# SLA & Automation

A serious support operation runs on commitments and repeatable process: response targets you can stand behind, and workflows that don't depend on someone remembering to act. This article covers the SLA engine, the automation model, and how the helpdesk extends into the rest of your stack. Configuration steps are in the signed-in documentation.

## SLA Policies

An SLA policy defines, per priority, the targets your team commits to: **first response**, **next response**, and **resolution** times. Policies are flexible enough for real operations:

- **Business hours** — targets can count only working time, respecting a timezone, working days, and holidays, so a Friday-evening ticket isn't "overdue" Monday morning.
- **Pause on the customer** — the clock stops while a ticket waits on the customer and resumes when they reply, so you're measured on time you actually control.
- **Tiered policies** — specific accounts (VIP or contract customers) can be given tighter targets that override the default, without changing everyone else's.

Customers never see your internal SLA mechanics — the portal shows a friendly *expected response time*, while your team works against the real targets.

## At-Risk and Breach as Events

As a ticket approaches its target, an **at-risk** event fires; if a target passes, a **breach** event fires. These aren't just red badges — they're real platform events you can act on: nudge the owner, notify a lead, or reassign automatically. That turns your SLA from a passive scoreboard into an active safeguard.

## Automation

The helpdesk runs on Karmaflow.ai's orchestration engine — the same automation layer as the rest of the platform. Tickets emit events at every meaningful moment (created, replied, status changed, escalated, solved, SLA at-risk, SLA breached, CSAT received), and workflows can react to any of them with branching, waits, AI-router steps, and actions that create or update tickets, message customers, or call external systems.

Because AI agents are actions in this engine, automation and AI compose naturally. A representative pattern: when a voice call or chat ends in an escalation, an automation opens a ticket with the transcript summary, routes it to the right team, and lets an AI agent draft the first response — all without a human lifting a finger until it matters.

## Extensibility

The helpdesk is built to fit into a real stack, not to be an island:

| Surface | What it enables |
|---|---|
| **REST API** | Create and manage tickets programmatically from your own systems |
| **MCP tools** | Let any connected AI agent — including external ones — file and update tickets over the Model Context Protocol |
| **Event triggers** | React to any ticket event in an orchestration, or forward it to an external webhook |
| **Native channels** | Email, chat, SMS, voice, and web forms are wired in without integration work |

## Why It Matters

Response-time commitments and automated process are what separate a support *operation* from an inbox. Karmaflow.ai gives you SLA targets with real-world nuance (business hours, customer-wait pausing, tiers), turns approaching-breach into an actionable event, and lets you automate the repetitive parts with the same engine — and the same AI — that powers everything else on the platform.

> **Ready to define your policies and workflows?** The signed-in documentation walks through SLA setup and building automations step-by-step. [Sign in to Karmaflow.ai](https://omni.karmaflow.ai/auth/login) to read the guides.