AI makes software cheaper, UX research makes it safer

Lately I have been thinking a lot about how cheap software is becoming to build. With AI assistants, even someone like me, a designer rather than a developer at my core, can create fairly sophisticated systems. Things that once required large teams, long timelines, and serious budgets can now be explored, built, and iterated on by a small group or even a single person.

That shift changes the way I think about UX research.

For a long time, research made obvious sense because building the wrong thing was expensive. Code was hard to change, delivery cycles were slow, and mistakes could linger for years. Prototypes were a safe way to learn before committing to production. Research helped reduce that risk.

Today, the cost of building has dropped dramatically. Interfaces can be generated, refined, and rebuilt quickly. Many interaction patterns are well established and AI tools already reflect a huge amount of collective design knowledge. In some situations, you can build something reasonable, ship it, and improve it without doing much formal research at all.

That does not mean research has lost its value. It means its purpose has changed.

AI is good at recognising what usually works. It is much less capable of understanding why people behave the way they do in specific environments. It does not feel the pressure of compliance, the fear of making a costly mistake, or the personal risk that comes with using the wrong system in the wrong way. These things shape behaviour far more than most UI decisions.

In complex or regulated organisations, people rarely use software exactly as intended. They adapt it to survive their day, they build workarounds, and they avoid features that feel unsafe. None of this is obvious from looking at screens or metrics. It only becomes clear when you spend time understanding the context people operate in.

This is where I now see the real role of UX research.

Research is less about validating designs and more about understanding reality. It helps uncover hidden constraints, unspoken rules, and the reasons behind behaviour that can otherwise look irrational. It creates clarity in situations where building faster does not necessarily mean learning faster.

There is also a practical side that often goes unspoken. In large organisations, research helps align people. It creates a shared view of the problem and gives teams confidence to move forward. This matters even more when delivery speeds increase, because decisions are made faster and the cost of misunderstanding grows.

Not all research is worth doing. Testing familiar patterns or obvious interactions adds little value. Experience, good judgement, and AI support can often cover that ground. What still matters is research that helps teams understand risk, consequence, and complexity.

As a designer, this shift feels important. My role is less about protecting design decisions and more about helping teams understand the systems they are building for. AI makes software easier to create, but it does not remove uncertainty. UX research, done well, still exists to reduce that uncertainty where it really matters.