How we built an AI assistant for an enterprise IT service desk
Blog post from Redpanda
Redpanda describes an AI IT service assistant embedded in Microsoft Teams that helps employees find knowledge-base answers, report incidents, and submit service catalog requests while continuing to use an organization’s existing ITSM forms, workflows, permissions, approvals, and records. Built for an enterprise customer, the assistant resolved 36% of employee requests through knowledge articles during its first week, preventing those interactions from becoming tickets, while other requests were classified, checked against current form requirements, completed through chat, routed through normal ITSM processes, and verified after submission. The system uses Redpanda’s Agentic Data Plane and a task-specific MCP server to limit the assistant to defined actions rather than broad ITSM access, preserves employee identity for changes through OAuth connections, and uses a separate read-only account for form definitions and verification. Governance features include access policies, data restrictions on tool arguments and responses, content guardrails, audit trails and tracing, and model-spending limits, although the text notes that available controls depend on configuration and not all were necessarily enabled for the customer. The approach is intended to reduce manual ticket preparation without transferring ownership of critical decisions, restricted information, approvals, or incident closure away from IT staff.
| Trend | Post Mentions | Total Month Mentions | Posts | Companies | MoM |
|---|---|---|---|---|---|
| MCP | 5 | 11,794 | 1,209 | 231 | +11% |
| LLM | 3 | 6,513 | 1,228 | 231 | +8% |
| AI Agents | 2 | 6,921 | 1,636 | 267 | +23% |
| Secrets Management | 2 | 3,062 | 600 | 151 | +50% |
| Observability | 1 | 3,060 | 732 | 183 | +11% |
| OpenTelemetry | 1 | 846 | 169 | 64 | +19% |
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.