Ship now, cry later

Ana and Egle talk to Niloo Roshani, a Design and Research Leader at NielsenIQ, about how the UX research function is changing in 2026: why her team are becoming thinking partners rather than the ones checking the "yes, we did research" box, why AI adds speed but not critical thinking, and why the most efficient research is sometimes no research at all. Along the way, the nugget AI summaries always miss, the kickoff that sets risk and reward up front, and a term coined live that belongs on a LinkedIn banner: Stakeholder FOMO. Plus the surprising comeback of Service Design, who would have thought. If AI only hands back what you expected to hear, who on your team is listening for everything else?

Episode 3: Care before taste or JUST DESIGN THE PAGE. 

Ana and Egle talk to AJ, Head of Design at The Times, about what the tools actually change and what they don't: why AI adds risk before it adds speed, why he'd rather see care than "taste" in a portfolio, and why he can't tell you what any of his thirteen designers are working on this week (on purpose). Along the way, the token audit almost nobody has run, a Figma prototype that took 659 versions, and one man in a research session who moved an entire industry by asking where the scratches were. If the tool amplifies whatever you already are, what is it amplifying in your team?

Why do so many teams run every ceremony in the agile handbook and still not trust each other? Ana and Egle talk to Marion - a UX-designer-turned-agile-coach and trained mediator - about the difference between performative process and the slower, harder work of actually building trust: showing up consistently, asking for help without treating it as weakness, and staying grounded in your own judgment as AI reshapes how fast everyone can move. What's your process actually protecting - your team, or your own discomfort with trusting them?

Notes
In the episode, Marion mentions a stat about workers trusting AI more than their manager, and credits it to Gallup. We went back to check - it's not Gallup, it's Oracle's 2019 "AI at Work" study with Future Workplace: 64% of workers said they trusted a robot more than their manager*. 

Small correction. Chasing it down turned up something more interesting than the original stat.
That number hasn't really moved since. A 2025 survey put a similar question at 67%**. A 2026 industry report on heavy AI users put it at 74%***. Worth a caveat: several of these are run by companies that sell AI tools, so treat the exact percentages loosely. But the direction hasn't changed once in six years.
But here’s something really interesting here. When you ask people the actual authority question - would you accept AI as your boss - independent academic polling says no, overwhelmingly (Quinnipiac put it at 15%, this year****). People don't want AI in charge. They just don't trust the human in charge enough to bring them a problem or not reprimand them.

This gap was visible in 2019. Before most of us had used a chatbot, before "AI at work" was even a category. Which means this was never really an AI story. It's a six-year-old trust story, and AI just gave people an easier way to act on it.

Marion has the actual answer here:
"I think we need to use AI thoughtfully, and remember to come back to our own inner voice. Use AI as a support tool, but don't let it think for you."

We'd add: don't let the AI story distract from the leadership one either.
Go ahead, click play.


*oracle.com/corporate/pressrelease/robots-at-work-101519.html

**resume.ai/reports/ai-vs-management

***glean.com/work-ai-institute/reports/work-ai-index-report

****techcrunch.com/2026/03/30/ai-work-boss-supervisor-us-quinnipiac-poll)

Ship Now Cry Later is a new podcast for designers and design leaders who are done pretending the environment doesn't matter.Episode 1: Roots Before Branches. What makes a designer or a team actually stable when everything shifts? Ana and Egle explore core values, craft identity, and the difference between flexibility and formlessness, with perspectives from industry voices on what resilience actually looks like in practice. How do you know what to hold onto when the role description changes again?