Karri Saarinen: Beyond AI Hype to Product Insight

Saarinen intentionally stopped tracking AI model updates and new tools for a season to gain perspective on the core difficulties of product development, a move he noted allowed him to refocus on the challenges of building great products.

Upon his return, he realized that despite the emergence of new models and agents, the fundamental challenge of crafting exceptional products remains inherently difficult, leading him to question whether the industry's intense focus on new AI tools genuinely translates into improved products or enhanced team performance.

The tech industry has historically prioritized scaling operations and optimizing output by implementing specialized roles, refined processes, and data-driven experimental approaches.

Saarinen warns against the risk of companies becoming 'software factories,' where critical thinking and decision-making are outsourced to automated processes or external entities.

He stresses that customers do not purchase lines of code or experiments, but rather complete products that meet their needs and desires, distinguishing organizational efficiency from true product value.

Building products inherently yields two distinct outcomes: the tangible product itself and the invaluable knowledge acquired by the team throughout the development process.

The demanding process of design—encompassing questioning, engaging with customers, and debating potential solutions—is the primary arena where significant learning and deep understanding are cultivated.

Saarinen suggests that if teams bypass the struggle of creation through excessive automation, they risk losing the profound understanding essential for developing great products.

A team's capability is fundamentally defined by its collective understanding of its operational space, inherent talent, cultivated taste, and sound judgment.

Leading companies sustain their competitive advantage by fostering a culture of shared knowledge that delineates what truly matters and what does not.

This institutional knowledge encompasses a comprehensive history of past decisions, critical customer insights, and an acute awareness of technological limitations.

Formula 1 steering wheels serve as an example, evolving from traditional designs into highly specialized tools, a transformation driven by decades of user experience and iterative refinement.

The unique demands of high-speed racing necessitate specific fine motor movements, leading to a design that diverges significantly from standard car steering wheels.

Saarinen warns that excessive AI automation risks severing the link between execution and the crucial learning process, potentially eroding a company's long-term competitive advantage.

Linear automates bug investigation by integrating Sentry with other internal tools, allowing AI to diagnose code issues and significantly reduce the time engineers spend on these tasks.

Engineers are encouraged to utilize this saved time to engage directly with customers through platforms like Slack channels and various feedback loops, fostering closer relationships.

The overarching goal is to transcend mere execution, enabling teams to concentrate on developing superior judgment and enhancing overall product quality through more meaningful interactions.

Many individuals tend to concentrate on what AI can execute rather than what it can impart to the team in terms of knowledge and understanding.

Linear harnesses AI to consolidate diverse feedback streams, including insights from sales calls, support emails, and internal discussions, transforming AI into a central repository for institutional context.

Saarinen observes that product teams within large organizations frequently lose direct contact with customer problems due to the overwhelming volume of information they encounter.

To counteract this, he implemented an AI 'watcher' agent designed to aggregate customer feedback on a daily basis.

This AI generates a daily briefing specifically on AI workflows, providing him with insights into diverse customer usage patterns without requiring extensive personal time investment.

The process is instrumental in identifying evolving best practices in a nascent field where established standard procedures have not yet emerged.

Saarinen noted that increased AI usage could inadvertently lead to employees working in silos, consequently diminishing shared learning opportunities across teams.

'Quality Wednesday' is a mandatory weekly initiative where team members dedicate time to identify and resolve one quality defect within the product.

The primary objective extends beyond merely fixing the defect, which could be a minor task like correcting a janky animation or a copy error, to actively training employees to spot imperfections.

Sharing these findings in a subsequent meeting allows all team members to collectively enhance their attention to detail and learn from one another's discoveries.

A 'feature roast' is an optional meeting where the team developing a new feature invites other colleagues to provide candid and unfiltered critiques.

The main objective is to gather raw, impersonal feedback, allowing the feature to be viewed through the fresh eyes of someone encountering it for the first time.

If colleagues express confusion about a feature during the roast, it serves as a strong indicator that actual users will likely experience similar difficulties.

Leads then synthesize this feedback into actionable issues, and the collaborative discussion helps employees internalize the team's shared quality standards.

Rather than treating AI solely as a tool for increased output, Saarinen suggests that product organizations should reallocate the time saved through efficiency toward deeper exploration and critical reflection.

This shift encourages product organizations to advance towards more abstract thinking and foster team-wide learning, rather than solely concentrating on output.

Prioritizing customer feedback, engaging in rigorous critiques, and maintaining a steadfast focus on quality are crucial for achieving superior long-term product outcomes.

Companies require a centralized platform to systematically store critical customer context and product development insights.

This shared knowledge repository benefits both human team members and integrated AI agents, creating a unified understanding.

AI agents can serve as an intuitive interface, allowing teams to query customer feedback and track the corresponding team responses effectively.

Automation should be judiciously reserved for tasks that are both repeatable and have minimal impact on core product differentiation.

Saarinen defines intuition not as a mystical force, but as the tangible accumulation of training derived from experience and continuous feedback.

This personal understanding, cultivated over time, progressively compounds into a sophisticated, team-wide collective intelligence, enhancing overall decision-making.

The ultimate objective is to deeply comprehend the underlying forces that drive software creation: human judgment and the decisions made by individuals.

Hiring practices should therefore prioritize a candidate's trajectory, innovative ideas, sound judgment, and refined taste, rather than solely focusing on their raw output.

Product management's role should evolve from merely overseeing code production to actively curating and enriching the context within which products are developed, emphasizing understanding over raw production metrics.