Table of contents
- What are AI assistants vs AI agents
- Key differences
- Graphite: How it blends assistant and agent modes in code review
- Frequently asked questions
What are AI assistants vs AI agents
AI assistants are reactive tools that respond to explicit user commands and perform tasks accordingly — examples include Siri, Alexa, or ChatGPT that await each prompt to act
AI agents, on the other hand, are proactive, goal-driven systems. Once given an objective, they autonomously plan and execute workflows, sometimes using external tools or memory, with minimal ongoing prompting
Key differences
| Characteristic | AI assistant (reactive) | AI agent (proactive) |
|---|---|---|
| Autonomy | Requires frequent user input | Operates independently once given a goal |
| Control flow | Prompt → response | Goal → planning → execution |
| Memory & context | Often stateless or short-lived | May maintain memory, handle multi-step workflows |
| Complexity | Suited for simple, defined tasks | Built for complex, autonomous, long-running tasks |
| Computational load | Lower | Higher, due to reasoning, monitoring, and planning requirements |
How Graphite blends assistant and agent modes in code review
Graphite is a unified developer platform designed to help engineering teams ship higher-quality software, faster—with deep integration into GitHub, VS Code, and CI/CD workflows. Graphite embodies both AI paradigms in a single workflow:
- Diamond acts as the autonomous agentic layer, proactively reviewing every PR, surfacing critical issues, and offering instant suggestions without any user input.
- Graphite Chat serves as the reactive assistant, giving developers a conversational interface for deep dives, clarifications, targeted fixes, and quick action within the pull request.
Graphite demonstrates both models: Diamond acts autonomously with agentic behavior; Graphite Chat remains a reactive, prompt-driven assistant within developer workflows.
Frequently asked questions
What is the fundamental difference between AI assistants and AI agents?
AI assistants are reactive—they await and respond to user prompts. AI agents are proactive—once provided with a goal, they can independently plan and execute a series of actions to meet that objective
How do LLMs relate to both AI assistants and AI agents?
Both rely on large language models for natural language understanding. Assistants use LLMs to interpret prompts and return responses. Agents leverage LLMs for reasoning, planning, tool use, and autonomous action over extended tasks
Can you give real-world examples of each?
Assistants: Siri, Alexa, ChatGPT help with routine tasks on demand. Agents: Autonomous systems like code review tools, content moderation systems, or Graphite's Diamond—once initiated, they act independently toward their goals
What are the limitations of AI agents?
Agents are more complex to build, can be costly, prone to errors in unsupervised operation, and require monitoring to avoid going off-track
Why would a development team use Graphite?
Graphite integrates both approaches: Diamond automates high-signal, context-aware code reviews (agent), while Graphite Chat offers an interactive assistant for on-demand feedback. Together, they streamline review workflows, enforce standards, and assist in PR management with less friction
How should I decide between AI assistant vs AI agent for a use case?
If you need simple, prompt-driven help (like drafting, summarizing, searching), an assistant fits best. If you're aiming to automate complex workflows or repeatable multi-step tasks with minimal fuss later, agentic solutions offer more autonomy and efficiency—though at higher complexity.