Skip to main content

​
What is the Call Details panel?

Click any row in the Call Log and the Call Details panel slides open beside the table. It is the complete record of a single call: the audio recording, the analysis and metadata, the full transcript, the data your script collected, and a notes area for your team. The same panel opens beside the Dashboard when you open a call from the Activity feed. The panel header reads Call Details above the contact’s name and the call’s date and time, with two icons on the right — one to download the call as a PDF and an X to close the panel. While a call is still in progress the panel updates live — the transcript streams in as the conversation happens, an Actions panel lets you steer the call, and the analysis and collected data appear shortly after the call ends. See Live Monitoring. The panel has four tabs: Details, Transcript, Data and Notes.

​
Recording player

Directly under the header sits the recording player: a waveform of the call audio with a play/pause button, an elapsed-time counter and a mute toggle. Click anywhere on the waveform to jump to that point in the call. The recording loads when you open the panel — you’ll briefly see Loading recording…. If no audio exists for the call (for example, a call that never connected), the player shows Recording not available. If the audio exists but cannot be played, it shows Failed to load recording — that message clears when you open a different call.

​
Details tab

The first tab is a one-screen digest of the call, in three sections.

​
Call Analysis

The results of Post-Call Analysis for the call. At the top you’ll see the crux of the call (a headline of eight words or less), a short summary and any recommended next steps. Below that, the Classification section lists the standard analysis fields — including the call’s Outcome, whether an opt-out was requested, whether recording/monitoring was disclosed and approved, whether a calendar booking was made, whether the contact asked for a human or asked if they were speaking to an AI, and a sentiment score from −1.00 (negative) to +1.00 (positive). If your script defines its own analysis fields, they appear in a Custom Fields section with the labels you configured. A footer line records when the call was analysed and which outcome set was used. Where a value was established during the call itself — recording consent, for example — that in-call value is what you see; it is not second-guessed from the transcript afterwards. If the call hasn’t been analysed, the section reads No analysis available for this call.

​
Call Objectives

If the call’s script defines Call Objectives, each objective is listed with its result:
  • A green tick — the objective was met
  • An empty circle — the objective was not met
  • A Pending badge — the result hasn’t been determined yet
Objective results are editable. Click any objective to toggle it between met and not met — useful when you review a call and disagree with the automatic assessment. The change saves immediately and flows through to the Objectives section of your Dashboard.

​
Metadata

A field-by-field record of the call:

​
Editing the end reason

Hover over the End Reason row and a pencil icon appears. Click it to reclassify how the call ended, choosing from:
  • AI Hangup
  • Failed
  • User Hangup
  • Voicemail
This is useful when the automatic classification got it wrong — for example, a call marked Voicemail that actually reached a person. The change saves immediately (End reason updated) and is reflected wherever the call appears, including Call Log filters. The full set of end reasons and what each means lives on the statuses reference. If the caller hung up, that is always what the end reason records — even if the assistant was in the middle of ending the call at the time.

​
Transcript tab

The full conversation, rendered as chat bubbles — the assistant’s turns on the right, the contact’s on the left, labelled with the contact’s first name. If the call produced no conversation, the tab reads No transcript available. On live calls the transcript streams in as the conversation happens, and a scroll-to-latest button appears if you’ve scrolled up while new turns arrive.

​
What produced each line

Assistant bubbles carry a small footer showing the part of the script that produced them, as Flow name · Step name. Where the assistant spoke for a specific reason rather than from an ordinary step — a safety refusal, an answer from the knowledge base, an intent pathway confirmation, or a recovery after a check failed — that reason is shown in place of the step label, so you can tell at a glance why the wording was what it was. Lines spoken by a team member during a handoff are labelled with their name.

​
Tool runs

Every tool the assistant ran appears as a marker row between the bubbles, labelled with the step it belongs to (or Tool call where there is no step name). Hover a marker to see the tool that ran. A tool that took its fallback route is tagged (otherwise) — that is a normal route, not an error. Long runs of tools collapse: only the first is shown, with a + N more control to expand and Show less to collapse again.

​
Text messages

Any SMS sent or received during the call appears inline in the transcript, in its correct place in the conversation, with the full message text. A text that failed to send appears in the system notes instead.

​
Refined transcription

Transcripts are polished a few seconds behind the live conversation: a correction pass reconciles what was heard against the whole exchange and fixes mis-transcriptions of names, numbers and short answers. Where a contact’s turn was adjusted, a small sparkle icon appears on the bubble — hover it to see the original transcription. Trivial clean-ups such as punctuation or capitalisation are applied silently and are not marked. Because the assistant acts on what it hears in the moment, the final transcript can occasionally be more accurate than what the assistant responded to — which explains the rare case where an in-call decision doesn’t quite match the transcript text you read afterwards.

