‹ Support & Field Services recipesSupport & Field Services · Agent Blueprint
Recurring issue radar
The same issue, reported three different ways this week. Cluster detection across tickets and calls, before it becomes an incident.
Side by side: token usage, with and without Paperbase
~115k tokens saved · 82% less work for the agentWithout Paperbase
~140k tokens
- Read the week's tickets and calls.
- Manually spot the same problem described differently.
- Ask agents if anything kept coming up.
- Miss the pattern until the incident page pings.
With Paperbase
~25k tokens
- Pull the week's tickets, calls, and customer interactions.
- Ask Paperbase to cluster issues by underlying theme, regardless of wording.
- Rank clusters by volume and velocity.
- Flag any that grew within the week.
Agent prompt
You are scanning for recurring issues this week. Using Paperbase: 1. Pull the last 7 days of tickets, calls, and customer interactions. 2. Cluster them by underlying Theme, not surface wording. 3. Rank clusters by volume and how fast they grew. 4. Flag any cluster that looks like an incident forming. Output: - Recurring issue clusters, ranked. - For each: the descriptions (quoted), the volume, and the trend. - The two clusters most likely to escalate, with a recommended owner.
Placeholders in {double_braces} are inputs the agent will ask for at runtime. Give the prompt to any agent connected to Paperbase, and the rest grounds in your own memory graph.