Skip to main content

Scripts

What is a script?​

Script is customer interaction logic defining sequence of system actions and responses. Scripts are used for automating phone calls and handling incoming calls.

Script types:​

  • Campaign scripts — for outbound calls (sales, surveys, reminders)
  • IVR scripts — for incoming calls (voice menu, routing)

Main Functions​

  • Creating scripts — "Create" button in top panel
  • Editing — click on script row
  • Search and filtering — by name, type, status
  • Bulk deletion — select multiple scripts

Information in table:​

  • Type — campaign or IVR
  • Name and description — script identification
  • Status — current state (Draft, Ready, In use)
  • Subscriber group — target audience for calls
  • Database — customer data source
  • Creation date — when it was created

Creating a Script​

Step 1: Basic Information​

  1. Click "Create" → select type (campaign/IVR)
  2. Fill in name and description
  3. Select subscriber group (database with phone numbers)
  4. Select database (customer information source)

Step 2: Script Training​

Option A: Training with operator

Why training calls are needed:

  • System training — AI studies your communication style, intonation and dialogue management
  • Knowledge base creation — system remembers typical customer questions and your answers
  • Business adaptation — script adapts to your business specifics and terminology
  • Quality improvement — more examples, better system imitation of your communication
  • Logic testing — you can check how system reacts to different situations

How it works:

  1. Select internal number for calls
  2. Make training calls to customers (recommended 10-20 calls)
  3. System records and analyzes your dialogues
  4. Based on analysis, automatic script is created that imitates your communication style

Option B: Audio upload

How to properly prepare audio files:

  • Recording quality — use good microphone, avoid background noise
  • Dialogue structure — each file should contain complete dialogue from start to finish
  • Naturalness — speak normally, don't read from paper
  • Variety — include different customer types: interested, doubtful, refusing
  • Duration — optimally 2-5 minutes per dialogue, no more than 10 minutes
  • Quantity — recommended 10-20 quality dialogues

System file requirements:

  • Format: WAV
  • Size: up to 50 MB per file
  • Quality: minimum 16 kHz, 16 bit (recommended 44.1 kHz, 16 bit)
  • Encoding: stereo
  • Content: only voice dialogues, no music or sound effects

Upload process:

  1. Prepare audio files according to requirements
  2. Upload files to system through interface
  3. Wait for processing
  4. Check speech recognition quality

Step 3: Logic Configuration​

Important: Dialogue nodes are automatically generated by system based on your training calls or uploaded audio files. You cannot manually add or delete nodes — they are created by AI based on analysis of your dialogues.

What can be configured:

  1. Review created logic — system automatically creates nodes based on your examples
  2. Configure node content — edit text that system will speak in each node
  3. Add actions — configure transitions between nodes, conditions and additional actions
  4. Test script — check created logic work

What cannot be done:

  • ❌ Manually add new nodes
  • ❌ Delete existing nodes
  • ❌ Change dialogue structure (node order determined automatically)

What can be done:

  • ✅ Edit text inside nodes
  • ✅ Configure transitions between nodes
  • ✅ Add conditions and additional actions
  • ✅ Test and correct logic

Node Actions​

Each script node can perform various actions during conversation with customer. You can configure multiple actions for one node, and they will be executed sequentially.

1. Voice Information​

What it is: Main action of each node — system speaks text to customer.

Why needed: This is what customer hears when reaching the node. Without this action, node won't work.

How to configure:

  • Write text that bot should speak
  • Use fields for message personalization

Usage examples:

  • Welcome to our company! — regular greeting
  • Hello, #Name! Glad to hear you — personalized greeting
  • Your order *#OrderNumber* is ready for pickup — order information

2. Send SMS​

What it is: Sending SMS message to customer directly during conversation.

When to use:

  • Send order confirmation
  • Forward important information
  • Send link to website or app

How to configure:

  • Write SMS text
  • Choose whether to ask customer permission
  • Add text bot will speak when sending (if needed)

Examples:

  • Your order #OrderNumber accepted. Wait for manager call
  • Payment link: #PaymentLink
  • Thank you for contacting us! We'll call you within an hour

3. Transfer to Operator​

What it is: Transferring call to live operator or manager.

When to use:

  • Customer asked complex question
  • Need specialist consultation
  • Customer wants to speak with manager

How to configure:

  • Select operator or department for transfer
  • Decide whether to ask customer permission
  • Write text bot will speak during transfer

Example situations:

  • Transfer to tech support for technical questions
  • Transfer to sales manager when wanting to buy
  • Transfer to accounting for payment questions

4. Send Email​

What it is: Sending email to customer during conversation.

