AI Agent Wizard and Builder
AI Agents handle phone conversations for your firm. They can answer inbound calls, place outbound campaign calls, collect task answers, transfer callers, send follow-up messages, book appointments, and summarize calls afterward.
The AI Agent Wizard helps you create a strong first draft. The full Agent Builder is where you test, refine, publish, and maintain the agent over time.
When to Use the Wizard
Use the Wizard when you want guided setup. It asks for the agent's purpose, recommends a template, fills in common defaults, and creates a draft agent that you can review before publishing.
Start here:
- Go to Agent Builder.
- Click Create Agent with Wizard.
- Answer the onboarding questions.
- Review each Wizard screen.
- Click Create Draft Agent.
- Test and publish from the full Agent Builder.
Wizard Onboarding
The Wizard starts with three setup choices:
- Practice areas — the legal or operational area the agent supports.
- Contact type — whether the agent talks to new leads, existing clients, or inbound callers.
- Primary goal — the job the agent should complete, such as qualifying a lead, collecting case information, scheduling an appointment, connecting to an attorney, or answering and routing inbound calls.
These answers shape the recommended template, generated prompt, transfer rules, welcome messages, and available setup steps.
Wizard Screens
Name & Voice
Set the internal agent description, agent purpose, language, and voice. Use a name your team can recognize later, such as the case type, campaign type, or intake workflow.
Call Criteria
This screen defines what the agent should collect or verify. Depending on the goal, you can connect a form or provide free-text qualification guidance.
When you connect a form, the agent can follow the form's pages, questions, conditional logic, and disqualification rules during the call. This is the best option when the conversation needs to fill tasks or intake fields.
Actions
Choose what the agent can do during or after the call:
- Live transfer to a team member or intake line.
- SMS follow-up using a message template.
- Email follow-up using a subject and body template.
- Appointment booking through a connected scheduling integration.
Transfer settings can include cold or warm transfer, human detection, transfer triggers, and business-hours availability.
Welcome Messages
Configure what the agent says at key moments:
- outbound welcome
- bad timing response
- voicemail script
- inbound welcome
- after-hours message when configured
Message fields can use variables such as , , , or .
Review & Create
The Review screen summarizes the agent description, selected voice, call criteria, actions, welcome messages, and knowledge base status. Use the edit links to revisit any step, then create the draft agent.
Working in the Full Agent Builder
After the draft is created, Kayse opens the full Agent Builder. This page gives you the advanced controls that the Wizard intentionally keeps out of the first setup flow.
Use the Builder to manage:
- the agent prompt and conversation rules
- connected forms and task collection
- knowledge bases
- call transfer, SMS, email, and appointment booking
- phone number assignments
- voice, language, responsiveness, interruption sensitivity, and call settings
- voicemail behavior
- background sound
- post-call statuses and variables
- test calls and transcript history
- version history and publishing
Forms, Tasks, and Mapping
Form-based agents use a form to guide the conversation. During a call, the agent asks the form questions, follows conditional logic, detects disqualifying answers, and saves answers back to the related tasks.
Mapped task answers can become Kayse custom fields or CRM fields when mapping is configured. This is how information collected by an AI Agent can become structured data for reporting, workflows, and CRM sync.
See Smart Tasks & Forms in Kayse for the task setup flow.
Test Calls
Always test an agent before publishing.
- Open the Test Call panel in Agent Builder.
- Choose a testing number.
- Enter the phone number to call.
- Fill any variables needed for the test.
- Click Test.
- Review the live transcript, call summary, sentiment, saved fields, task outcomes, and disconnection reason.
If the agent has a connected form, test calls also show task outcomes so you can confirm whether questions were completed, skipped, or left incomplete.
Editing and Updating Agents
Agents save changes as drafts until you publish them. You can open an agent from the Agent Builder list, adjust the prompt or settings, run test calls, and publish when ready.
When a form-based agent uses a system-generated form prompt, Kayse can detect when the linked form changed and offer to rebuild the prompt from the latest form. Your other settings, such as voice, phone numbers, and transfer behavior, stay in place.
Publishing and Version History
Publishing makes the current draft live for calls handled by that agent. Kayse saves a version snapshot each time you publish, including the prompt, form connection, voice, phone numbers, transfer settings, messages, and other key settings.
Use version titles and descriptions so your team can understand what changed. If you need to review or restore an older setup, open the agent history and load a previous version.
Additional Settings
After the core setup is working, review these advanced settings:
- Phone numbers — assign inbound, outbound, or both directions.
- Call settings — control max call duration, silence timeout, ring duration, and noise handling.
- DTMF — allow callers to press keypad options during a call.
- Voicemail — choose whether the agent hangs up, leaves a message, or plays static text.
- Post-call analysis — configure statuses and variables that summarize the call result.
- CRM variables — use CRM and case data inside prompts and post-call workflows.