Building Your AI Workforce: Coordinators and Specialists
Learn how to design a practical AI workforce with coordinators, specialist subagents, tools, knowledge, and owner controls.
Building Your AI Workforce
Building an AI agent workforce in Anakin is completely different from designing a single monolithic chatbot. Anakin uses a specialized multi-agent architecture where "Coordinators" handle user intent and "Specialist Subagents" handle specialized tasks.
In this guide, we'll walk through exactly how to build this workforce, configure components, and explore the HIH (Human-Intelligence-Human) approach.
1. Creating the Coordinator (The Router)
Your Coordinator Agent is the entry point. When a visitor sends a message, the Coordinator interprets it to understand the request. It does not answer deep technical questions directly. Instead, it securely routes the context to a Specialist designed specifically for that capability.
To start:
1. Navigate to your Agent Dashboard.
2. Click Deploy Agent and set its type to Coordinator.
3. Set the coordinator's initial instructions to focus strictly on routing logic.
2. Adding Specialist Subagents
Specialists are deeply focused agents that only handle one domain (e.g., Sales, HR, Customer Support). Because their prompts are much smaller and tightly constrained, they are much less prone to hallucination.
Under the Coordinator configuration, you can assign Specialist Subagents.
Click the Deploy Agent button to spawn a new specialist and choose its Coordinator.
3. Configuring Behavior and Prompting
Once you click into a Specialist, you must configure the specific Behavior and Prompts.
Best Practices for Specialist Prompts:
- Tell the specialist exactly what persona to adopt (e.g., "You are a senior mortgage advisor").
- Provide clear boundaries (e.g., "Do NOT answer questions about technical support; refer them back to the coordinator.")
- Explain how they should sequence their questions instead of bombarding the visitor.
4. Enabling Built-In Tools
For a Specialist to take action (like booking a meeting, saving a lead, or sending you a notification), you must attach Tools.
When you toggle on a tool (like the Scheduler or Lead Collector), the specialist learns when that tool is available and how it should be used. The agent can then offer the right action naturally during the conversation.
5. Adding Domain Knowledge (The Knowledge Base)
If your Specialist needs specific company context, such as product catalogs, PDFs, or FAQs, you must build its Knowledge Base.
When you upload items, Anakin organizes them so the specialist can find the most relevant details while answering. Visitors get responses grounded in your actual materials instead of generic guesses.
Understanding HIH: Human-Intelligence-Human (Double-Faced AI)
Anakin implements a groundbreaking architectural pattern we call HIH (Double-Faced AI). Your agent looks and responds differently depending on who is talking to it.
The Visitor's View (Public Interface)
When a regular visitor interacts with the agent via your embed or public link, the agent acts as your polite, professional workforce. It will guide them through funnels, answer questions, and try to collect leads.
The Owner's View (Command Hub / Testing Chat)
As the Owner, you do more than test the public chat. From your Dashboard, you can guide your agents, add business updates, and correct behavior in real time.
Owner Capabilities:
- You can say: "Hey, we have a server outage right now. If anyone asks about slow load times, tell them we are working on it."
- The agent acknowledges your command.
- The very next visitor who says "Why is the app so slow?" will receive the updated context seamlessly.
This means you can steer your agents in real time using natural language, much like sending a quick Slack message to your human team, without ever touching the baseline Behavior Prompt.