LinkedIn's Full Stack Builder Model: AI-Augmented Product Development

Deep Dive

The discussion begins by addressing the rapid pace of change in the job market, citing a projection that 70% of necessary job skills will change by 2030. This necessitates a reimagining of how product development occurs, moving beyond traditional, complex processes that have led to organizational bloat and slow feature cycles, often taking six months for a single feature. The source introduces the "Full Stack Builder" (FSB) model at LinkedIn as a radical new approach to product development that fully embraces AI capabilities.

The core of the Full Stack Builder model is to empower individuals to take an idea from conception to launch, irrespective of their traditional role or team. This model emphasizes a fluid interaction between humans and machines, aiming to collapse the development stack and return to craftsmanship. Key traits for these builders are identified as vision, empathy, communication, creativity, and judgment, with a strong emphasis on the latter. The goal is to increase experimentation volume and quality while reducing the time from idea to launch, making organizations more nimble, adaptive, and resilient.

The implementation of the FSB model involves three pillars: platform, tools and agents, and culture. Significant investment is required in re-architecting core platforms to be AI-ready, as off-the-shelf AI tools often do not work with enterprise codebases without extensive customization. This includes building composable UI components and ensuring that AI can reason effectively over the company's specific codebase and design systems.

Internal tools and specialized "agents" are being developed to automate tasks outside of the core builder traits. Examples include a trust agent that identifies vulnerabilities in product ideas, a growth agent that critiques ideas based on LinkedIn's unique growth processes, and a research agent trained on member personas and past research. An analyst agent is also being built to query the LinkedIn graph, reducing reliance on SQL queries and data science teams.

A significant learning has been that customization is crucial; third-party tools like Cursor, Devin, or Figma require adaptation to work with LinkedIn's specific environment. Different teams have gravitated towards different tools for similar functions, necessitating a strategy to converge on a manageable set of options. The focus has been on building specialized agents, with the intention of developing an orchestrator layer to allow these agents to work together seamlessly.

The investment in AI has heavily focused on the "idea to design" phase, as coding has already seen significant acceleration. While coding and maintenance agents are mature, the "idea to design" area, which is material to the quality of work and empowers everyone, has received more recent attention. The creation of these internal tools and agents has required substantial dedicated work, including gathering and cleaning relevant data, with a focus on providing AI with specific context rather than broad access to all company information.

The pilot program for the Full Stack Builder model is underway with a substantial part of the organization using the tools and providing feedback. Benefits observed include saving hours of work per week across various roles, improved quality of insights and discussions, and increased experimentation volume. Designers and PMs are reportedly picking up bugs directly from tickets, and there is a high demand for access to these tools.

A key insight from the pilot is that top performers are adopting AI tools the fastest, contrary to expectations of a leveling effect. These individuals have an innate drive to improve their craft and stay at the cutting edge. This highlights the critical importance of change management, as simply providing tools is insufficient; incentives, motivation, and examples of success are necessary for broader adoption.

Change management tactics involve celebrating wins, making tools exclusive to create a sense of urgency, and updating performance reviews to reflect the adoption of these new ways of working. The Associate Product Builder program is being launched to train individuals in coding, design, and PM skills together, replacing the previous Associate Product Manager program. This program will serve as a training ground and a model for broader organizational adoption.

It is emphasized that not everyone needs to become a full-stack builder; specialized roles will continue to exist, but there may be a reduced need for as many specialized individuals as in the past. The shift is towards a "full-stack mindset" rather than just a title. The advice for companies starting this journey is to focus on the platform, tools, and culture, with a strong emphasis on the latter. Patience and investment in these areas are crucial for long-term success, and over-communication of vision and progress is vital.

In a lightning round, Tomer Cohen recommends the books "Why Nations Fail," "Outlive," and "The Beginning of Infinity." He enjoys the Hebrew podcast "One Song" and a similar English podcast, "Song Exploder." A product he loves is the voice mode in ChatGPT, and he notes that Tesla's Grok feature in cars is a step in that direction. His life motto is "better than being," emphasizing continuous growth and evolution. He expresses pride and gratitude for his 14-year tenure at LinkedIn, viewing it as a mission aligned with his purpose, and is excited to explore new problem sets and areas of investment.