Multi-Agent HR Hub-and-Spoke

The gap

Most AI demos treat HR like a single chatbot problem. In reality, HR work spans compliance, payroll, legal, and document generation — and each area has different rules, different data sensitivity, and different failure costs. A single generic agent either gives shallow answers or risks saying the wrong thing in the wrong context.

The solution

I built a live prototype using Hermes Agent in which a central hub decomposes incoming HR requests and routes work to specialist agents. Each specialist handles its own domain: compliance, payroll, legal, or document templates. The hub then synthesizes the specialist outputs into a coherent, deliverable result.

What makes it hard

The most complex part was the routing logic. Each request has to land in the right specialist context without allowing sensitive information to leak across agents. A payroll question should not accidentally expose PII to the document specialist. A compliance check should not be answered by the general-purpose hub.

Capabilities

• Document generation from structured inputs.
• Version history and audit trail for every output.
• Compliance monitoring with domain-aware checks.
• PII-aware routing that keeps sensitive data in the right context.

Where it is headed

The longer-term vision is HRIS-agnostic integration through APIs. HR teams would interact with their systems in natural language instead of navigating a different, complex interface for every platform.