When to use:

  • Send contract or documents
  • Forward detailed information
  • Send instructions or manuals

How to configure:

  • Write email subject
  • Create email text
  • Select sender mailbox
  • Decide whether to ask customer permission
  • Save as template — you can save email for reuse

Examples:

  • Sending contract for signing
  • Forwarding product catalog
  • Sending usage instructions

Working with templates:

  • Saved templates can be used in other scripts
  • You can choose from ready-made templates when creating email
  • Templates help standardize customer correspondence

5. Working with Data​

What it is: Automatic processing and saving of customer data using AI during conversation.

When to use:

  • Need to create new database record
  • Required to update existing customer data
  • Need to extract structured information from speech
  • Need to record customer attitude to offer

Action types:

Record Creation​

  • Create new record - NER - value — add new record to information database
  • Supplement record created in node — supplement previously created record with new data

Data Update​

  • Enter data - NER - value — update data in database using AI to determine value
  • Enter data - NER - identifier — update data in database using AI to identify record
  • Enter data - Tonality — update data with simple values: yes/no

Data Retrieval​

  • Get data - NER - identifier — find existing record by extracted data
  • Get data - NER - identifier (Continue node) — continue working with previously found record

How to configure:

  1. Select action type from list of available operations
  2. Select database:
    • Customer database — specific customer data
    • Information database — reference information
    • NER — special database for entity extraction
  3. Select target field — where to save data
  4. Configure data extraction:
    • NER types — organizations, addresses, names, money, dates, numbers
    • Tonality — yes/no for simple responses
    • Custom fields — your own data types
  5. Specify search field if needed — for operations with existing records

Usage examples:

Creating new record:

  • Customer names their company → system creates new record in information database
  • Customer provides address → added to address database

Updating customer data:

  • Customer provides new phone → "Phone" field updated in their record
  • Customer specifies order amount → "Order amount" field updated

Working with tonality:

  • "Do you agree to installment?" → "yes" or "no" saved
  • "Are you interested in additional service?" → customer attitude recorded

Retrieving information:

  • Customer names company → system finds existing record
  • Customer provides order number → system loads order data

Work features:

  • AI automatically extracts needed information from customer speech
  • System understands context and can work with incomplete data
  • Data saved in real time during conversation
  • Can combine different operation types in one node
  • System tracks connections between nodes for operation continuation

Message Personalization​

You can make messages personal using customer data and information from your databases.

How it works​

Customer fields (#) — information about specific customer:

  • Name, surname
  • Phone, email
  • Data collected in previous calls

Database fields (*) — reference information:

  • Service prices
  • Office addresses
  • Product descriptions
  • Any other information

Telegram bots (@) — links to your bots:

  • For channel subscription
  • For receiving notifications
  • For communication via messenger

Usage examples​

In greeting:

  • Hello, #Name! — system will say "Hello, John!"
  • Welcome, #Name #Surname! — "Welcome, John Smith!"

In order information:

  • Your order *#OrderNumber* is ready — "Your order №12345 is ready"
  • Cost is *#Price* rubles — "Cost is 5000 rubles"

In SMS:

  • #Name, your order #OrderNumber accepted — "John, your order №12345 accepted"
  • Payment link: #PaymentLink — will insert real link

With Telegram bots:

  • Subscribe to our channel @BotName — will add bot link
  • Receive notifications via @NotificationsBot — notification bot link

How to Work with Actions​

Adding Actions​

  1. Open node for editing
  2. Click "Add Action"
  3. Select needed action type
  4. Fill required fields
  5. Save changes

Editing Actions​

  • Click on action in list
  • Change needed parameters
  • Save changes

Deleting Actions​

  • Click delete button next to action
  • Important: "Voice Information" action cannot be deleted — it's required for each node

"Working with Data" Action Features​

  • System tracks sequence — some actions available only after others are executed
  • Can create new NER types — click "+" button to add special fields
  • Actions linked between nodes — system remembers created records for subsequent use
  • Automatic context determination — AI understands which record data belongs to

Execution Order​

Actions are executed in order they were added to node. Order can be changed by dragging actions in list.

Troubleshooting​

Script Won't Create​

Reasons: no subscriber group selected, internet problems, insufficient examples Solution: check all fields, ensure connection stability, add more examples

Insufficient Training Calls​

Problem: system requires minimum 10-15 quality dialogues for creating effective script Solution:

  • Make more training calls with different customers
  • Conduct dialogues naturally, as you usually communicate with customers
  • Cover different scenarios: interested customers, doubtful, refusing
  • Use real situations from your business

Script Shows Error​

Actions: read error description, check subscriber group settings, ensure database correctness

Audio Upload Problems​

Common problems:

  • Poor recording quality — re-record with better microphone
  • Background noise — use noise reduction or re-record in quiet place
  • Incorrect encoding — ensure stereo format

Solution: check file compliance with system requirements, reconvert or re-record if necessary

Can Active Script Be Changed?​

Limitations: active scripts can be edited carefully, changes may affect campaign work

Cannot Add/Delete Nodes in Script​

Explanation: Dialogue nodes are automatically created by AI based on analysis of your training calls or audio files. This is done so system can accurately reproduce your communication style.

