Key Moments
Roles aren't converging—they're expanding | Tamar Yehoshua (Atlassian CPO)
Lenny's PodcastKey Moments
Product managers are becoming 'AI builders,' expanding their roles beyond traditional tasks to include coding and prototyping, accelerating product development cycles significantly.
Key Insights
In the AI builder era, product managers are increasingly writing code, with one PM submitting 26 pull requests in a month for Confluence's Remix feature, surpassing many engineers.
Atlassian's Confluence Remix and Slides features were launched in 6 weeks and 8 weeks respectively, a significant acceleration from the previous estimated 6 months for similar AI-dependent enterprise software.
The 'Robo Claw' project, built from scratch, saw product managers initially contribute code but later shifted focus to strategic direction and unblocking engineers, leading to faster progress.
For large, complex software like Jira, PMs are not directly contributing code to the main codebase but are leveraging AI tools to automate tasks like prototyping, feedback classification, and bug fixing, leading to 3x productivity.
Atlassian's 'AI Mastery Index' assesses PMs on a scale of 1-5 across six AI capabilities, aiming for most to reach a level 3 proficiency, with an expectation to excel (level 5) in areas most critical to their team.
Over 1,000 employees have participated in Atlassian's 'AI Builder Weeks,' creating more than 120 new workflows and focusing on specific skills like prototyping, evals, and bot building.
Product management roles are expanding, not disappearing, with the rise of 'AI builders'.
The narrative around job displacement due to AI has shifted from 'your job will disappear' to 'roles are converging.' In large companies like Atlassian, this convergence is manifesting as an expansion of responsibilities, particularly for product managers (PMs). The concept of the 'AI builder' is prevalent, where individuals, including PMs, are leveraging AI tools to accelerate development and take on tasks previously outside their purview. This shift means PMs are now expected to do more with less, enhancing their effectiveness and ability to deliver value to customers.
AI tools are automating traditional PM tasks, freeing up time for higher-impact activities.
Historically, PMs spent time on administrative tasks like taking meeting notes, tracking action items, and coordinating with marketing. AI tools, however, can now automate many of these functions. At Atlassian, the CEO can get instant updates on feature releases via their AI product, Rovo, demonstrating the accessibility of information. This automation allows PMs to focus on core responsibilities such as identifying product-market fit, building beloved products, and ensuring business viability. The key change is not the 'what' of product management, but the 'how,' with AI tools becoming integral to this process.
Product managers are increasingly writing code and building prototypes to accelerate development.
In product development, the metaphor of rowing versus steering is used to describe the PM's role. For Atlassian's Confluence, a new feature called 'Remix' was launched in just 6 weeks, a drastic reduction from the estimated 6 months it would have taken previously. This acceleration was partly due to the PM taking on coding tasks. For instance, a PM with no prior coding experience partnered with an engineer to build a front-end helper tool, enabling her to submit 26 pull requests in one month. This allowed engineers to focus on higher-level activities. Similarly, AI was used to automate design bug fixes, resolving 14 issues in an hour. This hands-on approach, where PMs are 'rowing' by contributing code, is crucial for speeding up development cycles, especially for enterprise software that requires compliance and security.
AI accelerates product development through rapid prototyping and iterative feedback loops.
The development of Confluence Slides, which turns a page into a presentation in seconds, was launched in 8 weeks, another testament to AI-driven acceleration. The product team embraced a radically different approach, prioritizing speed. They used AI for tasks like prompt evaluation and design-to-code automation. For example, linking Figma designs to actual code and automating fixes reduced development time significantly. For Jira, a 20-year-old product with a massive user base, the goal was to make it AI-first. This involved building AI features that helped customers use AI in their own software development. PMs utilized tools like Loom for rapid prototyping, where UI changes or Figma designs with voiceovers could generate actionable items that automatically triggered a coding agent. This streamlined the process, generating compliant code in the Atlassian design language and significantly speeding up the handover to developers.
The role of the PM shifts between 'rowing' and 'steering' based on project stage and type.
