💡 Top tip — Read the TL;DR below to grasp the core transformation, then dive into the sections that matter most for your team's AI journey.

Tomer Cohen (LinkedIn CPO), "Why AI is disrupting traditional product management"

EXPLORE THE FULL SUMMARY 👇

LinkedIn isn't tweaking product development—they're rebuilding it from first principles. The full stack builder model centres on a provocative idea: what if anyone, regardless of their original function, could take products from idea to launch?

This isn't theoretical. A user researcher on their team saw an open PM role on the growth team. She used LinkedIn's AI tools to build the necessary skills and successfully transitioned into the role. That career path - researcher to growth PM - wasn't even considered possible under traditional structures.

The programme replaces LinkedIn's Associate Product Manager (APM) track with the Associate Full Stack Builder programme. Engineers, designers, researchers, marketers—all can become builders if they demonstrate the capability. The boundaries blur: a designer might analyse growth data using the analyst agent, whilst a PM generates design variations using the design agent.

When LinkedIn's head of partnerships personally built a product integration instead of delegating to his team, it sent a powerful message: this is possible for everyone.

Key Takeaways:

Whilst AI transforms execution, certain capabilities remain distinctly human. Cohen identifies five critical traits that builders must possess: vision, empathy, communication, creativity, and judgment.

Vision means profound understanding of unmet needs—seeing possibilities beyond the obvious. Empathy allows builders to align and rally others around ideas. Communication remains critical across almost every role. Creativity generates possibilities beyond what AI suggests—Cohen notes AI "brings back things you might not know about" but doesn't yet excel at next-level creative thinking.

But judgment stands supreme. Cohen calls it "the most important trait for a builder"—making high-quality decisions in complex, ambiguous situations. Some call it taste; he calls it judgment.

Everything else? "I'm working really, really hard to automate," Cohen states. The goal isn't making humans do more—it's making organisations more nimble, adaptive, and resilient. Matching the pace of change to the pace of response.

Key Takeaways:

LinkedIn didn't just give teams access to ChatGPT Enterprise. They built custom AI agents trained on LinkedIn's specific data, workflows, and member personas. These agents actively collaborate with teams and critique ideas.

The research agent trains on LinkedIn member personas (small business owners, job seekers, recruiters) using past research, support tickets, and world knowledge. When a team presented a marketing spec for small business owners, Cohen asked the research agent to critique it. The feedback was so insightful it shifted the team's entire direction.

The growth agent helps teams identify which features have the biggest growth opportunity. The UXR team uses it to prioritise research based on potential impact, not urgency.

The analyst agent eliminates bottlenecks. Trained on how to query LinkedIn's massive graph, it removes data science dependencies. Queries that previously took days or weeks happen instantly.

The design agent remains a work in progress—different teams gravitated toward different design AI tools, creating standardisation challenges LinkedIn is still resolving.

Key Takeaways:

Individual AI agents provide value. But agents working together create exponential value. This is where LinkedIn invested heavily.

Cohen emphasises they've built an orchestrator enabling agents to collaborate. The research agent and growth agent can work together, going back and forth rather than operating sequentially. This agent-to-agent communication creates compound effects.

The orchestration layer required significant engineering investment - it's not something off-the-shelf tools provide. But the returns justify the effort. When the research agent identifies user needs and the growth agent simultaneously assesses feature impact, decisions accelerate whilst quality improves.

This mirrors how human teams work best - not sequential handoffs, but fluid collaboration. LinkedIn's betting that AI agents will collaborate the same way, with orchestration infrastructure enabling that fluidity.

Cohen notes this investment level separates superficial AI adoption from transformation. Point solutions help individuals. Orchestrated systems transform teams.

Key Takeaways:

LinkedIn fundamentally redesigned team structure. They've moved from large teams to "pods"—small groups of full stack builders assembled for specific missions, then reassembled.

Cohen uses the Navy Seals analogy deliberately. Seals cross-train across multiple areas but specialise in the mission, not the function. They operate in small pods, move quickly, and reassemble as needed.

LinkedIn assembles pods of full stack builders coming together. It's less about "can I have an engineer, designer, PM working together" and more about "folks who can flex across". They tackle something for a quarter, then LinkedIn reassembles those pods to different missions.

The goal: speed, adaptability, and flexibility. Teams that previously bloated now operate with surgical precision. This addresses what broke as teams grew—the ability to match the pace of change to the pace of response.

Cohen sees this as the organisational model that wins in the future: nimble, focused, continuously adapting.

Key Takeaways:

LinkedIn fundamentally changed what gets measured. Cross-functional capability now carries equal weight to functional depth.

If you're a PM, your designers rate your design thinking. Your engineers rate your technical understanding. This 360-degree evaluation across disciplines is now standard.

Cohen rolled this out methodically. First, his direct reports received cross-functional 360 feedback. If you came from a PM background, designers rated your design thinking—creating immediate motivation to develop those skills.

Hiring now explicitly looks for full stack builder traits. The upcoming performance cycle includes these cross-functional evaluations as formal review criteria. Cohen deliberately moved slowly to avoid diluting the programme.

