TicketScope now scores every ticket for SLA risk before you read it

MSP service desks don't have a ticket problem. They have a triage problem.

The queue has 40 tickets. Most of them are fine. Two or three of them are about to blow up if no one touches them in the next hour. The ticket list view doesn't tell you which ones.

Priority helps. But priority is set at creation by a human who didn't know how the ticket would evolve. A Medium-priority "internet intermittent" ticket that turned into a full outage with an angry client note from an hour ago still shows as Medium.

Today I'm shipping TicketScope's ticket health panel — a new Pro feature that loads automatically in the sidebar on every ticket you open and tells you:

• Risk score (1–10): how likely is this ticket to become an incident if untouched?
• Sentiment trend: positive / neutral / negative / escalating — based on what clients have actually written
• Urgency: low / medium / high / critical — a specific SLA window, not just a feeling
• Days since last engineer note: the thing everyone knows but nobody surfaces
• Next action: one concrete sentence — "Call client now," "Post a status update within the hour," "Escalate to on-call — no response in 3 days on a reported outage"

How it works

The panel loads automatically after the ticket summary. You don't click anything. Open a ticket, see the score.

Under the hood, TicketScope sends the ticket notes and metadata to Claude Haiku. The model looks at:

• Critical language: "outage," "completely down," "SLA breach," "cannot work," "all users affected" — these are hard signals that bump the score immediately
• Sentiment language: "unacceptable," "frustrated," "ridiculous," "losing money" — negative sentiment, especially when it's escalating across multiple client notes
• Temporal signals: how many days since an engineer last posted? Is the most recent note from a client with no engineer reply?
• CW priority: P1/Critical tickets start higher, but the score can move above priority if the situation warrants

When the Anthropic API key isn't configured, a heuristic stub runs the same analysis locally — keyword matching, note-recency checks, sentiment word lists — with no external API call. Lower fidelity, but still surfaces the obvious fires.

What it looks like

The sidebar health panel shows a risk score bar (the bar fills from green to red), urgency and sentiment badges, a stale-days warning when the ticket has gone quiet, a recommended action callout in blue, and bullet flags for specific observations.

Ticket health
Risk  ████████░░  8/10
CRITICAL   ↓ escalating
Last engineer note: 3d ago
→ Call client immediately and escalate to on-call if not already done.
• No engineer activity in 3 days
• Critical keyword detected (outage/breach)
• Client expressed strong negative sentiment

The design decisions

One next action, not a list. Engineers at 8 AM with 40 tickets don't need considerations — they need a default. "Post a status update within the hour" is actionable. "Consider whether this warrants immediate attention given the context" is not.

Heuristic fallback. "Outage" and "3 days without engineer response" don't require a language model to detect. The stub runs in milliseconds with no external call. This matters for MSPs who don't want ticket text leaving their network, and for latency.

Automatic loading. The health panel loads in the background after the summary, without any click. If you have to remember to request the health score, you'll only use it when you're already worried about the ticket — the wrong time.

Try it

TicketScope is a Chrome extension that adds an AI sidebar to ConnectWise Manage. The free tier includes ticket summaries (50/month). Pro ($30/engineer/month) includes ticket health, similar past tickets, resolution drafts, time entry suggestions, client status updates, KB article generation, and vendor escalation drafts.

You can download it from GitHub and side-load it in Chrome today. No account required for the free tier.