Building AI products users actually want

In an era where artificial intelligence seems to dominate every headline and product roadmap, distinguishing between genuine innovation and AI for AI's sake has become increasingly difficult. A fascinating conversation between Sarah Sachs, AI lead at Notion, and Carlos Esteban from Braintrust provides a refreshing perspective on building AI products that transcend the hype cycle. Their discussion cuts through the noise to deliver practical insights on creating AI tools that genuinely improve users' lives rather than merely checking a technological box.

Key Points

• User-centricity trumps AI sophistication — The most successful AI products solve real user problems rather than showcasing technical capabilities. Both experts emphasize that companies should resist the temptation to build AI features without clear user value.

• AI development requires cross-functional collaboration — Unlike traditional software development, effective AI products demand tight integration between product managers, designers, engineers, and data scientists from the earliest stages.

• Rapid iteration with user feedback is essential — The path to building valuable AI products involves continuous testing, learning, and refining based on real user interactions rather than theoretical assumptions about what might work.

• AI products should augment human capabilities — The best AI implementations enhance human strengths rather than attempting to replace them, creating a complementary relationship that produces better outcomes than either could achieve alone.

Expert Analysis

The most insightful takeaway from this conversation is the fundamental shift in product development methodology required for AI. Traditional software follows a relatively linear path from specification to implementation, but AI products demand a more experimental, hypothesis-driven approach. This distinction explains why many companies struggle with their AI initiatives despite having strong technical teams.

This matters tremendously in today's business landscape because organizations are investing billions in AI capabilities without necessarily adapting their product development processes. According to McKinsey, companies that successfully implement AI can expect a 40% increase in productivity, but those gains remain theoretical without the right development approach. The "build it and they will come" mentality simply doesn't work with AI products, which require continuous refinement based on user interaction patterns.

Beyond the Discussion

What the conversation doesn't fully explore is how organizational structure impacts AI product success. Companies like Microsoft have reorganized entire divisions to better support AI initiatives, creating cross-functional pods rather than maintaining traditional siloed departments. This