Product Tribe: The gap between AI product rhetoric and what teams actually do

Issue 245 of the Product Tribe newsletter published an analysis by Jakub Sirocki examining a gap the AI product discourse largely ignores: the difference between claiming to use AI and actually using it for the decisions that determine product quality.

The data point

According to the Figma AI Report 2025, only 31% of designers and product teams use AI for core design and product work. The remaining 69% use it for peripheral tasks — generating placeholder content, writing copy, handling administrative output — while core product decisions, discovery, and strategic framing remain largely untouched by AI.

The argument

Sirocki’s interpretation is not that adoption is slow or that teams are resistant. The argument is more specific: teams that skip research and discovery to jump straight into AI generation produce generic, error-prone outputs that lack strategic grounding. The speed gain is real but the quality loss is also real, and it typically shows up later when generated output does not hold up against user testing or stakeholder review.

The article also makes a point about what the pre-AI numbers reveal. If a team spent 60 to 70 percent of its time in a design tool before AI became available, that reflected a process problem, not a design productivity ceiling. AI can accelerate the same flawed process without fixing the underlying issue. The bottleneck was usually not how fast screens were drawn but how clearly the problem was understood before drawing began.

What this means for product managers

The practical implication is about sequencing. AI-assisted generation works best after discovery has produced clear requirements, constraints, and a defined user problem. Using it earlier — to generate personas, synthesize research, or frame problem statements — risks producing the appearance of rigor without the substance.

The 31% figure is also a reminder that adoption statistics about AI in product work tend to measure access rather than depth of use. When evaluating whether a team is genuinely AI-augmented, the question is not “does the team use AI” but “does AI touch the decisions that actually determine product quality.” For most teams in 2026, the honest answer is still no.