Admins create AloAi Voice Agents from the Agents menu. You choose whether the agent receives or places calls, pick a starting point, and answer a short set of setup questions that Aloware uses to write the agent’s instructions.
This article walks through each step, from choosing a direction to publishing the agent.
Choose the direction and starting point
Open the Agents menu in your Aloware admin account.
The Agents menu in the left navigation.
Click New agent and select Voice. The Create agent screen opens.
The New agent button on the Agents list.
The Create an agent window with the Voice option.
Under Direction, select one:
Inbound: the agent answers calls for Q&A, support, and other use cases.
Outbound: the agent places calls for initial reach-out to your contacts.
Direction options on the Create agent screen.
Under Templates, select a starting point:
Blank agent starts from scratch. The agent is created right away and opens in its configuration screen, without Guided Agent Setup.
Create your agent using AI asks you a few questions in Guided Agent Setup and writes the agent's behavior for you.
A template starts from a predefined setup for a common business scenario. The templates shown depend on the direction you selected.
Click Create agent.
The direction you select sets how the agent opens and closes a call and which settings appear in the agent editor. For the differences, see Getting Started with AloAi Voice Agents.
Templates on the Create agent screen.
Starting from a template
Templates are built around common business scenarios. A template can include the agent's instructions, conversation behaviors, tools, voice, and model settings, along with a default model tier.
Use the category tabs to narrow the list. For outbound agents, the tabs are All, Booking, Candidate Screening, Lead Qualification, and Outbound Sales. Each template card shows its name and category, for example Generic Appointment Booking under Booking.
Template category tabs on the Create agent screen.
Before you create the agent, you can review a template's details to see the purpose, conversation behaviors, tools, and voice it includes.
When you create an agent from a template, the template's configuration is copied into the new agent. The agent is not linked to the template:
You can change any setting after the agent is created.
Later changes to the template do not update agents already created from it.
Creating your agent using AI
Create your agent using AI is for admins who know what the agent should accomplish but are not sure how to configure it. You describe your business, use case, and goals, and the setup uses your answers to write the agent's behavior.
The Create your agent using AI option.
Complete Guided Agent Setup
Guided Agent Setup collects the essentials in order. A step rail shows where you are: Audience & timing, Knowledge, Voice & greeting, and Review. You can move back to any earlier step without losing your answers.
Who does it talk to, and when?
This step collects four things.
Purpose of this agent: describe the agent's purpose and objective in 2 to 3 sentences, up to 600 characters. The field shows an example. Required.
Who is calling in: describe who the agent speaks with and where the calls come from, such as existing customers or new leads, up to 400 characters. Required.
Line assignment: select one of your organization's lines, or leave it as No line assigned. An agent without a line cannot take live calls.
Operating hours: choose 24/7 or Business hours. If you choose Business hours, set the hours on the Lines & scheduling tab after the agent is created.
Next is unavailable until Purpose of this agent and Who is calling in are filled in. Both answers are used to write the agent's instructions.
The Audience & timing step.
What does it need to know?
Add the information the agent uses to answer callers, such as price lists, service descriptions, and policies. Upload files, or enter a web page address in Add a web page and click Add page. The agent indexes only the page at that address, not the whole site. Files can be PDF, Word, Excel, PowerPoint, CSV, text, Markdown, HTML, XML, or JSON, up to 50 MB each. Sources you add here are attached to the new agent.
This step is optional. To continue without a source, click Next. The review step shows that no sources were added. You can add sources later from the agent's Knowledge Base panel.
The Knowledge step.
How your sources become a knowledge base
The Knowledge Base Builder turns your sources into a facts-only knowledge base the agent answers from. It separates facts about your business from instruction-style content, such as persona, tone, call flow, objection responses, guardrails, greeting and goodbye, tool rules, and phrasing scripts. That content is held back and shown to you in a report, so nothing is discarded without you seeing it. This separation applies to uploaded documents and typed text.
The result you review includes:
A facts table index at the top, with topic sections below it. Section headers are phrased the way a caller would ask the question.
The source behind every fact the builder keeps.
Facts that disagree across your sources, shown side by side for you to settle.
Gaps, where your sources do not cover a fact. The builder never fills a gap with invented content.
Sources it could not read, listed as unreadable with their origin named.
A proposed split between the high-frequency facts suggested for the agent's instructions and the rest, which the agent looks up when a caller asks.
A verify-against-source list to work through before the agent goes live.
Each fact appears once, facts that change over time are dated, and claims that come only from your own marketing copy are labeled that way. Your sources keep their separate structure, which keeps lookups quick during a call.
To finish, review the result, settle conflicting facts, edit anything that needs correcting, work through the verify-against-source list, and approve it. Nothing attaches to the agent until you approve it.
How should it sound, and what does it say first?
Under Voice, preview the voices and select one. Each voice shows its accent and gender.
Review the Opening greeting. It is prefilled with a default, for example "Hi, how can I help you today?" Keep it or rewrite it, up to 450 characters. The agent speaks the greeting as written at the start of a call.
Under Communication style, select Professional, Professional + conversational, or Default communication style. Professional is selected by default. The style applies the matching Agent Handbook behaviors, and AI disclosure and scope boundaries apply under every style.
The Voice & greeting step with the communication style options.
Here's your agent: review and publish
The final step, Here's your agent, summarizes the setup: purpose, who calls in, direction, line, operating hours, knowledge sources, voice, opening greeting, and communication style.
Settings you left empty stay in the summary in muted text, for example "No line assigned" or "None yet, add them later." Click Edit next to a row to return to that step. Your other answers are kept. Direction has no Edit link because it is set on the Create agent screen.
Click Publish agent when you are ready. You can publish with any combination of set and empty settings. The agent opens in its configuration screen, where you can edit any setting.
The review step with the Publish agent button.
Moving through the setup
You can move through the setup in these ways:
Use the step rail to return to an earlier step. Your answers are kept. If a change makes a later answer no longer apply, you are warned before it is discarded.
Use the skip link to jump to the full configuration screen at any point. Your answers so far carry over.
If you exit the setup before publishing, your progress is discarded. Nothing is saved automatically.
An agent created through Guided Agent Setup is the same as one built on the full configuration screen, so every setting stays editable after you publish.
The instructions Aloware writes for you
When you finish the setup, Aloware writes the agent's instructions from your answers. The instructions are organized into standard sections: Identity/Role, Goal, Conversation steps, Objections, Guardrails, and Tools.
The greeting, goodbye, and pronunciation are saved as agent settings, not written into the instructions.
Rules the platform already applies on every call, such as keeping replies brief, are not repeated in the instructions. Speech normalization is controlled by the Normalize Text for Speech setting in Speech Settings, which is on by default.
The instructions appear in the agent's instructions box, where you can edit them.
If you regenerate the instructions, you are warned before your edits are overwritten.
If the use case has many conditional paths or actions, a warning appears. The instructions are still written.
To review and refine the instructions, see Writing instructions and setting conversation behavior.
After you publish
Place a test call before using the agent with real callers. Check the greeting, tone, routing, and overall behavior. For testing options, see Testing and reviewing your agent.
If the agent has no line yet, assign one on the Lines & Scheduling tab. See Lines and scheduling.