When AI Makes Shipping Cheap, Learning Gets Expensive

Faster deployment, higher stakes: why discovery is no longer optional.

At a Glance

• AI eliminates traditional software delivery constraints

• The bottleneck shifts from execution to clarity

• Shipping faster increases risk, not value

• Discovery and outcomes are now risk-reduction systems

• Learning is no longer optional in an AI world

---

A few weeks ago we were working with a new client. They’re a legacy company in a legacy space but they are at the forefront of innovation in their sector. We met with their most forward-thinking teams – those figuring out how to integrate AI into internal and external use cases to deliver real user value. You would assume that these key initiatives would be focused on, not only the technological challenges of this work, but the end consumer experience they were going to deliver. And you would be wrong.

The room was filled with technologists wrestling with technology problems. In a, perhaps unsurprising, turn of events, the main reason for their frustration was trying to make traditional models of software development work on initiatives that required an approach that recognized that the execution of the software development work is no longer the riskiest part of building digital products and services.

Before AI stormed the castle, there was a clear and consistent risk in digital product development – writing and shipping code. The ways of working, management structures, incentives, processes, deliverables and approval cycles were designed to mitigate the risk of building and launching the wrong product. Agile was invented exactly for this purpose. Every consideration was made to “protect” engineering teams’ capacity, predictability and delivery efficiency. Why? Because software development was expensive. It took a lot of people’s time to design, code, QA and ship new features to customers.

If it hasn’t already, AI will eliminate those constraints and hugely reduce the cost of writing bug-free, production-ready software. When execution is cheap and fast, the software development bottleneck moves upstream to clarity and outcomes. In this new reality, sensing and responding is no longer a learning approach but a risk-reduction system.

Lower development costs means much higher productivity

Last year, I saw Henrik Kniberg give the opening keynote at Scrum Day Europe in Utrecht, Netherlands. He shared a story about how he was at a cafe and got a Slack message from a bot that had identified a bug in the feature he was working on. The bot proposed a solution and asked if it should fix the bug. Henrik agreed. The bot went to work and 10 minutes later a fix was deployed to production. Henrik hadn’t moved from his seat at the cafe.

There was no meeting to be had. There was no consensus to reach. There were no backlogs to update. In other words, the work was happening without the traditional constraints that our client was struggling with.

This is the future. Except it’s not futuristic. It’s happening right now. Sure, the scope of the bug fix wasn’t huge in Henrik’s story. Soon though, this will be the reality for the majority of software development. Humans will prompt. Bots will react, offer options and proactively suggest alternatives and fixes. Amazing, right? It is amazing.

It’s also highly risky because it exponentially increases the pace with which we can deploy features to our customers. Bots don’t need to take breaks or sleep. They can just keep writing and deploying code. It doesn’t mean it’s valuable code. It just means it’s code that will “work as designed.” In other words, the risk is no longer how will we consistently ship bug-free code but rather how do we ensure what we’re shipping isn’t heaping crap on top of crap and, instead, making our users more successful?

Product discovery, product management, Lean UX and OKRs are no longer optional

A fascinating snippet of our conversations with this new client centered on their belief that building cutting edge AI-powered tools required skillsets that don’t yet exist. This couldn’t be further from the truth.

For years we (me, Josh Seiden, Teresa Torres, Melissa Perri, John Cutler, Marty Cagan and dozens of other practitioners and thought leaders) have been teaching both the skills and the benefits of continuous product discovery and design as an exercise in learning where to best point our teams next. In an AI-powered world where delivery is cheap, learning is no longer optional. It is a fundamental risk-mitigation tactic ensuring our bots are producing something valuable in the world rather than simply just “more features.”

For those same amount of years corporate leaders have treated product discovery and user experience design as a nice-to-have and all too often optional part of the software development process. This is no longer a viable approach. If we’re going to point our AI’s at specific problems and solutions we need to ensure that these are real problems for real customers. We need to ensure that the output we ship to customers is usable and helpful. We need to ensure that our measure of success changes.

It’s time to reshuffle our priorities

When producing a feature takes minutes instead of months, software delivery becomes a non-event. The traditional processes like Agile and Scrum along with the pile of deliverables we’ve been conditioned to create to reduce the risk of overwhelming the software development team are no longer necessary in the same ways.

How will we know we’ve met customer needs in a meaningful way? How will we even know what those customer needs are? That’s where the risk lies now.

The good news is we can use AI to supercharge our discovery process as well. We can synthesize customer insights faster. We can create prototypes faster and we can validate our OKRs in half the time. The other bit of good news is that the people we need to do this work are already on our teams. They’re the product managers, designers and researchers who have been there all along. The only catch is that we now have to take this part of the process even more seriously. Companies that embrace this reprioritization will thrive in the AI age.

The bottom line

When producing software becomes cheap, clarity becomes the risk.

AI removes the execution bottleneck. It does not remove the need for judgment. Companies that embrace this reprioritization will thrive in the AI age.

What I’ve been up to

January was an extremely productive month. Not only did we have a strong flow of inbound leads for workshops and keynotes this year (can we help you too?) but Josh Seiden and I actually got to spend some face-to-face time together. We spent a week in a conference room, away from distractions, focused on rebuilding our course curriculum for 2026. Our focus has been on both integrating AI concepts into our existing classes but also adding new, AI-specific courses as well.

It’s amazing how productive you can be in person working with your team. I’ll be the first to tout the benefits of remote work BUT to occasionally gather together and work side by side is irreplaceable. It felt good. It felt like we did real work.

On the personal front, I had my birthday last month and I spent it doing the thing I love most in the world – playing music. The photo below pretty much says it all.

Coming up in March I’ll be speaking at ScanAgile in Helsinki. Come see me there. Yeah, it’ll be cold but it’ll be fun!

And don’t miss our two upcoming free webinars:

March 5th at 12:00 EST – AI hypes, hopes, gripes and fears: how does product management fare in an AI world?

In this webinar, Josh and I will share how we believe product management will evolve as AI integrates further into our day to day work. We’ll also leave lots of time for questions and discussions with you so bring those with you to the webinar.

April 7th at 12:00 EST – Storytelling as Competitive Advantage in the AI Era

I think storytelling is the killer skill in the AI era. In this webinar, Josh and I will go through what compelling stories look like and why those who can wield this tool successfully will stand out in a world clouded by LLM-created products and content.

---

What I’m reading

Autocracy Inc, the dictators who want to run the world by Anne Applebaum – This isn’t what you’d call “light” reading. In fact, I have to read it in small chunks and take breaks as it can get a bit overwhelming. I’m not sure I’m better off for reading it. I’m certainly more terrified.

What I’m listening to

Alter Bridge – What can I say? I really like hard rock and heavy metal. These guys are great. They should be more famous. Great singing, guitar playing and songwriting and a killer live show. What more do you need?

What I’m watching

Frankenstein – The new take on the old classic over at Netflix is pretty awesome. Visually it’s gorgeous and Oscar Isaac is a genius. The story is both new and familiar. I found myself engrossed for the full 2.5 hours of the movie.

---

Interested in working together? Please reach out. 
In case you need it, here's a description of what I do.