How Every Scaled Operations with Agentic Workflows
the gist
The Every team explains how they use an 'AI sandwich' workflow—framing goals and reviewing output—to automate complex growth engineering and marketing tasks, allowing them to ship faster with fewer manual hours.
The 'AI Sandwich' Workflow
Every’s team describes their operational philosophy as the "AI sandwich." Humans occupy the top and bottom layers: defining the problem, setting the strategy, and performing the final review. The middle layer—execution—is delegated entirely to agentic workflows. By treating AI as a collaborator rather than a tool, the team shifts from manual implementation to high-level orchestration, allowing them to focus on creative strategy while agents handle the repetitive "fake work" of data segmentation, email drafting, and pipeline management.
Automating Growth Engineering
The team demonstrates how they move from idea to execution in minutes. In one instance, a growth idea proposed in Slack was converted into a fully segmented email campaign by an agent (Codex) while the human lead went to the gym. The agent handled audience segmentation, drafted personalized emails based on historical performance, generated social imagery, and scheduled the sends. This approach allows engineers like Yash Poojary to automate A/B testing pipelines, moving away from manual dashboard configuration to focus on higher-level questions like user psychology and conversion strategy.
Tooling and Stack Integration
The team emphasizes the importance of interoperability between tools. They rely on a stack that includes Claude, Codex, Cursor, PostHog, Framer, and Descript. A key insight is the use of Model Context Protocol (MCP) to allow these tools to communicate. For example, they use agents to reorganize Notion databases, manage video editing workflows via Descript, and spin up cloud servers for testing, effectively offloading maintenance tasks that previously caused team burnout.
Lowering the Barrier to Entry
Every launched their "All Access" membership and "Builder Pack" to address the cost and accessibility gap for individual builders. They argue that the current era of software development requires a specific, expensive stack of AI tools to remain competitive. By providing credits and access to their internal stack, they aim to help solo builders adopt the same agent-first mindset, enabling them to operate with the output capacity of a much larger team.