Assumptions as code: SiriusXM’s approach to platform prioritization

This episode begins with introductions of Eleanor Millman and Mina Tawadrous and their roles within SiriusXM. They discuss the challenges faced by platform engineering teams, particularly in scaling prioritization across multiple teams with limited resources. The importance of aligning technical priorities with broader organizational goals and the unique context of platform engineering—where value isn't directly tied to revenue—is emphasized. SiriusXM's efforts to develop a more formalized prioritization framework that incorporates multiple impact factors such as developer speed, reliability, cost, and trust are introduced.

The discussion highlights the transition from traditional models like RICE (Reach, Impact, Confidence, Effort) to a custom, multi-factor prioritization system tailored to SiriusXM's needs. The team iteratively refined their approach, incorporating data and leadership signals to weight factors differently depending on strategic goals. They also emphasized the importance of quantitative data and data-driven decision-making to manage complex, cross-team projects effectively, thus enabling better resource allocation and stakeholder alignment.

A core part of the episode is the innovative concept of 'assumptions as code.' This approach involves storing and validating assumptions within a central repository, making them visible, conflict-resolvable, and up-to-date. It facilitates transparency and collaborative resolution of assumption conflicts, which often exist in fragmented communication. The process includes using AI tools to surface relevant assumptions and validate or invalidate them by engaging team members, thus reducing misalignment and stale assumptions.

AI plays a crucial role in SiriusXM's framework by assisting in the recall and validation of assumptions, significantly scaling their decision-making process. The AI tools help surface existing assumptions, validate new inputs, and generate PRs, making the process more efficient and scalable—particularly beneficial for smaller initiatives. This integration ensures that decision-making is informed by current, validated data, reducing manual effort and increasing accuracy.

Both panelists discuss the shift towards empowering 'builders' beyond traditional developers, including non-technical contributors or citizen developers. They address challenges like security, maintainability, and support, but also explore strategies to enable wider participation without increasing risks. The conversation includes plans for SiriusXM’s 2026 roadmap, prioritizing discovery of existing tools, reducing friction in contributions, enhancing code review workflows, and further integrating AI for onboarding, documentation, and scaling efforts.