Sometimes I get the question about build vs buy. Don’t get me wrong, I’m a developer by heart and I’d always pick build.

But as a manager or leader I most often pick buy.

The equation rarely adds up when you build something that’s not part of your core business. Unless your core business is faced with a tight budget.

And by equation I mean time. Hours spent building, making it better, bug-fixing, upgrading, new feature, planning, sorting, prioritizing it, knowledge sharing.

Think of customer feedback as two distinct signals. One is complaint-driven insights about why people leave. The other is a silent signal that explains why people stay and remain loyal.

When looking at complaints and help requests you’ll only capture the minority of vocal customers. But your satisfied users might have the key insights to what makes your product worthwhile. Your loyal users often value something completely different from what complaining customers expect.

Value alignment over problem resolution.

You might have a bug in the product but you might also have the wrong customer.

With consensus everyone must say yes. With consent no one says no.

I remember when we introduced this at Spotify. Leadership felt it took way too long to make decisions - almost like paralysis by analysis. But once understood it was a fairly simple shift to get teams moving faster.

There are very few decisions we can’t reverse if they fail. Consent helps us move quickly toward either failure or success. Continuing to analyze and second guess rarely moves us forward.

One challenge with consent is that someone needs to make a stand. If discussions are endless then it might not be consent that’s lacking. It might be that no one cares strongly enough to put a stake in the ground. Or worse it might be a sign that no one feels psychologically safe to make a mistake.

One pattern I see is that we approach all AI with a similar mindset, an industrial mindset that focuses on production metrics and output targets. While AI excels at production, Generative AI’s real value might lie in distillation and discovery.

Think of it as a lens for spotting connections and extracting insights rather than a factory churning out content.

Earlier AI focused on clear tasks (recognize this, predict that, sort these) and it was easy to measure if the results were better.

But generative AI opens new possibilities for exploration and creativity. Yet we measure it with industrial era metrics and we see it in all benchmarks that the Generative AI companies share when they release a new model.

The power isn’t in producing more, but in helping us see different angles and uncover unexpected patterns. Quality over quantity. Insight over output.

The question shifts from ‘How can we create more?’ to ‘How can we discover better?’"

I’m fascinated by us “lazy creatives.” We dive so deep into planning that we never actually create. We research endlessly. Buy the perfect tools. Watch countless tutorials.

Right now I’m guilty of this with music. I’m setting up the perfect Ableton Live set. Buying gear. Watching performance videos. Even coding custom tools to make it “better.”

Have I made any actual music yet? Nope.

Bad art teaches you more than perfect planning ever will.

Here’s the truth though: Bad art teaches you more than perfect planning ever will.

Maybe instead of trying to get everything right we should just create something terrible on purpose. Make that awful track. Write that horrible story. Take those bad photos.

Because really. What’s braver? Having the perfect plan or actually making something?

Ever notice how some of your best ideas or solutions come while explaining something out loud? Just rambling out loud until something clicks? Rubber-ducking your way forward.

Most people stick to either talking or writing when processing thoughts. But you might get better results in combining the spontaneity in verbal processing and the structure of written words

Start talking. Let yourself ramble on a topic without filters. Just let the ideas tumble out.

Then bring it to AI. It captures your ramblings and reflects them back in text form. No judgment. No interruptions.

Now read the response. This is where scattered thoughts begin to make sense. Reading slows your brain down. Helps you spot patterns in your own ramblings.

It’s like having two thinking modes: creative outpour and thoughtful reflection. Your brain needs both to turn raw thoughts into clear ideas.

The AI revolution in video creation has me excited. It’s wild to think how storyboarding will evolve.

Directors sitting at their desks saying “I want a moody scene. Start with a low angle shot that slowly pans up to reveal a cityscape at dawn.” And then uses AI powered tools to create it.

Sharing vision becomes easier. Instead of struggling to explain what’s in your head you can show it. “Here’s the vibe I want” becomes a starting point not hours of back and forth.

Creating becomes less about technical skills and more about creativity. We’ll spend less time figuring out how to make something and more time deciding what stories need to be told.

Think of Harvey Pekar before he met R. Crumb. Amazing comic ideas trapped in his head because he couldn’t draw. Soon that barrier between imagination and creation disappears.

The tools are becoming easier. The ideas will become everything.

Alternative to estimates: do the most important thing until either it ships or it is no longer the most important thing. - Kent Beck

The challenge comes in knowing which thing is the most important thing. Similar to path of least resistance where water naturally flows downhill taking the easiest route. We tend to pick the things that are easier or use less time.

Most often we pick the path of most comfortable.

There is one rule you should never forget: Everything is saved on the internet.

Let me share a recent example. A company decided to use a shared ChatGPT account to save money. One day. the owner started a new chat with the topic “How to fire people”. This was visible to everyone using that account.

Even though you think a service is secure. moving data anywhere online means it can end up in the wrong hands. Your “deleted” content lives on in endless iterations as backups. archives. and screenshots.

And when you’re using shared accounts? The risks multiply. Your private information becomes accessible to everyone with access.

Always remember: Nothing is truly private online.

Want to write powerful questions for ChatGPT and its counterpart, the humans?

Focus on four things: construct, scope, assumptions, and reduction.

Construct “What”, “How”, and “Tell” are excellent question starters that can help with reflective thinking. “Why”, “Who”, “When” limit our responses.

Scope Ask from the point of view that you must take, be it a stakeholder or your own.

Assumptions In many great questions, there are inherent assumptions, make sure that you know yours and how to use them for benefit.

Reduction Do you want to reduce the output or response that you get? Example: “Tell me about yourself” vs “Tell me your three favourite movies, growing up”

Friendly reminder: answers can be strange, odd, and bad.

Every answer is someone’s interpretation of the truth. A question is a desire to understand that truth. - Darko Vukovic