will be able to find success. When building with AI, never settle; always strive for something that is uniquely delightful. 

The cost of delight has never been cheaper

The Old Way

A delightful product has always been valuable. Books have been written on how to structure your backlog so that delight is included (if you haven’t read Product Delight by Dr. Nesrine Changuel, pick it up; it’s a good one). 

But historically, building something uniquely delightful has been exceptionally difficult, and not because it’s actually hard to build. 

In the old way of product building, nearly everything was a trade-off. 

As the PM, I’d have a backlog of things I could ask my very expensive engineers to work on, and the things that would trickle to the top of the backlog almost always had some kind of return on investment. 

You could argue that a product manager’s primary role is to maximize the return on the company’s investment in engineering team members. 

In that environment, it is difficult to justify investment in weird stuff. It’s the premise of The Innovator’s Dilemma; it’s why the phrase “No one ever got fired for buying IBM” exists. It’s the antithesis of classical OKR’s.

In the face of uncertain ROI, it’s easiest to make rational, defensible positions. Invest in what you can measure. 

AI changes the equation. 

The New Way

Here’s a simplified example. 

One of the many products I’ve built recently is a custom web app that shows on the projector as you enter the suite. It welcomes you to the suite, tells you when your reservation starts, etc. Pretty simple. Nothing fancy. 

It’s June 15th. I see July 4th around the corner. I think “Man, it’d be fun to make the web app explode with fireworks before it unlocks.” 

In the old world, building this would have looked something like this: 

• Meet with the designer, tell them what I’d like it to look like. 
• Get the designs.
• Prioritize the work over another feature, probably one that was enabling a measurable, ROI-driving metric. 
• Get the engineers to focus on the fireworks and build and deploy it. 

Not impossible, but at the end of the quarter when I share with my leadership team how my team missed our metrics, saying “But I added fireworks to the lock screen” probably wouldn’t win me any awards. 

In the new world, it took about 30 minutes of back-and-forth with AI, and it was deployed. The cost? Essentially nothing. 

The Point I Missed

The point I failed to make in “Becoming a Product Builder” is that becoming a good product builder isn’t about replacing software engineering. It isn’t about how much you can output in 90 days. 

The point is, as the number of average products exponentially increases due to AI simplifying software development, the importance of learning how to move beyond the average to differentiate your products has never been higher. 

The point is, AI allows you to build delightful products in a way that sidesteps the classic ROI paradox most product teams face when trying to invest in disruptive or off-the-wall ideas. It lowers the barrier of entry to building good software, but it also lowers the barrier to entry of building great software if you just make a concerted effort to do so.

This is the paradox of AI. 

Yeah, it will probably make the average internet a vastly less interesting place filled with slop and weird videos of cats in parliament. 

But if you use it right, it gives smart people an opportunity to stand out amongst a sea of slop. 

So go create delight in your products. Don’t settle for good software; settle for great software.

⁃ Brian

The Context

Around 180 days ago I wrote “Must Go Faster“. I was grappling with what AI meant for my job. I’d spent the first 15 years of my career building products mostly the same way. Find the idea, prioritize it against the other ideas, work with the designers and engineers to understand the idea, work with everyone to build the product, measure its quality, communicate the outcome, repeat.

I think the reason I felt an inherent resistance to AI was that it felt as though it was fundamentally changing how building products would work. Enough of a change that it felt like my 15 years of hard-earned experience was about to be moot. The kid coming out of college was going to know more about building products than I was. It was a tough pill to swallow.

Reflecting on it, I think moments like this happen to every career, where something changes so drastically that your old experience becomes far less valuable. You’re faced with a choice when moments like this occur.

1. Put your foot down and say “new way bad, my way good”

2. Change.

I love change.

The Problem

The natural next step? Start building products the new way and get the experience. Use my existing experience as a multiplier to radically level up my product building game.

The problem was I was a people manager at a large org, with systems far too complex for me to just play around with. My job was mostly people and strategy; the work of building things was a bit too far outside my realm of responsibility, and my calendar was filled to the brim with the relationship and strategy work that comes with high-level management at a large organization.

The Solution

Around 90 Days ago, I quit that role. I decided to take the leap of faith, to follow the happy path I laid out in “Must Go Faster“, which sounded like this

“If AI is what its proponents believe it to be, AI will allow me to build things I could have never dreamed of building before at a pace previously unfathomable.”

I joined a growing startup, Another Nine, as the VP of Product and Technology. Now, that may sound like a title that would be filled with even more people management and strategy, but at a 10-person company, there can only be so many meetings. And my organization size? 1.