In the 'Robo Claw' project, initiated from scratch, the PM, Josh, and a designer initially built a prototype based on intuition. Once validated, engineers were added. Josh initially coded but realized his time was better spent on strategic direction and unblocking the team. He shifted from 'rowing' to 'steering,' which accelerated progress. His prior coding experience proved invaluable in understanding challenges and guiding the team more effectively. AI was also used to automate weekly reports, freeing him from manual tasks. This adaptability, moving between hands-on coding and strategic leadership, is key to maximizing impact. For Jira, a complex legacy system, PMs focused on 'steering' rather than direct coding due to the risks involved. They drove the AI-first strategy, aiming to embed AI capabilities into the software development lifecycle, significantly boosting productivity and releasing 22 user-facing features in 10 weeks.
AI empowers PMs to process vast amounts of feedback and insights efficiently.
Managing feedback for a product like Jira, with hundreds of thousands of customers, is a monumental task. Atlassian used AI agents to classify and sort feedback and bug reports received via Slack, automatically sending them to coding agents for fixes. This saved considerable engineering time. For user research, AI was employed to review over 900 customer insight videos from Jira Service Management. An AI agent classified and organized these insights, providing the engineering team with precise information on what needed to be fixed. This AI-driven approach to feedback and insight management allows PMs to act as clear navigators, ensuring that feedback is processed effectively and that reusable prototypes are built, ultimately accelerating project delivery.
Atlassian invests in AI upskilling its product managers through structured programs.
To ensure PMs can effectively leverage AI, Atlassian introduced the 'AI Mastery Index,' a framework to guide skill development across six capabilities, including tool usage, prompt writing, data insight automation, prototyping, and technical knowledge. This index is not a career ladder but a compass for growth, with an aim for all PMs to reach a level 3 proficiency and specialize (level 5) in critical areas. The company also runs 'AI Builder Weeks' quarterly, dedicating a week for PMs and designers to learn from external speakers and internal experts, then apply these skills to a project. Over 1,000 employees have participated, creating over 120 new workflows, focusing on skills like prototyping, evals, and bot building.
Measuring the impact of AI on productivity remains a challenge, but key metrics are emerging.
While the benefits of AI in product development are evident, quantifying its impact on productivity is still an evolving area. Atlassian, like many organizations, is exploring ways to measure this. Current metrics include pull requests merged into production (not just written), features delivered from idea to customer usage, and overall OKRs. The focus is on measuring both individual team speed and organizational velocity. While there isn't a perfect solution yet, Atlassian continuously experiments with different measurement approaches each quarter to better understand the results driven by AI adoption. The speaker invites insights on effective measurement strategies, highlighting that this remains an ongoing challenge.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Concepts
●People Referenced
Product Manager's AI Integration Cheat Sheet
Practical takeaways from this episode
Do This
Avoid This
AI Impact on Feature Development Timelines
Data extracted from this episode
| Scenario | Time Before AI | Time With AI |
|---|---|---|
| Confluence Feature (Remix/Slides) | Approx. 6 months | 6 weeks |
| New Feature in Existing Codebase | N/A (implied longer) | 6 weeks (Remix/Slides) |
| New Product from Scratch (RoboClaw) | N/A (estimated) | Initial MVP with Product/Design focus, then scaled with engineers |
| Jira AI Features | N/A (implied longer) | 10 weeks for 22 features (3x productivity) |
Common Questions
An 'AI builder' is a modern job title emerging in the tech industry, particularly in startups. These individuals are focused on developing and implementing AI solutions, often wearing multiple hats and taking on a broad range of responsibilities to bring AI products to life.
Topics
Mentioned in this video
A tool used for creating video recordings of UI changes or designs, which can then be used to generate actionable items for AI agents.
A long-standing Atlassian product that has undergone significant AI integration for its development lifecycle.
An Atlassian product where new AI-powered features like Remix and Slides were launched.
Communication platform used to receive feedback and errors, which are then processed by AI agents.
More from Lenny's Podcast
View all 56 summaries
29 minWhy Claude can’t be your PM (yet) | Anthropic CPO Panel
30 minStop planning for 2027: how OpenAI builds product 90 days at a time | Tara Sesha and Nan Yu (OpenAI)
95 minMolly Graham: The grief, burnout, and opportunity hiding inside the AI transition
23 minMarty Cagan: Strong Opinions, loosely held
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free