Quality Monitor
In this article
In this article group
Quality Monitor watches your customer feedback for quality issues, such as a jump in checkout complaints after a release or a bug that shows up in support tickets and app reviews on the same day. Enterpret Agent investigates each possible issue before it reaches you. When an issue is confirmed, Quality Monitor sends the alert or starts the work you set up, such as a Linear or Jira ticket or a PagerDuty incident.
Your existing monitors keep working. Their filters and Slack or email destinations stay as they are.
What changed
Quality Monitor used to stop at a spike. It flagged topics that jumped and wrote the alert from summaries of the feedback. Someone on your team still had to check that the issue was real, find the cause, and measure its impact.
Enterpret Agent now does that work before an alert is sent. In our benchmark, this cut false positives by 95%.
Two other things changed:
- An alert can start work. A confirmed issue can file a ticket, page on-call, or post to Slack or email.
- Alerts use AI credits. The investigation and the action behind each alert are Enterpret Agent work. See How Quality Monitor uses credits.
Your existing monitors
Each existing monitor appears under Running here on the Automations page, with its filters and destinations unchanged. Open it to add an action.
The Automations page shows the Quality Monitor templates and the monitors that run in your workspace. Sample data.
How it works
- Detect: Quality Monitor watches the feedback that matches your filter and flags unusual increases.
- Investigate: Enterpret Agent reads the feedback records behind the increase. It measures the supporting feedback, checks your Knowledge Base to tell expected behavior from breakage, and looks up recent releases when needed.
- Confirm or rule out: If the issue does not hold up, nothing is sent. If it holds up, the issue is confirmed.
- Act: The automation does what you configured, such as posting an alert, filing a ticket, or opening an incident.
- Inspect: Open the alert's session to review its report and actions, and ask follow-up questions.
The investigation rules out the noise that teams told us about:
- An outage that you already resolved.
- One customer who repeats the same complaint many times.
- A viral thread that does not reflect a wider problem.
- A feature that works as designed.
What Quality Monitor can do
Start from one of three templates. Each one runs when an issue is confirmed.
| Template | What it does |
|---|---|
| Quality Issue Alert | Sends the alert to Slack or email, with a trend chart when one is available. |
| Quality Issue Ticket | Creates one Linear or Jira ticket with the confirmed problem, measured impact, and customer evidence. If a matching ticket is already open, it adds the evidence to that ticket. |
| PagerDuty Incident | Creates a PagerDuty incident when a confirmed issue meets your on-call criteria. |
Choose a Quality Monitor template for an alert, a ticket, or an on-call incident.
You can also describe the action in plain language. For example, ask the agent to brief your Support team with what is known and a draft reply for affected customers. The available actions depend on the connectors and permissions set up for your workspace.
How to set up Quality Monitor
Go to Automations in Enterpret. Start from a Quality Monitor template, describe the automation you want in plain language, or build it step by step.
Step 1: Choose when it runs
Under Runs when, choose A quality degradation is detected. A Quality Monitor template sets this for you.
Step 2: Set the feedback to watch (optional)
Under Feedback to watch, click Add filter to limit the monitor to a product area, channel, customer segment, or tag. With no filter, the monitor watches all feedback.
Step 3: Choose what happens
- Report: Keep it on to send the alert report and daily-trend graph to Slack channels or email.
- Enabled connectors: Turn on the tools the agent can act in, such as Linear, Jira, Slack, or PagerDuty.
- Prompt: Tell Enterpret Agent what to do when an issue is confirmed. A template fills this in, and you can edit it.
For a ticket or incident template, replace the prompt placeholders with the exact Linear team, Jira project, or PagerDuty service and criteria. Enable the connected account and the actions it needs. Keep at least one Slack or email destination for the standard alert, even when an action also creates a ticket or incident.
Step 4: Name it and create it
Enter a Name, review the settings, and click Create.
A Quality Monitor automation that files confirmed issues in Linear and keeps the standard alert in Slack. Sample data.
