Building for current AI models is shortsighted, while building for a year ahead is too speculative. OpenAI's product teams target the anticipated model capabilities of the next quarter to stay relevant without being premature.
Unlike stable markets requiring deep theoretical strategy, the emergent nature of AI demands PMs shift from writing exhaustive docs to designing rapid, empirical tests focused on a single, core hypothesis.
As AI tools dramatically expand what's possible in a short time, a product manager's crucial function becomes reminding the team of this new reality and pushing them to pursue far more ambitious goals.
Unlike coding where output can be easily verified, AI for knowledge work must expose its reasoning process. Users need to trust *how* an answer was derived, not just the final output, making collaboration and transparency key product features.
Product teams at OpenAI are guided by three core questions: "Is this maximally accelerated?" for speed, "Are you mainlining it yet?" for intense dogfooding, and pushing for maximum ambition for scope.
Tara Seshan outlines a product evolution from simple chat interfaces (era one) to task-based agents (era two), culminating in a third era of persistent AI coworkers that collaborate with users and teams over time.
OpenAI's culture empowers every employee to act as a founder within their domain, fostering high ownership and speed with minimal top-down direction. This is a subtle but powerful departure from the typical founder-led model.
Tara Seshan suggests a disciplined approach to using AI for writing. Outsource tasks like summaries and status updates ('reporting'), but manually author strategy docs ('thinking') because the writing process itself is crucial for idea clarification.
Long strategy documents are no longer credible proof of thought, as AI can generate them. Tara Seshan now prioritizes interactive mocks, prototypes, and A/B test results as more effective tools for communicating and validating ideas.
Tara Seshan's time at Sutter Hill revealed its secret: testing a product's narrative and positioning ('product marketing fit') with hundreds of prospects is critical and should happen *before* committing to a specific product shape.
Tara Seshan posits that as AI agents increasingly handle tactical execution ('rowing'), the human role will evolve to focus on higher-level direction setting, feedback, and making opinionated judgment calls ('steering').