The exclusivity matters. By making full stack builder a programme people aspire to join, LinkedIn created a movement. Cohen shares: "People want to feel like there's success stories... it feels like it's a movement they want to be part of."

Key Takeaways:

Simply giving teams AI agents doesn't drive adoption. LinkedIn learned that incentives, examples, and cultural shifts matter as much as tooling.

Leadership modelling proves critical. When executives build products themselves, it sends a powerful signal. Cohen emphasises these examples are "really, really powerful" for driving adoption. The head of partnerships building his own integration showed everyone: this is possible.

Celebrations matter. You need all-hands highlights of successful transitions, clear promotion signals, and cultural momentum. The exclusivity of the programme - not everyone qualifies immediately - creates aspiration rather than mandate fatigue.

Cohen mentions doing "sessions today with CPOs and COs on this process" - the investment upfront gets the gains after. Even learning tools like Cursor or Figma requires investing hours before velocity increases.

The lesson: change management deserves equal investment to tooling. LinkedIn's success stems from combining technology (custom agents), structure (performance updates), and culture (leadership modelling).

Key Takeaways:

If you lived through the mobile transformation, this feels familiar. Companies that required all product specs to include mobile mocks created a forcing function that changed how everyone worked.

The mobile shift wasn't just adding a channel - it required rethinking information architecture, interaction patterns, and core workflows. Similarly, the AI transformation isn't about adding tools. It's about reimagining what's possible when you remove traditional constraints.

Cohen draws this parallel deliberately. During mobile, top-down mandates weren't enough—you needed examples, training, and cultural shifts. The same holds true now. LinkedIn combines directive change (performance review updates), enabling tools (AI agents), and cultural modelling (leadership building products).

What's different: Mobile was primarily about platform. The AI transformation unlocks latent capabilities across your team. That user researcher who became a growth PM had the strategic thinking all along. AI tools removed the technical barriers.

One critical lesson from mobile: start now, don't wait for perfection. Companies that waited for mature mobile tools paid a steep price. Waiting for "mature" AI means competitors are already pulling ahead.

Key Takeaways:

Off-the-shelf tools are starting points. The biggest gains come from customisation and integration.

Building the research agent meant training it on all past research, support tickets, and LinkedIn member personas. This requires dedicated engineering resources and deep integration with your data systems.

The orchestration layer represents another major investment. Getting agents to work together fluidly required building custom infrastructure. Cohen notes: "this does require that level of investment."

Time horizon matters. Cohen describes current agents as "MVP++"—they're rolling out more broadly internally over the next couple of months. This isn't a quick implementation. It's a multi-quarter effort requiring patience and iteration.

Training AI on the "right data, not all the data" requires weeks of work. Cohen's example: filtering through professional posts on LinkedIn to create golden samples. You should be ready to invest those hours before seeing velocity and quality gains.

Unexpected use cases emerge—the UXR team's adoption of the growth agent surprised Cohen. This suggests tools enable use cases you can't predict upfront.

Key Takeaways:

Beyond tools and structure, Cohen emphasises a fundamental mindset change: fall in love with continuous growth, not reaching a fixed state.

He describes this as "progress mode, iteration mode—it's not about reaching a state, it's about the journey." This matters because AI capabilities evolve so rapidly that any "arrived" state becomes obsolete quickly.

The practical application: measure your year-over-year delta. How much did you grow? What skills did you gain? Cohen frames this as "version 2026 versus 2025—what's the delta there?"

This eliminates FOMO (fear of missing out) because there's no destination to miss. You're simply committed to continuous evolution. The comparison isn't against others' positions—it's against your own previous version.

Top talent "has this tendency of continuously trying to get better at their craft." The full stack builder model accelerates this by removing artificial barriers between disciplines.

Cohen explicitly states: "I have a bias for change. I have a bias for positioning myself in a place where I can learn the most." This philosophy extends to his career—leaving LinkedIn after 14 years, driven by growth orientation. Even successful long-term roles eventually constrain learning velocity.

Key Takeaways:

LinkedIn's full stack builder model represents one company's ambitious answer to AI transformation. Whilst not every organisation will replicate their exact approach, the underlying principles apply broadly.

Start with the mindset shift before the tools. If your team views AI as an addon rather than a fundamental capability multiplier, adoption will remain superficial.

Invest in change management as heavily as tooling. LinkedIn's success stems from combining technology (custom agents), structure (performance review updates), and culture (leadership modelling behaviour).

Accept that this transformation is messy and nonlinear. Different teams will gravitate toward different tools. Unexpected use cases will emerge.

Most importantly, recognise this is a years-long journey, not a quarter's initiative. The companies that will win aren't those with the best initial strategy—they're the ones that start now and commit to continuous evolution.

The future Tomer Cohen describes is already arriving. The only question is whether your team will shape it or be shaped by it.

I'm genuinely intrigued by this approach. On the one hand it could unlock huge opportunities. On the other hand it could create chaos. I also somewhat loathe the direction of travel LinkedIn has taken of late, which has jaded my opinion of this approach.

Love to understand / hear your views on this!!

That’s a wrap.

As always, the journey doesn't end here!

Please share and let us know what you liked or want changing! 🚀👋

Alastair 🍽️.