Every time I land in the UK, I catch myself hesitating at the curb. Look right, not left. Walk on the left side of the street. It feels unnatural…until it doesn’t. After a few days, that behavior fades into the background, and my actions become automatic. Defaults are the quiet architecture of how we think and operate. They reduce friction, conserve energy, and let us move fast without having to reconsider every step. Without them, we would stall due to cognitive overload. With them, we rarely stop to ask whether they still make sense, and that’s precisely the risk.

In 2009, Blackberry was the fastest-growing company in the world. Its worldview was coherent. Phones were for productivity, and productivity meant email. The physical keyboard wasn’t just a feature. It was the product. When the iPhone launched, BlackBerry didn’t panic. Their default told them no serious professional would trade a keyboard for glass. However, users didn’t want a better email machine. They wanted a computer in their pocket. The keyboard did not die overnight, but what had been a feature became a bug.

This pattern repeats more often than we admit. Kodak built the first digital camera and shelved it. Blockbuster passed on Netflix because the store’s Friday night ritual felt permanent. These were not blind companies. They were leaders who operated with strong logic, deep customer insight, and disciplined execution. As Clayton Christensen pointed out, good “management” is often the problem. Listen to customers, optimize margins, and double down on what works. These behaviors reinforce existing defaults, but when the underlying assumptions and paradigms shift, they accelerate decline. A good default, applied after its expiration date, becomes a liability.

We’re in another reset now. For the last fifteen years, SaaS has been the dominant default. Software delivered via the cloud, sold per seat, monetized on subscription. It was a massive upgrade over on-premise systems and created enormous value. Entire categories and some of the most important companies of the era were built on the premise that SaaS was a superior way to deliver software.

AI is now challenging that foundation. When an agent can handle customer support, process invoices, draft contracts, and manage workflows end-to-end, the unit of value changes. You’re no longer paying for software access. You’re paying for outcomes. Per-seat pricing starts to look misaligned. Build vs. buy shifts when software can be generated in days rather than quarters. The cost structure compresses. The speed of iteration increases. AI-native companies reach product-market fit faster, with fewer people and less capital.

Companies that navigate moments like this don’t wait for confirmation. They interrogate their own assumptions early, while they still have the option to act. Netflix did this repeatedly, making three transitions from DVDs to streaming to original content, each time disrupting its own model before someone else could. Microsoft, after missing major shifts in the internet and mobile, reset under Satya Nadella by leaning hard into cloud and AI, even though the payoff wasn’t yet obvious, because they re-underwrote their defaults.

At Zappos, we went through smaller versions of this. In the early days, navigation drove the user experience; then search took over; and finally, mobile forced a full redesign. Each shift was painful. Systems broke. Teams resisted. Conversion initially dropped, and it took hard work and time to get them back up. After some initial bake time after each transition, the old model became unthinkable. We recognized there is an asymmetry to these decisions. Changing defaults can be extremely painful, but the pain is temporary. Keeping defaults is comfortable, but keeping the wrong defaults is fatal.

Currently, multiple defaults are being questioned simultaneously. Code as a moat. Humans as the unit of work. Software priced per seat. Static models. Long development cycles. None of these defaults will disappear overnight, but in the age of AI, the half-life of a good default is shrinking.

Peter Drucker once observed that no matter how obvious they appear, we must always question all assumptions (defaults), because in a surprisingly high percentage of cases, what is known to be true turns out to be only partially true, inaccurate, or simply the residue of a world that no longer exists. What was a feature yesterday is merely a default today and may become a bug tomorrow.

From time to time, reexamine and reunderwrite your defaults. All of them. For the ones that are obviously broken, please fix them quickly. Please carefully assess even those that still seem to be working. Those are the most dangerous because you would not think to question them until it is too late.