Skip to main content
After each conversation, Callers analyzes the transcript and records structured results alongside the recording. You decide what to capture with key questions: the budget a lead mentioned, the date they want to start, whether they confirmed their address. Answers appear on the contact, can be used as filters, and are sent in webhooks so your CRM gets clean fields instead of a transcript.

What post-call analysis produces

Key questions

A key question has a title, an instruction for the AI, and an answer type. You can also turn any script step into a key question: hover the step number and click the star (Make a key question). See Steps.

Answer types

Choosing a type fills in a starting instruction, such as “Pick just one option” for Single Choice. For choice types, add options with Add choice or add “Other”.
New Post-Call item dialog with the title Budget, a Prompt instructions field, and the type dropdown open listing Open Question, Single Choice, Multiple Choice, Yes / No, Number, Date, and Time

The answer type dropdown on a new post-call item.

Step questions vs. global questions

  • Step questions are tied to one step in the script, such as “What’s your budget?”. The AI looks for the answer around that step. Create them from the step itself.
  • Global questions look across the whole conversation, such as “Did the contact mention a competitor?”. Create them in Post-call Analysis.
Both kinds show up together in the conversation summary.

Add a step question

1

Open the step menu

In the script editor, open the menu on the step you want to analyze and choose Make a key question.
Steps of a script block with the star icon highlighted next to step 1, a greeting that uses the name, agentName, and companyName variables

Hover a step number to reveal the star that turns the step into a key question.

2

Review the generated question

Callers drafts a question from the step’s text. Edit the title, the instruction, and the type. Click Add Title to add more than one question for the same step.
A key question dialog showing the step text, an Introduction title tab with an Add Title button, the instruction This is an introductory message from the agent with the type Open Question, and Cancel and Add to Summary buttons

The key question dialog for a step: title tabs, the instruction for the AI, and the answer type.

3

Save

Click Add to Summary.

Add a global question

1

Open Post-call Analysis

Open the agent, click Edit, and choose Post-call Analysis in the settings sidebar. You can also open it from the Summary node on the workflow canvas.
2

Add a question

Click Add, then fill in:
  • Title: the label shown in the summary and used as the key in webhooks.
  • Prompt instructions: what the AI should extract, and how.
  • Type: the answer format.
Questions created from a step also appear in this list, tagged Connected to a step.
Write instructions the way you’d brief a colleague who listened to the call: “The monthly budget the contact mentioned, in US dollars. Leave empty if they didn’t say.” Say what to do when there’s no answer, so you don’t get guesses.
Use letters, numbers, and spaces in titles. The characters $ and . are replaced with _ because the title becomes a key in the webhook payload.

Where results appear

In the app

Open a contact and select the Summary tab to see the outcome and every key question’s answer for that conversation. On the agent’s contact list, use Filter by post-call items to find contacts by answer, for example everyone who answered Yes to “Has a car loan”. The export uses the same filters.

In webhooks

Every webhook sent after post-call processing includes the analysis. Each key question appears under summary, keyed by its title:
webhook payload (excerpt)
Type values in the payload are OPEN_QUESTION, SINGLE_CHOICE, MULTIPLE_CHOICE, YES_NO, NUMBER, DATE, and TIME. See Webhook payload for every field.

In actions and advanced webhooks

Key question answers are available as variables in advanced webhooks and post-call actions, so you can map each answer to a CRM field.

Extracted variables

Extracted variables are typed values the AI pulls from the transcript for a specific integration, such as a car model, an order number, or a list of products. You define them as transcript variables (text, number, boolean, or array) in an advanced webhook. Each one appears in extractedVariables with its value, description, and type. A variable the AI couldn’t find has an empty string as its value. Use key questions for answers you want to see in the app and filter by. Use extracted variables for values an integration needs in a specific shape.

Best practices

  • Ask for what you’ll use. Every key question should map to a decision, a report, or a CRM field.
  • Prefer closed types. Yes / No, Single Choice, and Number answers are easier to filter, chart, and sync than open text.
  • Keep options exhaustive. Add an “Other” option to choice questions so unexpected answers still have a place.
  • Test with real calls. Run a few test calls and check the Summary tab before launching.

Success criteria

Define the outcomes Callers classifies.

Webhook payload

All fields sent after each conversation.

Agent results and insights

Filter and analyze results.

Contact memory

What Callers remembers between conversations.