The beauty of this role is that someday I will get back to being a manager. As Another Nine grows, so too will my domain. I’ll get back to a part of the work I love, growing people and their careers. But for now, I am a 1 man organization, and all I do is build.

One of the many reasons I took this role is that it’s like a paid language immersion course in the new AI-driven ways of product building. Every day, all day, I sit with my bots and advance the the company’s technology any way I see fit. I have maybe 4 hours of meetings a week. The CEO trusts me enough that I barely have to convince him my ideas are any good. I have a growing company with a myriad of problems to solve. I have everything I need to learn, and I’m getting paid to do it.

The Result

Someday, this quantity of output might seem normal (and yes, it’s output, not outcomes; leave me alone you data-driven zealots I’m doing my best). But writing it right now feels like bragging, especially given the pace at which things were built in my first 15 years of product.

Here’s what I built in the first quarter at a new organization, from 0 to 100 PR’s.

Did you catch that last one? I built an iOS and Android app. I don’t know how to program. It took me 4 weeks.

This would be the equivalent output of an entire organization of people at previous organizations. Hell, just building Student Discounts in a past life took an entire quarter with multiple teams working on it. At Another Nine, we had the idea on a Monday, and it was live on Tuesday.*

Quick Asterisk

While I am proud of everything I have built, I know that building for small-scale at a small organization is a wildly different thing than building for large-scale at large organizations. While nothing I’ve built is a toy, I don’t know how it will hold up to the pressures of a growing business from a stability or security perspective. These are things I will have to learn along the way.

I also have to give credit to our development agency, Mtn Haus, who have:

• Built an amazing foundation of backend services and architecture before I arrived for me to layer these products on top of.
• Reviewed my PR’s tirelessly and with grace and given me the confidence and adjustments required to productionalize all of this.

Thank you to you all and the engineers who put up with a crazy product guy trying to build fast.

Loving the Bomb

The title of this post is a tribute to the movie “Dr. Strangelove, or How I Learned to Stop Worrying and Love the Bomb.” It’s an old classic, a satirical take on how a government determines that building a doomsday device capable of destroying humanity will save humanity. It’s all I could think of when thinking about how much I love building with AI.

Part of why I’ve built so much in the past 90 days is because when I said “all I do is build,” I was being somewhat serious. Yes, my workday is filled with building. But once my kids and wife go to bed, I spend another 1-3 hours almost every night building other things. Not because my boss expects it of me, but because it’s fun.

I’ve loved building things since I was a kid. Growing up, I wanted to be an architect, and since college, I’ve always found time to build things in the physical world. But in the digital world, I’ve never been able to build much myself; I’ve always had to build through people. I’ve always had to convince someone to help me bring what I wanted to see in the digital world to life. That’s the fundamental change AI has created. I can build software now, and it’s addictive.

So, is AI going to destroy us all? Probably. But damn if I’m not having a blast riding it down.

Perhaps you’ll remember the scene. Jeff Goldblum’s character from Jurassic Park, Malcom, is resting in the back of a jeep. His eyes open as a deep rumble is heard in the distance. Another rumble, closer this time. In the foreground, a puddle begins to ripple. He yells to Ellie, “We gotta get out of here, right now!”

The jeep accelerates as a T-Rex erupts from the tree line. “Must go faster” he states, as the enormous dinosaur closes the gap on their car.

For whatever reason, my Mom loved this line. So much so that it became a household staple. “Must go faster” has been quietly muttered in my family anytime something needs to speed up for the better part of the past 3 decades.

I’ve been thinking of it a lot lately. Except I’m Malcom, and AI is the T-Rex.

AI Isn’t the Problem

While I wouldn’t object to the idea that AI is a T-Rex level threat, that’s not really my point.

Sure, AI will fundamentally change the trajectory of many careers, mine included. Perhaps for better, perhaps for worse. Many jobs will be eliminated, and many jobs will be created.

In general, my belief is that the algorithm amplifies the extremes. From the pundits screaming “AI will replace everyone!” to “AI will eat itself and never outgrow its hallucinations!”, the truth with the internet often lies somewhere in the middle.

So why is AI making me feel like I’m constantly on the run?

When How Fast You Can Go Is No Longer How Fast You Can Go

I’m an optimizer. Perhaps it’s why I’m in my field.
I can’t stand wasted effort. A wrong turn on a trip is my happiness kryptonite.

For my entire career up to now, there really has only ever been one way to get work done. Me.

Got a presentation to build? Get the story on paper and start putting slides together.

Got a PRD to write? Get my notes and a cup of coffee, and get to writing.

There was no choice here; the work would get done as fast as I was able to get it done. The quality was as good as I was able to make it.

