Beyond Prompts: Lessons from a Year of AI in UX Strategy

AI isn’t here to replace what we do; it’s here to expand what’s possible. For our collaborative discipline, that’s an important distinction. It’s the difference between a threat to defend against and a capability to develop.

Our team was given real freedom to experiment with AI tools in our day-to-day UX work and figure out what stuck. A lot did. Faster synthesis, quicker first drafts, more ground covered. But the more interesting lessons came when things didn’t go as expected.

We’ve earnestly explored what AI means for our team via internal knowledge sharing, team AI summits, working sessions, and retros. What follows is a year’s worth of honest reflection: what we learned, what surprised us, and where we’re headed as the landscape continues to shift.

From Assistants to Agents: What Actually Shifted

A year ago, our AI usage was mostly experimentation using AI assistants: prompts in, responses out. As the year progressed, we learned how to prompt more strategically, create notebooks and agents, and dig into where AI might provide efficiencies in our workflows. AI isn’t just a layer in our process anymore; it’s becoming structural to the products we’re helping clients design. Agentic experiences, adaptive interfaces, systems that respond and learn — that’s the direction we’re heading.

However, that asks something more of UX practitioners. It’s not enough to just be fluent in prompting; we’re now thinking about the experiences users have and the AI implications for the systems making those experiences possible at the same time. That’s an expansion of scope, not a replacement of our core purpose.

What the Friction Taught Us

When we took stock of the past year as a team, three things kept coming up:

1. A lingering sense of disconnection. When AI is threaded through every stage of a process, ownership gets diffused, and everyone now has their own shadow collaborator. That sense of ownership — the time and effort that comes with truly exploring our ideas — quietly eroded in places. We felt it.
2. Looking for a nail with an AI hammer. There’s real competitive pressure in the market right now, and that pressure can short-circuit deeper understanding, replacing it with a sense of urgency that isn’t always justified. We found clients arriving already committed to AI as the answer, without the foundational clarity around data, structure, and goals that an effective AI product actually requires. The gap between wanting a solution and being ready to build one became part of our work in ways we hadn’t fully anticipated.
3. A productivity contradiction. AI output is fast. That sounds like efficiency until you’re the one responsible for reviewing it. However, more output meant more review time. The bottleneck didn’t go away; it just got redistributed.

The Friction in Collaboration

AI tools are almost always shared on an individual basis. One person, one prompt, with one clean result. What that doesn’t capture is what happens when that output enters a room full of smart, multidisciplinary practitioners during a critique, a client review, or a working session where four people have different relationships to what the AI produced.

We started calling it “AI-ing at each other” (a phrase I’m co-opting from one of my favorite design colleagues): sharing AI output in place of the alignment conversations that should have come first. The tool made production fast enough that stopping to align felt slow by comparison. But that’s where shared understanding lives. Skip it, and you don’t save time; you just move the friction downstream. For a team that values collaboration and ideation, this introduced real friction and real magic across the team that we don’t want to lose.

What We’re Keeping in Focus

Where has our exploration taken us? These are the things the friction pointed us back toward:

1. Collaborate with your people first. AI can accelerate a lot of things. Getting a team aligned around what they’re actually trying to accomplish isn’t one of them. The most innovative ideas will still come out of our brains, not from an LLM.
2. Keep the fundamentals in focus. Human-centered design principles are more important than ever. In a landscape that keeps shifting, they’re the stable thing to orient around.
3. Think about users and systems together. The experience someone has and the system powering it are increasingly the same conversation. Practitioners who understand the connection between the two are better positioned for success. It’s no longer just about creating the experience; it’s about orchestrating the experience and the systems behind it together.
4. AI expands the work. It doesn’t replace the people doing it. This fear is everywhere right now. Reframing it from replacement to expansion of capabilities keeps your team in the driver’s seat instead of waiting to see what role AI decides to leave for them. AI is pattern matching based on what already exists; it will never replace the innovation your teams can accomplish.

The tools will keep changing. The landscape AI exists within will keep changing. However, the discipline of keeping humans in focus doesn’t have to, and our exploration of AI has reinforced that as the core of the work we do as UX practitioners.

If your team is working through where AI fits, where it helps, and where it’s creating more friction than it’s worth, we’d welcome the conversation.

Photo Credit: Pawel Czerwinski | Unsplash