Solstice Helpdesk
A ticketing and support-ops platform with native Slack integration, AI-powered ticket automation, and human-in-the-loop guardrails on every AI action.
Industry
Technology & SaaS
Timeline
18 weeks
Team
3 engineers, 1 ML engineer
Category
Enterprise Application
Where things stood
Solstice's support team was split across email, Slack DMs, and a legacy ticketing tool, so tickets got lost between channels and nobody had a single view of SLA risk or team workload.
How we got there
Slack-native ticketing & ChatOps
Built two-way Slack integration so tickets can be created, assigned, and resolved directly from a channel or DM, with ChatOps commands to trigger workflows and coordinate incidents without leaving Slack.
SLA-aware routing and Kanban workflow
Built intelligent ticket routing with automated SLA escalations and a drag-and-drop Kanban board with WIP limits and swimlanes, so priority is visible at a glance instead of buried in a queue.
Autonomous AI agents with human-in-the-loop guardrails
Deployed an agentic chatbot that resolves common queries end-to-end and a log-analysis agent that surfaces root causes automatically — both routed through human approval for high-impact actions, with rate-limiting and prompt-injection guardrails on every model call.
What changed
Slack-native
Tickets created, assigned, and resolved without leaving Slack
~60%
Routine tickets resolved by the AI agent without human handoff
100%
High-impact AI actions routed through human approval
Tech stack
Slack API
Powers two-way ticket sync and the ChatOps commands run directly from a channel.
Claude API
Drives the agentic chatbot and the log-analysis agent, with guardrails on every call.
Node.js
Orchestrates routing, SLA escalation, and the human-in-the-loop approval queue.
PostgreSQL
Stores tickets, SLA state, and the full audit trail of AI actions and overrides.
WebSockets
Powers the real-time Kanban board and live SLA dashboards.
TypeScript
Typed every action the agent is permitted to take, enforced by the guardrail layer.