Note: Members can create automations and edit the automations they created, without Admin access.
What each alert includes
- A plain description of the problem and any likely cause supported by the evidence.
- The number of supporting feedback records, and a trend chart when one is available. A record count is not a count of distinct customers.
- The feedback records behind the issue, in the customers' own words.
- Links to the ticket, incident, or message the automation created.
- The alert's session, where you can review the report and actions and ask follow-up questions.
Quality Monitor checks recent alerts to reduce repeats. A growing issue may receive an update in its existing Slack thread rather than a new alert. Email does not receive these duplicate follow-up updates.
The session for one confirmed issue: the steps the agent took, the Linear issue it created, and where the alert went. Sample data.
How Quality Monitor uses credits
Quality Monitor uses AI credits only when it sends an alert or starts the work you set up.
| Activity | Credit treatment |
|---|---|
| Detecting unusual increases in feedback | Included in your plan. No credits. |
| An investigation that ends without an alert | No credits. Enterpret covers this work. |
| An alert that is sent, with its investigation and the action that follows | Uses credits |
| Follow-up questions in the alert's session | See the Session credit rules in How AI credits work in Enterpret |
Most possible issues are ruled out during the investigation, and you do not pay for those. When an alert is sent, its credits cover the investigation behind that alert and the action that follows.
How many credits an alert uses
For planning, allow around 4 to 8 credits per alert, with an average of about 6. Actions such as filing a ticket or opening an incident can increase usage because the agent does more work.
These are planning estimates, not fixed amounts. Each alert uses credits based on the work completed.
| Alerts per month | Estimated monthly credits |
|---|---|
| 10 | About 60 |
| 25 | About 150 |
| 75 | About 450 |
Quality Monitor credits come from the same balances as your other Automations. For how personal credits, organization credits, and limits apply, see How AI credits work in Enterpret.
See your usage
Open Settings → Credits to see the credits each automation used, listed by name under Automations. Users see their own usage. Owners and Admins can also see usage for every user.
Settings → Credits splits usage into Sessions and Automations, with one row for each automation. Sample data.
Use fewer credits
- Narrow the filter: A monitor that watches fewer product areas or channels has fewer possible issues to confirm.
- Remove duplicate monitors: Two monitors that watch the same feedback can confirm the same issue twice.
- Pause what you do not use: A paused monitor uses no credits.
Best practices
- One monitor per product area: Separate monitors keep alerts focused and send them to the team that owns the area.
- Give each alert an owner: Route tickets and alerts to a team that can respond.
- Reserve paging for critical issues: Add clear conditions before you send an issue to PagerDuty.
- Inspect and refine: Read the first few sessions, then adjust the filter or the action.
Frequently asked questions
Do I need to change my existing monitors?
No. Your monitors, filters, and Slack or email destinations stay as they are. Each monitor appears on the Automations page, where you can add an action.
Does Quality Monitor use credits when it finds nothing?
No. Detection is included in your plan, and an investigation that ends without an alert uses no credits.
Why did one alert use more credits than another?
Credits follow the work completed. A larger issue has more records to read, and an alert that files a ticket or opens an incident does more work than one that only posts a message.
Can I change how the investigation works?
You control the feedback filter and the action. The investigation runs the same way for every monitor.
Can I see what Enterpret Agent did?
Yes. Open the session for the alert to review its report, supporting evidence, and actions.
What happens when my organization runs out of credits?
Quality Monitor follows the same rules as other Automations. See What happens when credits run out.
How is Quality Monitor different from Escalation Shield?
Quality Monitor watches for emerging product and service issues across feedback. Escalation Shield watches for one customer who is about to escalate. Many teams run both.
Need help setting up Quality Monitor? Contact support@enterpret.com.
Quality alerts or recurring reports?
Quality Monitor is an Enterpret Agent automation for investigating potential quality issues and running configured follow-up work. For a daily report, weekly digest, or recurring product-area summary, use scheduled Agent automations. Use connectors for supported actions and external context, with the permissions configured for that automation.