What to do:

  • If need to change logic — add more training examples with needed scenarios
  • If need to fix text — edit content of existing nodes
  • If need to add new transitions — configure conditions and actions between nodes
  • If script doesn't fit — create new one with more examples

Node Action Problems​

Action Not Executing​

What to check:

  • Are all required fields filled (e.g., is operator selected for transfer)
  • Did customer give consent for action execution
  • Are external services working (SMS, email)

What to do:

  • Check action settings in node
  • Ensure all fields are filled
  • Check SMS gateway and mail server operation

Fields Not Substituting in Text​

Reasons: incorrectly written fields or they don't exist in database

What to do:

  • Check field writing correctness (#Name, *Price)
  • Ensure fields exist in subscriber group or information database
  • Check field name accuracy

SMS/Email Not Sending​

Reasons: mailboxes or SMS gateways not configured, customer didn't give consent

What to do:

  • Configure mailboxes in personal account
  • Check SMS gateway settings
  • Ensure customer consented to sending

Transfer to Operator Not Working​

Reasons: operator unavailable, queue not configured, customer didn't give consent

What to do:

  • Check selected operator availability
  • Ensure queue is configured correctly
  • Check that customer consented to transfer

System Not Collecting Needed Information​

Reasons: incorrectly configured information types, few examples in training data

What to do:

  • Check extraction type settings (names, amounts, dates, etc.)
  • Add more examples to training calls
  • Configure special fields for your information

"Working with Data" Action Problems​

Action Not Available in List​

Reasons: preliminary actions not executed, incorrect node sequence

What to do:

  • Ensure all necessary preliminary actions are executed
  • Check node sequence in script
  • Create needed records in previous nodes
Data Not Saving​

Reasons: incorrectly selected database, target field not configured

What to do:

  • Check database selection (customer/information/NER)
  • Ensure correct target field is selected
  • Check database access rights
AI Incorrectly Extracting Data​

Reasons: incorrectly selected NER type, insufficient examples in training data

What to do:

  • Check NER type correspondence to extracted information
  • Add more examples to training calls
  • Create special NER types for specific information
Errors When Working with Existing Records​

Reasons: incorrectly configured search field, record not found

What to do:

  • Check search field settings for record lookup
  • Ensure record exists in database
  • Check data accuracy for search

Tips​

Creating Quality Scripts​

  • Use clear names and descriptions
  • Test on real dialogue examples
  • Group by types (campaigns/IVR)

Working with Automatically Created Nodes​

  • Remember: nodes are created by AI, cannot be added/deleted manually
  • Plan ahead — include all needed scenarios in training calls
  • Be consistent — use same communication style in all examples
  • Cover different situations — more variants, better system creates nodes
  • Test result — check how system reproduces your logic

Effective Training Calls​

  • Prepare in advance — think through main communication scenarios
  • Be natural — speak normally, don't try to "play role"
  • Cover different situations — successful sales, refusals, customer questions
  • Use real data — call real numbers from your database
  • Record results — note which approaches work better

Preparing Quality Audio Files​

  • Use professional equipment — good microphone and quiet room
  • Plan structure — each dialogue should have clear beginning and end
  • Test quality — check recording before uploading to system
  • Organize files — give clear names to files (e.g., "Sale_customer_interested.wav")

Working with Node Actions​

Action Planning​

  • Think through sequence — which actions should execute in what order
  • Use confirmations — for SMS, email and transfers include consent request
  • Test integrations — check mailbox and SMS gateway operation

Text Optimization​

  • Personalize messages — use customer fields to create individual messages
  • Make texts natural — avoid robotic phrases
  • Check length — SMS should be brief, email — informative

Collecting Customer Information​

  • Configure correctly — choose suitable types for extracting needed information
  • Use special fields — for specific information of your business
  • Check results — configure correct processing of customer responses

Transfer Management​

  • Choose suitable operators — consider specialization and workload
  • Prepare context — pass all collected information to operator
  • Plan fallback options — what to do if operator unavailable

Working with Data​

Data Operation Planning​
  • Think through sequence — first create records, then update them
  • Use correct NER types — choose suitable types for extracted information
  • Plan node connections — system remembers created records for subsequent use
Data Extraction Optimization​
  • Use specific NER types — create your own types for your business
  • Test extraction — check how AI understands your data
  • Add context — more examples, more accurate extraction
Database Management​
  • Separate data types — use customer database for personal data, information database for references
  • Check integrity — ensure search fields are configured correctly
  • Document structure — keep track of created fields and their purpose

Work Organization​

  • Plan scripts in advance
  • Use templates of successful scripts
  • Analyze results in statistics and optimize