In a way, this way of working is constantly operating on the most optimal frontier. There were no “wrong turns”. Sure, I could use some hotkeys or learn some tips and tricks or leverage templates to make marginal improvements. But in general, there was no way to exponentially increase my output. As fast as I could go, was as fast as I could go.

No more.

With AI available, I never feel like I’m going as fast as I could go. As fast as I should go. It’s like having an endless queue of employees at the ready, if only you could manage them effectively.

Must, go, faster.

Decisions, Decisions

These days when I work, nearly every task has hundreds of moments where I could implement AI to potentially accelerate the work. And every time, I have to decide, should I? Which way will be the most efficient way to create the output I want? (Perhaps I should create an agent to help me decide.)

Being constantly barraged by micro-decisions on the route I take as I try to do work leaves me with a nagging feeling that what I’m doing isn’t enough, a creeping doubt that I’ve made a wrong turn somewhere, that I’ve wasted effort.

I find myself halfway done with a task thinking “Bet I’d already be done if I’d used AI”. Or, worse, using AI for an hour on something and having to scrap it completely because the AI solution just isn’t working.

For me, AI sells the optimizer’s dream, but so far has been unable to deliver.

Damned if I Do, Damned if I Don’t

I suppose the simplest solution to avoid these decisions is to simply always use AI. Take the decision out of it. AI is the default path and should only be skipped when it doesn’t work.

Honestly, it’s not a bad solution.

Benefit #1: It will cause me to learn the most about leveraging AI in my work. That seems like a wise career move (see above re: job elimination). Adaptation to this new way of working is being preached as pivotal to success in my industry, and it likely contributes to my persistent feeling of needing to go faster.

Benefit #2: It’ll help reduce decision fatigue. If AI as the default is just “how I work”, I don’t have to belabor decisions on when to apply it and when not to.

Benefit #3: On the whole, I’ll probably get more done if I do it right. Granted, the first year or so will likely be slower as I learn all the ways to appropriately implement it. But eventually it should be faster… right?

A Leap of Faith

Most signals indicate the above to be true. Eventually, it probably will be faster. But there are some frustrations with the AI-first path that make me hesitate when considering making the jump.

Frustration #1: Checking Work is Less Fun Than Creating Work

Here’s the thing. I like my work. I like creating things. I like crafting stories through data or presentations in a way that people find useful. I like that I worked hard on it, that I toiled away on a presentation, adjusting the positioning to be just so for it to draw your eye to the right information. And when people like it, I feel proud of my work.

With AI, I haven’t learned to feel that ownership. So, instead of feeling like the things I’m doing with AI are my creations, they feel like I’m just checking over someone else’s work.

Frustration #2: Non-linear progression on specific tasks

Historically, when I work on something, progress is mostly linear. There is almost never a situation where work that I do is completely thrown away wholesale. Components can be reformatted, formulas reused, structures altered to meet changing needs.

With AI, this is only sometimes the case. Sometimes, because of the probabilistic nature of AI and the ease of creating ouput, there are large chunks of work that are just…garbage. I’ll go down a path with an agent that ultimately results in something unusable. Once I reach this point, trying to get the AI to unwind the issue often creates even more complexity in the system, so I find myself just restarting from scratch. I would imagine this frustration will lessen as I become more proficient, but still, it’s a pain point currently.

Frustration #3: Loss of Critical Thinking and Skills

This frustration is the one that keeps me up at night. There is an old man inside of me that just wants AI to get off my lawn. A fear that, if I don’t do the work the way I used to do it, I won’t be able to do it anymore.

It reminds me of passage (I’ve abridged it, but the meaning is ultimately the same) from Carl Sagan’s book A Demon Haunted World…

“We’ve arranged a global civilization in which most crucial elements profoundly depend on science and technology. We have also arranged things so that almost no one understands science and technology. This is a prescription for disaster.”

AI feels like a perfect illustration of this issue. There is a fear inside of me that true learning stops when I start allowing AI to do the hard part. Instead, AI learns, and I simply learn how to use AI.

The Advice

Unfortunately, I don’t really have any. Perhaps I wrote this to find some. Perhaps I wrote it to give more of a voice to the people in the middle.

+ On the one hand, if AI is what its proponents believe it to be, AI will allow me to build things I could have never dreamed of building before at a pace previously unfathomable. This is good.

– On the other hand, it makes my work far less enjoyable. It leaves me feeling like I will never reach the optimal frontier of my work again. “I could always go faster if I could just leverage… X”. It’s a T-Rex can that cannot be escaped. This is bad.

I wonder what AI would do?

For now, we just keep driving.