Skip to main content

​
What is Post-Call Analysis?

After every call finishes, Voxworks reads the full transcript and produces a structured analysis: summaries at three lengths, a single outcome classification, compliance and behaviour flags, a sentiment score — and, if you configure them, answers to your own custom questions. The results are stored with the call and drive the Analysis tab on every call’s detail view, the hover summary in the Call Log, and can be referenced from your automations. The analysis framework is configured per script in the Post-Call Analysis tab: Call Scripts → (your script) → Post-Call Analysis The tab has three sections — Call Outcomes, Standard Analysis Fields, and Custom Analysis Fields. Changes take effect when you save the script and apply to calls analysed from then on; each call keeps the results of the framework revision that analysed it (shown in the Analysis tab’s footer). You don’t have to configure anything: every call is analysed with the standard fields and the default outcomes even when this tab is untouched.

​
Call Outcomes

Every analysed call is classified with exactly one outcome. The section shows the default outcome set for your script’s call type — as the description puts it, “Every analyzed call is classified with a single outcome. These are the default outcomes for outbound calls; customize them to match this script’s use case.”

​
Default outcomes — outbound scripts

​
Default outcomes — inbound scripts

​
Custom outcomes

Switch on Use custom outcomes for this script to replace the default set with your own rubric. The editor is seeded with the defaults so you can rename, remove and add from there. Each outcome is a row with two columns: Add Option appends a row; a rubric holds between 2 and 15 outcomes. Switching the toggle off returns the script to the default set for its call type.

​
Telephony stamping

For outbound scripts, three outcomes are never decided by the AI. When the phone network already tells us what happened, the outcome is stamped directly from the call’s status and overrides any classification:
  • No answer (no_answer) — the call rang out
  • Voicemail (voicemail) — the call reached voicemail
  • Did not connect (did_not_connect) — the call dropped out or failed
These three appear as greyed-out rows marked auto at the bottom of a custom outbound rubric — as the editor notes, “The greyed-out outcomes are set automatically from telephony status and cannot be removed.” They are always appended to a custom outbound rubric, so a call that never connected is classified correctly even if you forget non-connections when designing your outcomes. See Call statuses for the statuses behind them.

​
Standard Analysis Fields

The Standard Analysis Fields section lists the fifteen fields produced for every analysed call. They are read-only — you cannot remove or edit them — and their keys are reserved, so a custom field can never reuse one. A sentiment label is derived from the score automatically: positive at +0.20 and above, negative at −0.20 and below, neutral in between. The recording and monitoring fields give you a per-call audit trail for consent — see Recording consent for the disclosure side.

​
Custom Analysis Fields

Custom fields extend the standard analysis with your own questions. As the section explains: “After each call, the AI answers them from the transcript and stores the results with the call. The question text is the instruction the AI follows — phrase it as a question about the call.” Click Add Field to append a field (it starts as a Yes / No field named custom_field_1), then expand the row to configure it. A script holds up to 20 custom fields.

​
Writing questions that work

  • Ask about the call, not the world — the AI only sees the transcript. “Did the caller mention a competitor by name?” works; “Is this company a good prospect?” asks for knowledge it doesn’t have.
  • Match the answer type to the question — a Yes / No field should be answerable with yes or no; a Number field should ask for a figure (“How many properties does the caller own?”); a Classification field should have options that cover every call, including a none-of-the-above code such as not_discussed.
  • One fact per field — two questions in one field get one muddled answer. Split them.
  • Keep keys stable — the key is how automations and exports refer to the answer, so rename the Display Label freely but treat the Field Key as permanent once automations depend on it.

​
Where results appear

The Analysis tab on every call. Open a call from the Call Log and select the Analysis tab. The crux renders as the heading with the short summary and next steps beneath it, followed by a Classification grid of the standard fields — Yes / No answers as badges, the outcome under its display label, sentiment coloured by its score and label. Your custom fields follow in a Custom Fields section under their display labels. A footer records when the call was analysed, which outcome set applied, and the framework revision. Before a call has been analysed the tab reads “No analysis available for this call.”; if analysis failed it says so and notes the call can be reprocessed. The Call Summary. The four-paragraph summary_long becomes the call’s Call Summary, shown on the call detail view. The Call Log hover card. Hovering a call shows its short summary together with the outcome label and sentiment. Automations. A Call Completed trigger can start a pipeline once a call finishes. See Automation conditions and field mappings for how a pipeline references values from a call.

​
Next Steps

  • Call details — where the Analysis tab, Call Summary and transcript live on each call
  • Automation conditions — branch pipelines on analysis results after each call
  • Data Collection — extract structured field values from the transcript alongside the analysis
  • Call Objectives — define per-call goals that are evaluated after each call