Internal Team Support Template

• What it does: Routes internal employee questions to topic-specific agents that answer strictly from attached knowledge bases (e.g. general office procedures, HR policy), with no guessing or invented answers.
• Best for: HR, operations, internal help desks, departments that need consistent answers from policy and procedure docs
• Apps used: File Search (two knowledge bases: general office, HR policy)
• Setup time: 10 to 15 minutes

Internal Team Support is a no-code, multi-agent workflow that gives employees one place to ask internal questions. A Router Agent classifies each query and hands off to a General Office Procedure Agent or an HR Policy Agent; each agent answers only from its own File Search knowledge base. You can add more topic agents and knowledge bases and extend routing. This template acts as a starting point: any app, instructions, and agents can be updated as required for your own use case—simply ask Max (Agent Builder) to update the workflow as needed. Last verified from workflow config on March 2026. Includes common failure modes and fixes.

---

Problem this solves

• The same internal questions (office procedures, HR policy, IT, facilities) get asked repeatedly; answers are inconsistent or scattered across docs and people.
• HR and ops spend time repeating policy instead of handling exceptions and improvements.
• You want a single entry point for internal Q&A that stays grounded in your actual docs and does not invent answers.
• Scaling support: add new topics by adding agents and knowledge bases instead of more manual triage.

---

What this agent does

This template can:

• Classify incoming internal queries by topic and hand off to the correct specialist agent.
• Route follow-up questions to the same agent when they relate to the initial query; reclassify or inform the user when no suitable agent exists.
• Answer general office procedure questions using only the attached knowledge base (no guessing).
• Answer HR policy questions using only the attached HR knowledge base (no guessing).
• Politely say when information is not in the knowledge base instead of inventing answers.
• Support extension: add new handoff agents and File Search tools for more topics (e.g. IT, facilities, travel) and update the Router instructions.

---

How it works

1. Employee sends a question via the trigger (e.g. internal chat, widget, API).
2. Router Agent analyzes the query and classifies it (e.g. general office procedure vs HR policy).
3. The Router hands off to one specialist agent: General Office Procedure Agent or HR Policy Agent. If no category fits, the Router informs the user that the inquiry cannot be addressed.
4. General Office Procedure Agent receives the message, queries the General_Office File Search tool (its knowledge base), and responds with answers only from that content.
5. HR Policy Agent receives the message, queries the HR_Policy File Search tool (its knowledge base), and responds with answers only from that content.
6. If the answer is not in the knowledge base, the agent replies that it does not currently have that information.
7. Follow-ups that relate to the same topic stay with the same agent; unrelated follow-ups are reclassified or declined.

---

Requirements

• Two knowledge bases (or more if you extend): e.g. general office procedures (PDFs/docs) and HR policy (PDFs/docs). You upload these to the File Search tools.
• Router classification rules: clear topic categories that match your handoff agents (the template ships with general office and HR).
• Optional: replace the example organization/department name (“University of Carolina – Communication Department”) in agent instructions and sticky notes with your own.

---

Setup guide

1. Clone the template and open it in AgentSpace.
2. Knowledge bases: Enable and configure both File Search tools. For General_Office: connect or create a vector store and upload your general office procedure docs (PDFs, docs). For HR_Policy: connect or create a vector store and upload your HR policy docs. The template ships with these tools disabled and no vector store; you must add documents for the agents to answer.
3. Router: Review the Router Agent instructions. Ensure the classification categories and handoff names match your specialist agents (General Office Procedure, HR Policy). Update if you add or rename agents.
4. Agent context: In the General Office Procedure Agent and HR Policy Agent instructions (and sticky notes), replace the example organization/department with your company/department name so responses stay on-brand.
5. Extend (optional): To add a new topic (e.g. IT or travel): create a new agent node, add a new File Search tool and upload the relevant docs, add an edge from Router to the new agent and from the agent to the tool, and update the Router instructions to classify and hand off to the new agent.
6. Run Test Workflow with sample questions for each topic. Confirm routing and that answers come only from the correct knowledge base; confirm “I don’t have that information” when the answer is not in the docs.
7. Deploy to your internal channel (widget, chat, or API).

Common issues and fixes

• No or wrong answers: ensure each File Search tool is enabled and has a vector store with the right documents. Re-upload or update docs if content changed.
• Wrong routing: tighten the Router Agent classification instructions and examples so topics map clearly to the right agent. Add examples for edge cases.
• Agent invents answers: the template instructs agents to answer only from the knowledge base; if you see guessing, reinforce in instructions: “Do not invent or guess; if not in the knowledge base, say so.”
• Follow-ups go to wrong agent: the Router is set to send related follow-ups to the same agent; if behavior is wrong, clarify in the Router what “related” means for your topics.

