GitHub Copilot has introduced stacked sessions and pull requests, a feature that lets developers chain multiple AI-assisted coding tasks together in a single workflow. Instead of starting fresh with each new session, developers can now build one task on top of another, with the AI maintaining context and creating dependent pull requests that stack on top of each other.

What Are Stacked Sessions and Why Do They Matter?

A stacked session is a series of tasks in the same repository where each session builds off the previous one. When you complete one session and create a pull request, you can immediately start a new session that branches from that work, rather than having to manually integrate changes or lose context. The GitHub Copilot app automatically creates what's called a "stack," which is a series of pull requests in the same repository where each pull request targets the branch of the pull request below it, forming an ordered chain that ultimately lands on your main branch.

This matters because software modernization projects often involve multiple interconnected tasks. Cassidy Williams, Senior Director for Developer Advocacy at GitHub, shared a real-world example: she spent over a decade neglecting a personal project built with React 15, Less CSS preprocessing, and outdated react-bootstrap components. Attempting this modernization manually would have taken weeks. Instead, she used the GitHub Copilot app to create a plan, execute the first phase of styling updates, discover compatibility issues, pivot to a different branch, and then stack a second session on top to handle react-bootstrap replacement, all without losing progress or context.

How Does the Stacking Workflow Reduce Developer Friction?

Traditional development workflows force developers to choose between two bad options: either make one massive pull request that handles everything at once, or manually manage multiple branches and risk losing context between sessions. Stacked sessions eliminate that friction by automating the dependency chain. When Williams realized her initial modernization plan conflicted with an older development branch she'd partially updated, she could simply ask the Copilot app to create a new session targeting the correct branch. The app closed the previous pull request, ported her styling decisions to the new branch, and created a fresh session without requiring manual rebasing or context loss.

The feature also combats scope creep, a common problem in AI-assisted development. Because developers don't have to write all the code themselves, it's tempting to expand a single pull request into a massive 10,000-line changeset that handles everything at once. Stacked sessions encourage breaking work into smaller, reviewable chunks by making it trivial to create dependent tasks. Williams noted that this approach felt "magical" compared to her previous experience with manual dependency management.

Steps to Implement Stacked Sessions in Your Workflow

• Start with a comprehensive plan: Use the GitHub Copilot app's Plan mode to outline your entire project, including dependencies, constraints, and decision points before diving into code changes.
• Create your first session and pull request: Execute the initial phase of work, test it thoroughly, and create a pull request once you're satisfied with the results.
• Stack a new session on top: When you're ready to tackle the next phase, ask the Copilot app to create a new session that branches from your existing pull request, rather than starting fresh from main.
• Let the app manage dependencies: The Copilot app automatically creates stacked pull requests that target the branch of the pull request below them, maintaining the chain without manual intervention.
• Review and merge incrementally: Each stacked pull request can be reviewed and merged independently, reducing the risk of large, hard-to-review changesets.

What's the Current State of Enterprise Copilot Adoption?

While stacked sessions represent a significant productivity improvement for individual developers, enterprise adoption of copilot tools remains cautious. Developers accept only 33% of GitHub Copilot's suggestions, even in mature deployments, according to recent data. MIT NANDA research found that approximately 95% of generative AI pilots show no measurable profit-and-loss impact, suggesting that many organizations struggle to translate AI capabilities into business value.

Forrester research indicates that most enterprises are still testing copilots in narrow, scoped pilots rather than rolling them out broadly. The challenge isn't typically the underlying AI model's capability, but rather the quality and accessibility of the business context the model can see. An enterprise copilot is only as good as the data it can access, and many organizations lack the curated, access-controlled business data needed to ground these tools effectively.

How Is Microsoft Consolidating Its Copilot Strategy?

Microsoft is moving to simplify its fragmented copilot ecosystem by combining Chat, Cowork, and Code into a single unified application, scheduled to launch by the end of summer 2026. The company currently maintains approximately 80 products bearing the Copilot name, creating confusion about which tool serves which purpose. The new super-app will let users switch between three main modes within a single interface: Chat for routine queries, Cowork for lengthy multi-step tasks, and Code for programming tools.

The unified Copilot may also feature a toggle between personal and corporate Microsoft 365 Copilot accounts, allowing users to switch contexts without leaving the application. This consolidation reflects Microsoft's recognition that fragmentation has become a barrier to adoption. In March 2026, the company merged the consumer and enterprise divisions of Copilot under unified leadership to address this issue.

For developers using GitHub Copilot specifically, stacked sessions represent a tangible productivity gain in the near term, while the broader enterprise copilot market continues to mature. The feature addresses a real pain point in complex modernization work, where maintaining context across multiple dependent tasks has historically required significant manual effort. As more organizations move beyond pilot programs and into broader deployment, features like stacked sessions may become essential for managing the scope and complexity of AI-assisted development at scale.