​
Flagging a response

Hover any message and a flag icon appears — Flag this response. Click it, optionally type what was wrong (“What’s wrong here? (optional)”), and press Flag it. The message keeps an amber flag afterwards. Flags are the fastest way to teach the AI what to fix. They are read whenever your script is reviewed and become the raw material for AI recommendations on that script. Click an existing amber flag to reopen it — Flagged for AI improvement. From there you can edit the note and Save, or Remove the flag entirely. Some flags appear without you adding them: when a call is reviewed, anything the review considers serious enough is flagged on the exact turn where it happened. These behave the same way. Editing the note on one endorses it, and it then feeds into the recommendations for that script; removing it dismisses it, and that turn is not flagged again.

​
Explain this moment

Beside the flag icon is a sparkle icon — Explain this moment. Click it to ask the in-app assistant what happened at that point in the call. It looks at the message, the surrounding conversation and the system notes for that moment, and explains in plain language.

​
System notes

Below the transcript is a collapsible System notes panel — the call’s automated audit trail. It records what the platform did and why: which condition a step chose, which checks failed or timed out, a text message that could not be sent, a transfer that was attempted, or a warning that the call was ended early or cancelled from outside. The panel header shows a count for each severity. Expand it and filter with All, Errors, Warnings and Info. Messages that have notes are marked in the transcript — click one to jump straight to its first note, and click a note to jump back to the message it belongs to.

​
Live calls

While a call is in progress an Actions panel sits under the transcript, with the caller’s Live sentiment shown on the right and updating as the call goes on. Expand it to take over the call — see Live Monitoring for what the controls do.

​
Data tab

Everything your script’s Data Collection fields extracted from the call. A count at the top shows progress at a glance — e.g. 4 of 6 fields collected. Each field is listed with its label, a Required badge where applicable, its description, and the extracted value formatted to its type — Yes/No badges for booleans, option labels for choice fields, formatted dates and times, clickable links for URL, email and phone values, and a dash where nothing was found. Two empty states tell you why the section might be blank:
  • No data collection fields configured for this script. — the script doesn’t define any fields
  • Data extraction has not been performed for this call. — fields exist, but the call hasn’t produced values (for example, it hasn’t finished yet)

​
Variables

Below the collected fields, the Variables section lists every value the agent had available on the call — “The values that were available to the agent on this call.” — grouped into branches for contact values, custom variables, global values and Custom Variables objects. Branches start collapsed, showing a count such as 6 fields or 3 items; click to open one. Booleans render as Yes and No, and empty fields are hidden. Each branch has its own copy button, and Copy all at the top copies the whole tree. This is the fastest way to see what the agent actually knew — for example, whether a value your script expected from a CRM lookup ever arrived. The same values can be read in a post-call automation using the call.variables_data.* prefix.

​
Notes tab

A free-text notes area for your team. Click anywhere in the empty area (Add your first call note.) to start typing. Notes support Markdown formatting — headings, lists, links, bold and code all render once saved. Save with the tick icon, or discard your changes with the X; the panel shows Unsaved changes or No changes yet below the editor so you always know whether there’s something to save.

​
Line quality

Every call also carries a quality summary of the caller’s own line — background noise, echo, loudness, cross-talk and dropped audio, sampled throughout the call and summarised at the end. It is measured passively and never affects the call itself. It is not shown in the app; if a run of calls transcribed badly and you want the line quality behind them looked at, speak to the Voxworks team for further custom configuration.

​
Per-call cost

Every completed call records its cost against the call record, along with a breakdown of how the time was spent — see Usage. Cost isn’t shown in the detail panel itself; your running balance and billing details live under Settings → Billing.

​
Exporting a call as PDF

Click the download icon in the panel header to generate a Call Report PDF for the call. The report includes:
  • The contact’s name and the call’s date and time
  • The metadata grid (Type, Script, Batch, Origin, Destination, Duration, Status, End Reason, Call ID)
  • The Call Analysis section, with its Classification and Custom Fields sub-sections
  • The Objectives with met/not-met indicators
  • A Data Collection table — Field, Value, Confidence and Status for each configured field
  • The full Transcript
The file downloads as <contact-name>_call_report.pdf — ready to attach to a CRM record or share with someone who doesn’t have a Voxworks login. A toast confirms PDF downloaded.

​
Next Steps