OpenAI's Tara Seshan on the Rise of AI Coworkers

Seshan argues that designing for the current state of AI models is a guaranteed path to failure, as is attempting to predict model capabilities a year in advance. The only successful approach involves a pragmatic, short-term build cycle that accounts for rapid advancements.

In AI product management, prolific and empirical work significantly outperforms lengthy theoretical documents. The primary objective is to develop a testable prototype as quickly as possible to gather user feedback and iterate. This approach requires product managers to embrace a more ambitious and action-oriented mindset, prioritizing hands-on experimentation over abstract reasoning.

Tara Seshan envisions a future where work is less about manual execution and more about strategic direction. AI agents will handle the 'rowing' or the execution of tasks, allowing humans to focus on 'steering' the overall direction and making high-level decisions. The level of abstraction for this 'steering' role will continuously rise, demanding that humans provide intuition and make opinionated calls that autonomous agents cannot yet replicate.

AI agents are rapidly evolving to handle higher levels of abstraction, shifting towards a persistent, coworker-style collaboration model. This transition means interactions with AI are moving beyond one-off requests to sustained engagement within multi-agent team environments. Seshan suggests this co-worker model represents a more natural and effective interface for human-agent collaboration, facilitating deeper integration into daily workflows.

Essential components include local agent data access, robust cloud infrastructure, and strong reliability pieces. Seshan emphasizes that effective, non-isolated agents require a solid foundation of data connectivity and dependable systems, making these practical elements as vital as raw intelligence.

The most effective users of AI tools don't merely automate routine tasks; they leverage AI to significantly expand their capabilities across diverse fields like design, engineering, and product management. What once distinguished a 'unicorn' is becoming the baseline for the AI-augmented worker, with ambition emerging as the primary differentiator as simple tasks become trivialized.

These tools have dramatically raised the ceiling on what a small team can achieve, moving the goal from standard execution to 'unreasonably' ambitious outcomes. It's the PM's job to challenge conventional thinking and encourage teams to aim higher.

Seshan warns that building for the current state of models is a failure, as is attempting to predict capabilities a year out. The optimal strategy involves targeting model capabilities projected 2 to 3 months into the future, designing the core product to flexibly accommodate inevitable and rapid improvements.

In the pre-AI era, product perfection and polish were often seen as paramount. However, the current AI product environment prioritizes speed and conviction over meticulous perfection. Seshan argues that getting a transformative product into users' hands early, even if imperfect, is far more valuable due to the rapid pace of technological advancement. Early deployment allows for quicker feedback and iteration, leveraging the inherent transformative nature of AI technology.

This emphasis on fluid collaboration challenges traditional craft-based roles, requiring individuals to adapt and contribute wherever needed to ensure overall success.

Despite AI's advancements, human brains remain essential for accountability, expression, and team cohesion. Humans provide the ultimate ownership of outcomes, particularly in regulated industries. The artistry and opinionation required in software creation are aspects AI cannot yet autonomously replicate. Furthermore, fostering team enthusiasm and maintaining interpersonal relationships within a workplace remain uniquely human responsibilities, highlighting the irreplaceable role of human connection.

The process of outlining, editing, and refining ideas manually is critical for clarifying one's thoughts and should remain a human-led effort, as outsourcing it risks losing the depth of personal understanding.

This method encourages feedback by presenting ideas with rough edges, making it easier for colleagues to contribute and shape the final product. Perfectly polished ideas, conversely, can inadvertently discourage feedback, leading to less collaborative and potentially less robust outcomes.

While product-market fit is widely recognized, Seshan argues that 'product marketing fit' is significantly underrated. This concept involves defining how a product is discussed and marketed, often even before it's built. Testing the narrative and positioning through pitches and communication is treated as a necessary prerequisite to product development, ensuring that the story resonates before significant resources are invested in building.

Seshan uses the analogy of rereading the same book at different life stages to illustrate how personal growth changes one's perspective. Each reread reveals new themes and understandings that were previously missed, from basic plot points to complex historical or human dynamics. This flexibility in interpreting experiences is crucial not only for personal development but also for managing one's career and adapting to new challenges.

Reflecting on modern tools versus historical constraints, Seshan compares the capabilities of an iPhone today to the resources available to a filmmaker like Kurosawa when creating 'Rashomon.' With vastly more powerful tools at hand, creativity is no longer limited by a lack of resources. This abundance of technology should prompt individuals to question their own excuses for not elevating their ambitions and striving to create better, more impactful work.