Turn on your own AI Receptionist
Outcomes
Three parts make a receptionist
Every AI Employee, including the AI Receptionist, is built from the same three parts. Configuring yours means deciding who it is, what it can do, and what it knows. Flip each one:
Your receptionist serves your business, so it lives in your workspace: Partner Center. Everything in this step happens there.
Meet your workforce
Open AI → Workforce. The page that loads is your AI Workforce, and every AI Employee you have sits on it as a card.
You already have more than one. The Chat Receptionist answers anyone messaging your business. Ash is the voice receptionist and answers the phone. The AI Sales Assistant works on your recorded sales calls. Alongside them sit any custom employees, and a Create button in the top right for building your own.
Two things on every card are worth reading before you click anything:
- The buttons differ by employee. The Chat Receptionist offers
ChatandConfigure. Ash offersCall,Try it, andConfigure, because you test a voice receptionist by phoning it. - Autonomy is shown on the two receptionist cards as a level, and it is a label rather than a control. The info icon beside it explains the four levels; there is nothing to set. Custom employees show no autonomy at all.
The Starter plan includes the AI Chat Receptionist, which is the one this step configures. The Professional plan and above add the AI Voice Receptionist and the social channels, Facebook, Instagram, and WhatsApp. Two gates run in parallel and it is worth keeping them apart: your own subscription decides what you get, and a client gets whichever edition of Conversations AI is active on their account. This guide breaks down what each plan includes.
Labfind your receptionist
Go to AI → Workforce.
Name it, and read what it already does
The Configure page is one screen with everything on it. On the left: the photo, the Name field, the Autonomy level, and a View prompt button. On the right, in order: Triggers, Profile, Capabilities, Knowledge sources, and Automations.
Start with the name, because the AI knows its own name and introduces itself with it. Then read before you write, because two things are already done for you.
Under Capabilities you will find Capture leads: "Collects contact information, creates a new contact, and lets the contact know someone on your team will follow up." That is the capability that turns a chat into a record in your CRM, and it is on before you touch anything.
Labmake it yours, then read the role
Go to the receptionist's Configure page.
Reading first is the habit worth forming. The role that ships is a working one, and most first-day changes are small additions to it rather than rewrites.
Give it knowledge
Knowledge sources on the same page is where you choose what the AI can pull from when it answers. Your website homepage is already one of them, included by default, which is why the receptionist can say something useful before you configure anything at all. Every source lists its address, when it last refreshed, and how many pages it trained on, so you can always see exactly what the AI is working from.
Adding one takes two clicks rather than one. The Add knowledge field first offers knowledge you already have; + Add new knowledge at the top of that list is what opens the choice of three:
- Text for content related to your business, such as FAQs or custom information
- Website to retrieve content from a website
- File Upload for a document or file
Labadd the rest of your website
Go to the same Configure page, under Knowledge sources.
Anything that lives nowhere on your site (your hours, your prices, what makes you different) goes in through Text as a written entry, and a price list or a policy document goes in through File Upload. For every source type, crawl mode, and refresh option, the Knowledge Base guide covers it, and AI → Knowledge Base is where you manage sources across every employee at once.
Put it to practice
Chat on the receptionist's card opens a conversation between you and your employee. It is private to you and it keeps a history, and it is where you try your receptionist out: you chat with it, you audit the answers, and you fix what is off. There is no separate test page to find.
Labinterview your receptionist
Go to AI → Workforce, then Chat on the Chat Receptionist card.
Change one element at a time, whether that is the role, a capability, or a knowledge source. Changing three at once tells you nothing about which one worked.
Put it on your website
The widget does not add itself. Installing the code is the step that turns a working configuration into a receptionist your visitors can reach, and it is also what connects the two halves of this lesson: a widget you install shows up back on the employee as a trigger, one of the events that invokes it.
Labcreate a widget you can embed
Go to Conversations → More → Conversations Settings, on the Web Chat card. (Administration → Conversations Settings reaches the same page.)
Widget type reading Standard and your receptionist assigned to it.The widget that ships with your account has a Widget type of System, and its ⋮ menu offers only Edit. Embed codes live on Standard widgets, which is why the step above creates one.
Labembed it on your site
Go to your new widget's ⋮ menu on the Web Chat configuration page.
Conversations, and the contact is saved to your CRM.Go back to the receptionist's Configure page and look at Triggers. Your widget is listed there now, as an event that invokes this employee, with its own Settings. That is the shape of the whole system: a channel arrives, a trigger fires, and the employee answers with the role, capabilities, and knowledge you just set.
Open Conversations and find the chat you just had through your own widget. Whatever a visitor sends now lands here, and this is the inbox you will check in the morning.
Everything deeper, voice configuration, custom capabilities, and building receptionists for clients in their Business App, has a path of its own: Hire your first AI Employee.
Knowledge Check
Three quick questions on finding your workforce, the three parts, testing, and getting the widget live.