Skip to main content

Conversational AI Agents

D
Written by Denise Abdullah
Updated this week

Release Announcement: Conversational AI Agents (Early Access)

Release Date: March 2026

Availability: Early access — contact your account manager to enable this feature for your organization.

Overview

Moja now supports Conversational AI Agents — voice-based AI that can answer inbound calls, have natural conversations with callers, and collect structured data as tags. When the agent has gathered the information it needs, it hands the call off with the captured data attached for routing and reporting.

Why This Matters

Collecting caller information before routing has traditionally meant IVR menus or live agents. Conversational AI Agents add a third option: a natural voice conversation that gathers the data you need — without the rigidity of keypress menus or the cost of a human on every call.

The agent collects information as structured tags, which means the data flows directly into your existing routing logic and call reporting.

What's Included

AI-Powered Tag Collection

Define the tags you want collected — caller name, zip code, service type, or any custom field. The AI agent asks callers naturally and captures their responses as structured key/value pairs. You provide descriptions for each tag to guide the agent's questions.

Multiple LLM Model Options

Choose from 40+ language models across providers including OpenAI, Anthropic (Claude), Google (Gemini), and others. The model picker displays estimated latency and cost per minute to help you balance conversation quality with response speed.

Agent Handover with Captured Data

When the agent finishes collecting information, it triggers a handover that passes the captured tags along with the call. The data is available for downstream routing decisions and appears in your call records.

Tag Visibility in Call Life Log

Agent-captured tags appear directly in the Call Life Log. You can see exactly what the AI collected during the conversation and what was applied to the call — displayed as a clear key/value table with both the agent's raw captures and the final applied tags.

Creating an Agent

  1. Navigate to Agents in the sidebar

  2. Click Create Agent

  3. On the Agent tab, fill in the following:

    • Name — a descriptive name for your agent

    • Description — internal notes about this agent's purpose

    • First Message — the greeting callers hear when the agent picks up (e.g., "Hi, thanks for calling! How can I help you today?")

    • System Prompt — instructions for how your agent should behave, what to ask, and how to respond

    • Language — select from 30+ supported languages

    • Collection Tags — the data you want collected from callers. Each tag has a name and optional description to guide the agent

    • Voice — choose from a library of natural-sounding voices

    • TTS Model — select a text-to-speech model family

    • LLM Model — choose the AI model powering the conversation (40+ options with latency and cost estimates)

    • Temperature — adjust how creative vs. consistent the agent's responses are

    • Max Duration — set the maximum call length (default: 10 minutes)

  4. Under Advanced Settings, optionally configure:

    • Turn Eagerness — how quickly the agent responds (Eager, Normal, or Patient)

    • Interruptible — whether callers can interrupt the agent mid-sentence

    • Stability and Speed — fine-tune voice output

    • Audio Tags — vocal characteristics like accents or tone

  5. Use the Knowledge Base and Tools tabs if needed for your use case

  6. Save the agent

Adding an Agent to a Call Flow

Once your agent is created, work with your account team to add it to a call flow. Your agent can be used as a primary destination, a fallback for missed calls, or an after-hours handler within your existing routing logic.

Getting Started

  1. Request access: Contact your account manager to enable Conversational AI Agents for your organization

  2. Create your agent: Follow the steps above to configure voice, prompt, tags, and model

  3. Test: Place test calls to verify the agent collects the right data and handles conversations as expected

  4. Review: Check the Call Life Log to confirm captured tags and handover data appear correctly

  5. Go live: Work with your account team to add the agent to your call flow

Example

An insurance network handling 500+ daily calls configures an AI agent with three collection tags: Insurance Type, Zip Code, and Current Coverage Status. When a caller reaches the agent, it has a brief conversation to gather this information, then hands the call off with the captured data as tags — visible in the Call Life Log and available for routing. No IVR trees, no hold queues.

Questions?

Reach out to your account manager or contact support for help getting started with Conversational AI Agents.

Did this answer your question?