# Multi-Tool Workflow Decomposition ('Hack Your Own Multi-Agent System')
Deliberately splitting an AI-assisted build across several separate tools, each dedicated to one sub-task, instead of doing everything inside a single tool's context. Framed informally as 'hacking your own multi-agent system' by chaining single-purpose tools together.
Related to Context Engineering, of which this is a tool-orchestration tactic, and to Solve It's Notes/Prompts/Code Message System, which curates context within a single tool rather than across tools.
A Notion AI custom agent can be explicitly instructed to dispatch multiple sub-agents in parallel — each pulling from a different data source (Slack, a task database, a metrics tool, a meeting-transcript store) — and then combine their outputs into one artifact. The instruction to the parent agent is to "fan out" across sources as separate sub-tasks rather than processing them sequentially in one pass, which is materially faster and keeps each sub-agent's context scoped to one source. Demonstrated by Ryan Nystrom's scheduled daily meeting pre-read; see Personal AI Briefing Agent (Glean-Based).