Post-Call Processing: Data Collection & Evaluation

After every call, your AI can automatically extract structured data and evaluate call success. This turns raw conversations into actionable insights — no manual review needed.

Two Features, One Goal

Feature	Purpose	Output
Data Collection	Extract specific information from calls	Structured values (text, numbers, dates, yes/no)
Evaluation Criteria	Measure if call goals were achieved	Success / Failure / Unknown

Both features work automatically after each call ends — you define what to look for, AI does the analysis.

Data Collection

Extract specific information from every call and store it in a structured format. Perfect for CRM updates, lead qualification, and analytics.

Go to Agents → select agent → Data tab.

Creating a Data Field

Click Add Field
Enter a Field Name — use lowercase with underscores (e.g., customer_budget)
Select Data Type:
- Text — Names, addresses, free-form responses
- Number — Budgets, quantities, prices (with decimals)
- Integer — Whole numbers only
- Yes/No — Binary answers (true/false)
- Date — Appointment dates, deadlines
Write Extraction Instructions — Tell AI what to look for

Writing Good Extraction Instructions

Be specific about what counts as a valid value:

Good example:

Extract customer's budget amount.Look for: explicit amounts, price ranges, "can spend up to"
Return: Maximum amount mentioned, or null if not discussed

Another good example:

Does customer have an existing system?Look for: mentions of current equipment, repairs, replacement
Return: true if they have one, false if they don't, null if unclear

Common Data Fields

Field	Type	Use Case
`customer_budget`	Number	Lead qualification, pricing discussions
`preferred_date`	Date	Appointment scheduling
`has_existing_system`	Yes/No	Replacement vs new installation
`property_address`	Text	Service location
`number_of_units`	Integer	Quote preparation

How Values Are Stored

After each call, AI extracts values in the specified format:

Found — Value is stored (e.g., 15000 for budget)
Not discussed — Stored as null
Invalid format — AI attempts conversion or returns null

Extracted data appears in call details and can be sent to your CRM via webhooks.

Evaluation Criteria

Automatically score every call against your success metrics. See at a glance which calls achieved their goals.

Go to Agents → select agent → Eval tab.

Creating Evaluation Criteria

Click Add Criteria
Enter a Name — Short, descriptive (e.g., "Appointment Set")
Write an Evaluation Prompt — Define what counts as success, failure, or unknown

Writing Effective Evaluation Prompts

Structure your prompt with clear outcomes:

Check if client expressed clear purchase intent.Success: Explicit agreement, scheduling, price acceptance
Failure: Rejection, "not interested", unresolved objections
Unknown: If unclear from conversation

Another example:

Was a specific appointment scheduled?Success: Date and time confirmed, address verified
Failure: Declined, wanted to "think about it"
Unknown: Discussed but not finalized

How Results Appear

After each call, every criterion is marked:

Success (green checkmark) — Goal achieved
Failure (red X) — Goal not achieved
Unknown (yellow question mark) — Cannot determine from conversation

Common Evaluation Criteria

Criteria	What It Measures
Client Ready to Buy	Purchase intent expressed
Budget Qualified	Budget within your service range
Appointment Set	Specific date/time confirmed
Contact Info Collected	Got callback number or email
Objections Handled	Concerns addressed successfully

Using Data in Webhooks

Both extracted data and evaluation results are included in webhook payloads. Use this to:

Update CRM records automatically
Trigger follow-up sequences based on outcomes
Build dashboards and reports
Route leads to sales reps based on qualification

See Integrations → Webhooks to set up data forwarding.

Best Practices

Start simple

Begin with 2-3 data fields and 2-3 evaluation criteria. Add more as you learn what information matters most.

Be specific in prompts

Vague instructions lead to inconsistent results. Define exactly what counts as success vs failure.

Use consistent naming

For data fields, use snake_case names that match your CRM fields. Makes integration easier.

Review results regularly

Check if AI is extracting data correctly. Adjust prompts if you see consistent errors.

Frequently Asked Questions

When does processing happen?

Immediately after each call ends. Results typically appear within 30 seconds.

Can I change criteria after calls have been processed?

Yes, but changes only apply to future calls. Past call results are not re-evaluated.

What if caller didn't mention the information I'm extracting?

The field will be stored as null. This is expected — not every call will contain every piece of information.

Is there a limit on data fields or criteria?

No hard limit, but keep it focused. Too many fields slow down processing and make results harder to use.

Can I export this data?

Yes, via webhooks or through the call export feature in your dashboard.