Prioritizing with AI: how it works

Every message that reaches Loopline gets a priority within seconds. This article explains what the model looks at, where it is deliberately cautious — and how you stay in control of the rules.

What the model sees

Loopline reads the text of a message together with a few pieces of context. It never looks at data outside your workspace.

• Urgency signals: words like “down”, “can’t log in” or “charged twice”.

• Sentiment: whether the customer sounds frustrated, neutral or happy.

• Customer context: plan, account age and previous conversations.

• Channel: a one-star app-store review is weighted differently than a feature idea in the chat.

From signals to a priority

The signals are combined into one of four levels: urgent, high, normal and low. Each level comes with a suggested owner and a due date, based on the rules your team has set up.

The model explains every decision in one line — for example, “Urgent: payment issue, customer on Business plan”. That makes it easy to spot a wrong call and correct it.

Where the limits are

Automatic prioritization is a starting point, not a verdict. Irony, very short messages and mixed languages are harder to read. In those cases Loopline marks the message as “needs review” instead of guessing.

How you adapt the rules

1. Open Settings → Prioritization.

2. Add keywords or customer segments that should always count as urgent.

3. Set response targets per priority level.

4. Review the weekly accuracy report and adjust where needed.

Every manual correction is used to fine-tune the suggestions for your workspace — so the inbox gets calmer the longer you use it.