AI-Powered Employee Onboarding and Internal Support

New employees had to navigate multiple systems to find information related to HR, internal processes, security, benefits, tools, and procedures. This created lengthy onboarding processes, repetitive questions to HR and support teams, dependence on colleagues and managers, difficulty locating the latest policy versions, and lost productivity during the first weeks of employment.

The organization also faced inconsistent employee experiences across departments and locations and wanted a personalized internal assistant capable of providing reliable information 24/7 using official organizational knowledge.

Rootstack implemented an AI Employee Assistant using RAG to centralize access to corporate knowledge and provide contextual answers through natural-language conversations. The solution connected to more than five internal knowledge sources, including Google Drive, Confluence, SharePoint, internal wikis, employee manuals, corporate policies, knowledge bases, and tool documentation.

The architecture used semantic search and vector embeddings to retrieve relevant information before generating responses. User context—including department, role, location, employee type, access level, and authorized applications—was incorporated to personalize responses and ensure that employees only received information they were authorized to access. Permission validation was applied at the retrieval level, preventing the assistant from becoming an indirect mechanism for accessing restricted information.

Employees could ask questions about onboarding, system access, vacation policies, security procedures, expenses, equipment configuration, and purchase approvals without knowing where the information was stored. Responses included references to the original documents, with the source repository, relevant section, and document link when available. The assistant also maintained conversational context for follow-up questions and could escalate complex or out-of-scope requests to human support without requiring employees to repeat their situation.

For selected use cases, Rootstack connected the assistant to internal tools through REST APIs and MCP, allowing it to perform authorized actions such as creating requests, checking ticket status, retrieving internal information, providing forms, and initiating workflows. Docker was used for application deployment, PostgreSQL for persistence, and OpenTelemetry for observability and tracing.

The project was implemented over 5 months and delivered measurable improvements, including an estimated 60% reduction in onboarding knowledge search time, 70–80% of internal questions automated, an estimated 55% reduction in repetitive HR requests, and an estimated 65% reduction in policy and procedure search time. More than 95% of responses included sources, the assistant was available 24/7, and permission controls were applied at the retrieval level.