Why AI Still Needs to Ask Questions: The MCP Elicitation Revolution

When developers like GitHub’s engineers face a critical challenge—how do you stop AI from making assumptions?—the answer lies in a seemingly simple concept: elicitation. The Model Context Protocol (MCP) elicitation is reshaping how AI tools like GitHub Copilot interact with users by refusing to proceed until it gathers the right information.

The Problem With Default Assumptions

AI systems have a fundamental weakness: they operate on assumptions. When you ask GitHub Copilot or any AI-powered tool to execute a task, it often relies on default parameters that may completely miss what you actually want. This friction point—where user intent diverges from AI inference—creates friction in the development workflow. MCP elicitation flips this script by having the AI pause and ask clarifying questions upfront.

How MCP Elicitation Actually Works

The mechanics are elegant. When integrated into systems like Visual Studio Code’s GitHub Copilot, the MCP server performs a real-time check: Do I have all required parameters? Are there optional details that would improve the outcome? If gaps exist, the system initiates an elicitation prompt—essentially asking the user for missing context before proceeding.

Take a practical example that developer Chris Reddington (a notable figure in AI integration development) encountered: a turn-based game server. Initially, the system offered multiple overlapping tools for different game types. The AI agent would randomly select the wrong tool because tool names weren’t distinct enough. The solution? Consolidate and clarify: use schema-driven prompts that precisely define each option’s purpose, forcing the AI to request specific parameters like difficulty level or player name before launching the game.

From Technical Complications to User Clarity

Reddington’s development stream revealed the iterative path forward. The complications weren’t just technical—they were semantic. Tool naming matters. Parsing initial requests to identify only what’s truly missing matters. By refining these elements, the team didn’t just solve an engineering problem; they transformed how users interact with AI-powered features.

The refined approach means a user requesting tic-tac-toe doesn’t get generic defaults. Instead, the system intelligently prompts: “Difficulty level?” “Your player name?” “Board size preference?” Each answer personalizes the experience rather than forcing users into preset options.

Why This Matters Beyond Gaming

The implications extend far beyond casual applications. Every AI-assisted workflow—code generation, data analysis, content creation—suffers from the same assumption problem. MCP elicitation addresses a fundamental user experience gap: the distance between what users want and what AI delivers without sufficient context.

The Path Forward

As AI tools continue multiplying across development environments, the integration of MCP elicitation offers a template for intuitive interaction design. It acknowledges a core principle: better input yields better output. Rather than AI systems pretending to understand what you mean, they ask. Rather than users frustrated by incorrect defaults, they participate in shaping the outcome. This shift from assumption-driven to information-driven AI represents a meaningful evolution in how technology serves human intent.

The future of AI interaction isn’t about smarter algorithms making better guesses—it’s about creating pathways where users and AI collaborate transparently, one clarifying question at a time.