Customization knobs

• Add more topic agents and knowledge bases (IT, facilities, travel, etc.) and extend the Router handoffs.
• Replace example org/department name in all agent instructions and sticky notes.
• Adjust Router logic (e.g. more categories, fallback message when no agent matches).
• Use different models per agent if you want (template uses gpt-4.1 for General Office, gpt-4o for HR).

---

Apps used

(You can add more File Search tools for additional topics.)

---

Use cases

• HR help desk: Employees ask about leave, benefits, policy; one agent answers from HR docs only. Best for HR and people ops.
• Office procedures: Questions about facilities, equipment, access, general policies answered from procedure docs. Best for operations and admin.
• Department support: Multiple topics (e.g. comms, IT, travel) with one Router and one agent per topic and knowledge base. Best for mid-size teams and departments.
• Compliance-safe Q&A: Answers only from approved policy docs; no invention. Best for regulated or audit-sensitive environments.
• Scaling internal support: Add new topics by adding agents and knowledge bases without changing code. Best for growing organizations.

---

Example prompts and outputs

Example 1: General office

Prompt (employee): “What is the process for booking a meeting room?”

Output: Router → General Office Procedure Agent. Agent queries General_Office knowledge base and replies with the procedure from the docs. If not in the docs: “I don’t currently have that information.”

Example 2: HR policy

Prompt (employee): “How many vacation days do I get per year?”

Output: Router → HR Policy Agent. Agent queries HR_Policy knowledge base and replies from policy. If not in the docs: polite message that the information is not available.

Example 3: Follow-up

Prompt (employee): “And what about carry-over?”

Output: Router sends to the same agent (HR Policy) if the follow-up relates to the previous query; agent answers from HR_Policy again.

Example 4: No matching agent

Prompt (employee): “What’s the weather tomorrow?”

Output: Router has no matching category; informs the user that the inquiry cannot be addressed at this time (or similar message per your Router instructions).

---

Why you need this template

• One entry point for internal Q&A; consistent answers from approved docs only.
• No invented answers: agents are instructed to use only the knowledge base and to say when information is missing.
• Easy to extend: add agents and File Search knowledge bases for new topics and update the Router.
• Editable in AgentSpace: change categories, add topics, or replace example org name without code.

---

FAQs

1) Is this an AI agent template or a fixed automation?

It is a multi-agent workflow you can edit in AgentSpace (Router plus specialist agents). You control routing, knowledge bases, and how many topics you support. This template acts as a starting point: any app, instructions, and agents can be updated for your own use case—simply ask Max (Agent Builder) to update the workflow as needed.

2) Do I have to use two knowledge bases?

The template ships with two (General Office, HR Policy). You can keep both, use only one (and remove or repurpose the other agent), or add more agents and knowledge bases for more topics.

3) What if the answer is not in the knowledge base?

Each agent is instructed to reply that it does not currently have that information and not to guess. You can add more docs or a fallback (e.g. “Contact HR at …”) in the agent instructions if you want.

4) How do I add a new topic (e.g. IT or travel)?

Create a new agent node with instructions to answer only from a knowledge base; add a new File Search tool and upload the topic’s docs; connect the agent to the tool; add a handoff from the Router to the new agent and update the Router’s classification and handoff logic.

5) Can the Router send to more than two agents?

Yes. The template has two handoffs (General Office, HR). Add more agents and handoffs and update the Router instructions so it classifies and routes to all of them.

6) Why are the File Search tools disabled in the template?

So you must connect your own vector stores and upload your docs. Enable each tool and attach the correct knowledge base before testing; otherwise the agents have nothing to search.

7) What are the main failure modes?

Empty or wrong knowledge base (no or bad docs), Router misclassifying (unclear categories or examples), agent guessing (reinforce “only from knowledge base” in instructions). Fix by updating docs, tightening Router logic, and reinforcing no-invention rules.

8) Can I use this for external customer support?

You can, but the template is designed for internal employee support (office procedures, HR). For external support with booking and escalation, see the AI Receptionist template.

9) Where do I upload the knowledge base docs?

In AgentSpace, click each File Search tool node (General_Office, HR_Policy), enable the tool, and connect or create a vector store. Upload your PDFs or docs there. Each agent will query only the knowledge base attached to its tool.

---

Related templates

• AI Receptionist – external-facing Q&A, demo booking, and escalation (Cal.com, Zendesk, File Search).
• Personal Ops AI Assistant – calendar, comms, and retrieval in one workflow.
• Structured Output – fetch structured data for use in other workflows.
• Pre-Meeting Preparation – prepare meeting context and agendas.
• Project Manager – project planning and tracking with Linear and web search.