Pavel Samsonov - AI Can Build the Solution... You Still Have to Design the Problem

On this episode, I speak to Pavel Samsonov, Principal UX Designer at Justworks and author of The Product Picnic newsletter. Pavel has spent his career working across UX, service design and product management, including roles at Amazon and Bloomberg, helping organisations design better products by understanding the systems, processes and people behind them. We explore why great products start with better problem definition, how organisational silos undermine customer experience, why AI is making it easier to build the wrong things faster, and why genuine user understanding remains a uniquely human advantage.

Episode Highlights

• Design for understanding, not simplicity - Complex B2B products don't need to hide complexity; they need to present it in a way that users can understand, navigate and act upon confidently.

• Customer journeys don't follow organisational charts - Teams optimise their own domains, but customers experience the whole service. The biggest opportunities often lie in fixing the gaps between teams rather than improving individual features.

• Problem design matters more than solution design - Before discussing features or interfaces, ask whether you're solving the right problem, why it exists, and whether it's important enough for customers to actually care.

• Actionable beats visible - Dashboards, analytics and metrics only create value when they help someone make a better decision. Data without action is little more than decoration.

• Optimising your work can create someone else's workload - Shipping work isn't the same as completing work. Teams should think about who consumes their outputs and whether they're genuinely fit for purpose.

• AI accelerates production, not learning - AI makes it dramatically faster to generate prototypes and features, but it doesn't shorten the time required to validate ideas, learn from customers or understand real-world usage.

• High-fidelity prototypes can create false confidence - Just because AI can generate something that looks finished doesn't mean the difficult work of alignment, prioritisation, research and iteration has been done.

• Synthetic users aren't a substitute for real customers - Large language models can reproduce existing knowledge but can't uncover the tacit insights, unmet needs and market opportunities that come from talking to real people.

• Good product decisions require shared language - Cross-functional collaboration improves when teams focus on the decisions they're trying to make rather than debating ambiguous labels like "prototype", "MVP" or "research".

• Ask better questions before building faster - AI has made building dramatically cheaper, increasing the importance of asking why something should exist in the first place. Better problem framing remains one of the highest-leverage skills in product development.

Connect with Pavel

• Newsletter: https://productpicnic.beehiiv.com/
• LinkedIn: https://www.linkedin.com/in/pavel-samsonov-44ba2833/