Call log

The call log is where you find out what your agent is actually doing. If you only look at one screen in the first two weeks, make it this one.

Telephony → Calls

The list

Newest first, with the essentials on each row: when it happened, direction, the other number, how long it lasted, a quality score, a truncated summary and the cost.

Tabs filter by direction — all, inbound, outbound. Click the timestamp to open a call.

The quality badge is a fast triage tool: green is fine, blue is worth a look, red means read it. A dash means the call wasn't scored (either it's on a plan without AI analysis, or the call was too short to score).

Inside a call

Transcript

The full conversation, laid out like a chat — agent on one side, caller on the other.

Reading transcripts is the highest-value habit in this product. It's uncomfortable at first (agents say awkward things) and it's the fastest possible feedback loop for your instructions and knowledge base.

Overview

Direction, which agent took it, which number, when, how long, what it cost, the quality score out of ten, and the sentiment read.

Recording

An audio player, if recording is on and included in your plan. Hearing a call tells you things a transcript can't — whether the agent sounded rushed, whether the caller was losing patience.

Summary

One or two sentences on what the call was about and what happened. This is what makes a hundred calls a day reviewable instead of theoretical.

Important notes and improvements

Bullet points pulled from the conversation: things worth knowing, and specific suggestions for what the agent could have done better. The improvements list is effectively a to-do list for your agent instructions.

What went wrong

When a call went badly, this names the side it went wrong on — the agent, the caller, or a technical problem — with detail. Useful because "bad call" and "bad agent" aren't the same thing; plenty of calls fail because the caller was in a tunnel.

Extracted fields

The custom variables you asked for in Post-call — budget, property type, preferred day, whatever you defined. Structured data pulled out of an unstructured conversation.

Final notes

The closing takeaway for the call.

Call kinds

Not every row is a real customer call:

Test calls are logged so you can review them, and marked so they don't distort your statistics.

Statuses and why a call ended

• Active — happening right now. Duration, cost and summary aren't filled in yet.
• Completed — finished normally.
• Failed — never connected.

For calls placed by the call queue you also get an ended reason: answered, no answer, busy, declined, unavailable, or failed. That's how retries are decided.

What's filled in when

A call row appears the moment the call starts and gets enriched as things complete:

1. At the start — direction, numbers, agent, status active.
2. When it ends — duration, final status, and the charge to your wallet.
3. Shortly after — the transcript, then the AI summary, quality, sentiment and extracted fields.

So a call you open immediately after it ends may be missing its summary for a few seconds. That's normal, not an error.

What each plan gets

AI analysis is metered monthly. If you run out, calls are still logged and transcribed — they just don't get the AI layer. You can buy more with the AI Call Analysis add-on.

Where else calls appear

The same calls show up in several places, filtered for the context:

• On a contact — everything you've ever discussed with that person.
• On a number — the last 100 calls on that specific line.
• In Analytics — aggregated into trends.

Opening a call from any of those and pressing back returns you where you were.