Cocoa: Co-Planning and Co-Execution with AI Agents

K. Feng·Kevin Pu·Matt Latzke...Joseph Chee Chang
CHI·2026·23 citations

TLDRCocoa is presented, a system that introduces a novel design pattern -- interactive plans -- for collaborating with an AI agent on complex, multi-step tasks and saw the interleaving of co-planning and co-execution as an effective novel paradigm for human-AI collaboration.

How do people cite this paper?

(generated 20 days ago)

Cocoa's interactive plans design pattern and its concept of steerability have been adopted by subsequent work designing notebook-style interfaces for AI-assisted writing and research, referenced as a representative approach to user control in agentic systems spanning co-planning, co-execution, and flexible delegation, and cited broadly as motivation for structured interaction frameworks beyond chat, mixed-initiative controls that promote balanced human-AI interaction, and GUI-first designs for single-agent systems, while also serving as a reference point for AI-assisted scientific research tools and for studying what makes LLM-generated plans helpful to users.

Loading PDF reader...

Cocoa: Co-Planning and Co-Execution with AI Agents

K. Feng·Kevin Pu·Matt Latzke...Joseph Chee Chang
CHI·2026·23 citations

TLDRCocoa is presented, a system that introduces a novel design pattern -- interactive plans -- for collaborating with an AI agent on complex, multi-step tasks and saw the interleaving of co-planning and co-execution as an effective novel paradigm for human-AI collaboration.

How do people cite this paper?

(generated 20 days ago)

Cocoa's interactive plans design pattern and its concept of steerability have been adopted by subsequent work designing notebook-style interfaces for AI-assisted writing and research, referenced as a representative approach to user control in agentic systems spanning co-planning, co-execution, and flexible delegation, and cited broadly as motivation for structured interaction frameworks beyond chat, mixed-initiative controls that promote balanced human-AI interaction, and GUI-first designs for single-agent systems, while also serving as a reference point for AI-assisted scientific research tools and for studying what makes LLM-generated plans helpful to users.

Paper

Loading PDF reader...