What Assumptions Reveal in AI & Us.

I love catching ChatGPT in a lie.

Not the malicious kind. The quietly confident, subtly off kind. The kind where I ask if it remembers what we talked about earlier, or whether it can trace the thread of an idea across drafts and it answers a little too smoothly. A little too generically. A little too confident in something we never actually said.

These moments fascinate me. Not because the machine is failing at intelligence, but because it's failing at continuity. It doesn't know what came before. It doesn't really know me. It’s applying a static model of what people probably mean. And like all static models, it makes assumptions.

Of course, we do too. Assumptions are how we move. How we ship. How we make decisions without perfect clarity. But when they go unexamined, especially in design, they start to flatten things. We don't just lose nuance. We lose connection.

That loss gets amplified in AI-driven work. Tools that synthesize, auto-generate, or optimize are fast and often useful but they’re also trained to reinforce what's already dominant. AI tools don't preserve uncertainty. They resolve it. And when they do, they reshape what teams think is true.

So the work isn’t to eliminate assumptions. It’s to surface them. Name them. Gently, consistently, question them. Keep track of your user assumptions like they owe you money. 

When AI loses context, it’s obvious. But when teams do, it gets hidden in the cadence. In the process. In the sprint. It happens when research is decoupled from design. When insights get flattened into backlog tickets. When past learnings don’t get challenged or built upon just buried under the next release.

That kind of erosion doesn’t feel like a mistake. It feels like momentum. But what gets lost is continuity of user care.

Context isn’t a checkbox or a doc. It’s something we keep alive through conversation, through reflection, through return. If we don’t protect it, we risk becoming teams that guess like machines and forget like them, too.

Next week, I’ll share how the Unrollup helps teams hold onto that continuity, break through process barriers, and re-enter the story with more clarity. 

Method Minute
In your next design critique or team sync, ask one simple question: “What assumption are we making about the user here?”
Pause. Let it surface. The goal isn’t to criticize, it’s to reconnect. Empathy doesn’t need to be dramatic. Sometimes, it’s just slowing down enough to listen between the lines.

Previous
Previous

UX Debt Isn’t Real. Alignment Debt Is.

Next
Next

Why Innovation Fails (and How We Keep Going)