Instead of manually choosing between five different ChatGPT agents for each task, one journalist built a single "routing" agent that automatically recommends the right specialist for the job. The shift has transformed how she works, eliminating the constant mental overhead of remembering which AI to open first.

What Happens When Your ChatGPT Sidebar Gets Too Crowded?

Like many power users of artificial intelligence, Amanda Caswell, AI Editor at Tom's Guide, found herself drowning in custom GPTs. She had built specialized agents for fact-checking, brainstorming, research, editing, and reviewing content. Each one was useful in isolation, but together they created a productivity problem: every new task started with a decision about which agent to open first.

The real issue wasn't that the agents were bad at their jobs. It was that managing them became its own job. "Every single new task starts with another new decision as I wonder if I should open research GPT first or jump straight into writing and come back for editing?" she explained. The cognitive load of remembering which assistant handled which task, and in what order, was draining her productivity rather than enhancing it.

How Does a Routing Agent Actually Work?

Instead of creating yet another specialist agent, Caswell tried something different: she built one AI whose only job is deciding which AI should handle the task. This "routing agent" sits at the entrance of her workflow and acts like a project manager, evaluating each request and recommending the best sequence of specialists.

The routing agent uses a simple decision framework based on five core questions:

• User Goal: What is the person actually trying to accomplish with this task?
• Best Specialist: Which specialist agent is best suited to handle this specific work?
• Multiple Agents: Should more than one specialist be involved in the workflow?
• Sequence: In what order should the specialists work to maximize efficiency?
• Missing Information: Is any critical information missing before work can begin?

The prompt Caswell uses to set up her routing agent is straightforward: "You are my AI Routing Agent. Whenever I describe a task, identify my goal, decide which specialist agent should complete it, recommend the best sequence if multiple agents are needed and explain your reasoning briefly. If important information is missing, ask clarifying questions before continuing".

What makes this system intelligent is that it knows when not to suggest an agent. If she's rewriting an email in a softer tone to a neighbor, the routing agent won't recommend the fact-checking specialist. It understands context and applies judgment, not just pattern matching.

How to Build Your Own Routing Agent System

The process of setting up a routing agent workflow is straightforward and doesn't require technical expertise. Here's how to implement it:

• Identify Your Repetitive Tasks: Start by listing the three to five tasks you do most often, such as analyzing data, summarizing documents, brainstorming ideas, writing content, or generating images. These become the foundation for your specialist agents.
• Create Focused Specialist Agents: Build one custom GPT or AI project for each task. The narrower the focus, the better each agent performs. A brainstorming agent should only brainstorm; a research agent should only find sources and identify gaps.
• Build the Routing Agent: Create one additional agent whose sole responsibility is deciding which specialist should work on each task. Give it the decision framework above and let it learn your workflow patterns.
• Start Every Project with the Router: Instead of opening whichever GPT seems right, always begin with your routing agent. Describe your task naturally, and let it recommend the workflow automatically.
• Expand Over Time: As you discover new repetitive tasks, add new specialists for SEO optimization, social media posts, coding, spreadsheet analysis, or email drafting. The routing agent learns about each new specialist and incorporates it into future recommendations.

Caswell's own specialist agents were built for journalism work, but the system is flexible. A software developer might create agents for code review, debugging, documentation, and testing. A marketer might build agents for copywriting, audience research, campaign planning, and analytics. The structure adapts to your specific workflow.

Why This Approach Beats Having One "Do Everything" AI

Most people still treat AI like a single assistant that should be good at everything. But research and real-world use show that AI models perform better when each agent has one clear responsibility. "With any productive team, everyone has a specialty," Caswell noted. "That's why I decided my AI should work the same way".

The routing agent approach eliminates three major friction points. First, it removes decision fatigue; you don't have to remember which agent to open. Second, it creates consistent workflows; the routing agent automatically suggests the best sequence of specialists. Third, it scales without becoming more complicated; as you add new specialists, the router learns about them and recommends them when appropriate.

Caswell found that this system works across different AI platforms. "What I like about my system is that it still keeps the agents within the same platform. It works just as well with ChatGPT agents as it does with Claude," she explained. The routing agent concept is flexible enough to work with multiple AI models, giving users freedom to choose their preferred platforms without losing the organizational benefits.

The result is a productivity boost that surprised even the person who built it. "Routing agents have changed everything for me," Caswell stated. "Instead of asking myself which GPT to open, I now start every project with one assistant. Its only responsibility is deciding what happens next".