An agent for frontline supervisors: uses approved HR policies and real-world operating scenarios to practice difficult conversations and choose appropriate next steps.
Frontline supervisors handle hard conversations about attendance, safety, conflict, client complaints, performance, and communication, often unsure what policy says or how to word it. Inconsistent handling creates risk and pulls HR into routine questions.
I designed and built the agent, the workflow, the two MCP servers behind it, and the tests HR used to evaluate it.
Gives the GPT grounded answers from approved policies, manager guides, and escalation rules.
Pulls approved conversation guides, training, documentation templates, and manager resources.
Retrieves approved HR policies, manager guides, escalation rules, and conversation templates before answering.
I started with the HR team because they own the policies and are best placed to judge whether the guidance is accurate, appropriately cautious, and consistent with how HR coaches. Testing inside HR came before any supervisor use.
Two MCP servers. hr-policy-library holds approved policies, escalation rules, and conversation templates. learning-coaching-resources holds coaching guides, training, manager resources, and documentation templates. Escalation rules live in one place, so when HR is required is defined by data, not left to the model. The code and its automated tests are on GitHub.
Built on a shared quality framework as the approved source of truth, this concept answers process questions, identifies missing information on a request, categorizes recurring error themes, and produces coaching or QA summaries for team leads. Higher-risk areas involving sensitive personal data or employment decisions are intentionally kept out of scope.
Whether you're modernizing platforms, scaling AI, or rethinking your operating model — let's talk about what's possible.