LinkedIn

Tomer Cohen, LinkedIn's Chief Product Officer, noticed a crisis brewing—not in product, but in how product gets built. LinkedIn's data showed that 70% of skills required to do any job will change by 2030, yet the traditional product development process at LinkedIn had become glacially slow. What once took one builder—research, design, spec, code, launch, iterate—had fragmented into 15+ sources of information, design reviews, privacy reviews, security reviews, and layers of micro-specialized functions. A simple feature now required multiple teams, multiple codebases, and multiple sprints just to ship. "The work itself is not complex," Tomer realized, "but the process we made very complex."

Instead of optimizing the broken system, Tomer started from first principles. The Full Stack Builder model is built on three pillars: **platform**, **tools**, and **culture**. The platform work involved re-architecting LinkedIn's core systems so AI could reason over them—working closely with tool makers like Cursor, Figma, and others to customize their agents for LinkedIn's unique codebase and design systems. Off-the-shelf integrations didn't work; they required deep customization and alpha-mode partnerships. The tools pillar focused on building specialized agents, each owning one job: a trust agent (trained on LinkedIn's unique trust vectors), a growth agent (funneling past loops and tests), a research agent (trained on member personas and support tickets), and an analyst agent (querying the LinkedIn graph). Each agent was trained on carefully curated, gold-example datasets—not just raw access to everything. The coding pipeline already had progress (50% of failed builds fixed by the maintenance agent), but the idea-to-design phase was untouched until now.

Rather than mandate adoption, Tomer seeded the program through early adopter pods—small cross-functional teams (designer, PM, engineer) who got early access to the agents in exchange for feedback. The semantic search teams used these tools to build dashboards without waiting for design resources. A head of partnerships built a developer portal himself. A user researcher used the full stack builder tools to become a growth PM—a career transition that would have been impossible in the traditional model. These weren't forced changes; they were top performers gravitating toward the tools naturally. The program also launched the Associate Product Builder program, replacing the traditional APM curriculum with full-stack training in coding, design, and product.

The biggest learnings came from the culture piece. Tomer discovered that tools alone don't drive adoption—you need incentives, examples, and permission. Performance reviews now include 360 feedback from cross-functional peers, signaling that full-stack thinking is valued. Celebrating wins in all-hands meetings—like the user researcher who became a growth PM—creates FOMO and legitimacy. Different teams gravitated to different design agents (Figma vs. Subframe vs. Magic Patterns), forcing a trade-off between flexibility and standardization. And the data preparation was unexpectedly labor-intensive; reasoning over raw knowledge bases doesn't work—curating golden examples and providing weighted context took weeks. Early on, simply rolling out access didn't work; "I see a lot of companies roll out their agents and just expecting companies to adopt. It doesn't work this way," Tomer noted. The top 5% of talent adopted immediately, but scaling required deliberate change management.

After 4–5 months of dedicated work (plus months of setup), the pilot is live with pods across multiple teams. Teams are reporting saving hours of work per week through the analyst, prototyping, and product-jamming agents. Quality of insights and product decisions has improved visibly. The next-generation APB program launches in January, sending a cultural signal that full-stack building is the future career path. A full internal GA rollout is planned within a couple of months. The biggest insight: AI amplifies the best, not the average. Top performers are using the tools most and getting the best results. "Top talent has this tendency of continuously trying to get better at their craft," Tomer observed, and the Full Stack Builder model gives them the leverage to do it. The vision is clear—in a world where change outpaces response, nimble, cross-skilled pods of builders (like Navy SEALs, he noted) will be the organizations that win.