Product Teams Use AI To Solve Everyday Tasks
Product coaches Teresa Torres and Petra Wille released a June 30, 2026 episode of the Product Talk podcast arguing that most people underuse AI not because their problems are too small, but because they lack a habit of small, fast experiments. Torres described testing one to-do-list item with AI daily, and the pair recommend a 15-minute time-boxed session where the first results are expected to be bad. The episode advises ignoring peripheral tools like MCP servers or the newest plugins, focusing instead on a single task, and sharing results through community formats such as show-and-tell. Their suggested starter prompt: "I have to do this, how can you help?" For product and AI practitioners, the advice maps to a broader pattern in applied-AI tool adoption: narrow, repeatable experiments surface prompt patterns and failure modes faster than open-ended exploration.
For teams evaluating or onboarding AI tooling, the practical lesson isn't about a specific model or feature: it's a case for treating AI adoption as a habit-formation problem. Short, disposable experiments reveal a tool's real capability boundary faster than waiting for a large, well-defined use case, and they lower the switching cost of trying yet another plugin or agent.
What happened
In a June 30, 2026 episode of the Product Talk podcast, product coaches Teresa Torres and Petra Wille discussed the most common blocker people report when starting with AI: not knowing where to begin. Torres described her daily habit of testing one to-do-list item with AI, and Wille outlined a 15-minute time-boxing approach for trying a task even when the first results are poor. The hosts recommended filtering out noise such as MCP servers or the newest apps, and pointed to community formats -- show-and-tell sessions, YouTube, local meetups -- as faster learning accelerants than tool-chasing. Wille described iterating a MidJourney image from unusable to acceptable through repeated small attempts. Their suggested starting prompt: "I have to do this, how can you help?"
For practitioners
The episode's structure maps to established practices for applied-AI tool evaluation: pick a narrow, well-bounded task, define a simple success criterion, run short timed trials, and log which prompts and post-processing steps actually worked. That narrows the surface area enough to make a plugin, model, or agent's real strengths and failure modes legible within a single sitting, rather than after a multi-week integration effort.
What to watch
Whether this kind of lightweight, community-driven evaluation habit scales past individual practitioners into team-level processes -- shared prompt libraries or informal show-and-tell rituals inside product organizations -- is the more durable signal to track, more than any single tool the episode name-checks.
Key Points
- 1Teresa Torres and Petra Wille argue that a daily 15-minute AI experiment on one to-do item beats waiting for a big enough problem.
- 2The hosts advise ignoring new plugins and MCP servers and instead running narrow, timed trials to learn a tool's real limits fast.
- 3Community formats like show-and-tell and shared prompt logs accelerate learning more than chasing the newest AI app, the episode argues.
Scoring Rationale
Single-source practitioner advice from a well-known product-coaching podcast; useful but low-stakes and non-newsworthy beyond the two named hosts' own platform, consistent with a minor-tier score.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
