Darth: A Bot That Receives Everything and Dispatches

What I wanted to solve

In the previous part, I shared how I set up the newsroom in Hermes Cloud with separate roles for reporting, fact-checking, writing, art, and editing.

Once that pipeline was running smoothly, another practical hurdle showed up. I work during the day, and at night and on Sundays I build with AI agents. My spare hours need to count. By early October 2026, my setup on Grok Bot already had several specialized helpers: one for everyday administrative tasks, another for personal tax questions, another for legal doubts, alongside support for security, data analytics, and visual art.

Having multiple assistants introduces a friction point that most people overlook: deciding who to talk to. Every time an idea crossed my mind or an errand needed attention, I had to open the app, scroll through the list of conversations, recall which agent handled that specific topic, and launch the chat in the right place. Doing that on a phone between daily obligations felt slow and frustrating.

When I was in a hurry, I often fell into the trap of dumping whatever I needed into the first open chat window. The result was always weak. An assistant configured for general writing or administrative routines stumbled when asked to evaluate a contract clause or inspect security configurations.

I needed a single front door. My goal was to open one conversation, type whatever was on my mind, and know that the request would be understood, organized, and dispatched to the right specialist. On October 1, 2026, I assigned that responsibility to my primary assistant and named him Darth.

Quick summary

• What worked: funneling all requests through Darth, who handles intake and points to the right specialist for each topic.
• What it cost: $0 in new spending, using the assistants already created within the app.
• How long it took: a few hours on the evening of October 1, 2026, to map out the team and write the routing rules.
• Practical takeaway: having a front desk assistant saves time and prevents a single generic model from trying to do everything and returning shallow answers.

First: what is a dispatcher or front desk bot

For anyone who has never built a multi-agent workflow, the idea of a dispatcher bot might sound complicated, but it works just like an office reception desk.

Imagine walking into an office building or a medical clinic. Right inside the entrance sits a receptionist. You do not wander down the hallway opening doors to find out who can help you. You walk up to the desk, explain what you need, and the receptionist checks the schedule: “For accounting matters, head to the second floor; for your medical checkup, go to room four.”

Darth performs that exact role digitally. He operates with explicit instructions defining the scope of each team member. When I send a message, his logic identifies the intent and responds by pointing me down the right path, or by formatting the prompt in the exact way the specialist needs to work.

The front desk comparison helps clarify the logic, but it stops working at one point: in an office, the receptionist can pick up the phone, dial an internal extension, and transfer you right away. Inside an agent app, Darth does not take control of the keyboard to operate the other bot for me. He organizes the request and hands me a clean route so I can trigger the right specialist without getting lost.

What I used

To establish this dispatch system, I relied on simple, accessible tools:

• Grok Bot: the agent platform where my everyday personal assistants live.
• Darth: the lead agent serving as the front desk, coordinator, and chief of staff.
• Specialized assistants in Grok Bot: focused agents handling dedicated areas like ADM, Accountant, Legal, and soon after Analyst, Security, and Art.
• The bot map: a simple text document created on October 1, 2026, listing where each assistant runs and its official assignment.

How I built the dispatch workflow

To make this triage work smoothly in daily life, I structured the setup into five practical steps:

1. Drawing the inventory on paper. Before giving orders to the bot, I needed a clear view of my own team. On October 1, 2026, I wrote down every active assistant, where it ran, and what it was supposed to handle. Seeing that layout clearly made it possible to set boundaries between roles.
2. Writing Darth’s core instruction. The most important rule in Darth’s system prompt was forbidding him from trying to solve everything himself. I instructed him to recognize when a topic belongs to the Accountant, Legal, ADM, or another helper, focusing on triage rather than improvising technical answers outside his domain.
3. Choosing the name and tone. I named the assistant Darth in honor of Star Wars. Having a recognizable name makes it easy to remember who manages the reception desk and adds personality to daily interactions.
4. Testing routes with everyday scenarios. I began throwing mixed requests at Darth to test his behavior. When I sent something about tax deadlines or routine paperwork, he quickly identified the nature of the request and directed it to the right specialist.
5. Protecting the boundaries. Whenever Darth started drafting lengthy opinions on legal or financial questions, I tweaked his prompt to remind him that his job is receiving, sorting, and handing off, keeping the team disciplined.

What went wrong and what I learned

My biggest initial mistake was assuming the underlying model would instinctively know when to stop talking. Because these assistants are naturally talkative, Darth’s first reaction to any prompt was to generate a full answer, even when he lacked the deep context required for that topic.

If I asked about a contract term or a routine chore, he would generate three generic paragraphs before suggesting that I consult the specialist. That wasted reading time and failed to deliver the precision I needed. I had to tighten the prompt: once the topic is recognized, Darth should name the assigned agent and format the handoff clearly and briefly.

Another important lesson was the necessity of having a written bot map. Before sitting down on October 1 to document everything, I would occasionally forget which profiles I had already configured. At times I started drafting a new prompt from scratch, only to realize days later that another assistant had already been built for that exact job. Writing down a lean list of all bots resolved that problem entirely.

The conclusion is direct: artificial intelligence without triage quickly turns into mental clutter. Funneling incoming thoughts through a dedicated front desk saves time and shields your specialists from noisy, out-of-context requests.

Costs and limits

In financial terms, this change cost $0 in new spending. Darth and the other assistants were configured within the subscription for the app I was already using.

In terms of time, I spent a few hours on the evening of October 1, 2026, sketching the map, refining Darth’s routing rules, and testing the initial handoffs. That same day, the first draft of my journey’s opening essay was written.

Regarding the limits of the tool, it pays to stay grounded:

• Darth does not replace human judgment. He helps structure my thinking and routes tasks, but every final decision remains mine.
• He does not carry out the specialist work. His role is strictly intake and triage. If you try to turn your front desk into an all-purpose handyman, output quality drops immediately.
• The workflow requires clarity on my end. If I write an incoherent message blending three completely unrelated tasks into a single run-on sentence, Darth will need to ask clarifying questions before he can direct the request.

What I would do differently

If I were starting over today, I would set up Darth as a receptionist the moment I created my third assistant. Waiting until several profiles were already running meant spending days dealing with messy chats and scattered conversations. Setting up a front desk early makes expanding your agent team simple and orderly.

Who this is for and who it is not for

A dedicated dispatcher assistant is helpful for:

• Anyone using two or more AI agents for different chores who wants a single conversation to kick off any task.
• People who value mental clarity and want to stop clicking through endless chat windows.
• Beginners who want to build the habit of delegating with structure, recognizing that complex tasks benefit from dedicated roles.

This setup is not necessary for:

• People who use AI once a week just to ask a quick factual question.
• Anyone comfortable using a single generic assistant and unbothered by shallow answers.
• People looking for automated systems to make unattended personal or business choices.

Closing

Turning Darth into my front door brought order to my evening workflow with technology. Instead of juggling multiple chat tabs on my phone, I open Darth, drop whatever thought is on my mind, and get back an organized action plan pointing directly to the right specialist.

In the next part, I will cover token economy and the rule of having agents respond only with the final result, skipping the filler.

How do you keep your AI assistants organized day to day? Do you talk to each one separately or do you run requests through an entry point? Share your setup with me on X: @RodrigoesPorto.

Sources and links

• Official project profile on X: https://x.com/RodrigoesPorto
• Grok and xAI: https://x.ai/