The words behind AI assistants, voice agents and automations, explained without the jargon. Each term links to the page where you can see it working in Mayavi.
Your company's private space in Mayavi, holding your assistants, knowledge, tools, team and settings. Agencies often run one per client.
Short for "where is my order", the most common question online stores receive. AI agents can answer it from live order data.
Meta's platform for businesses to send and receive WhatsApp messages through software, needed to connect an AI agent to WhatsApp.
WhatsApp's rule that businesses can reply freely only within 24 hours of a customer's last message. After that, you need a message template approved by Meta.
A message one system sends to another automatically when something happens, such as a lead form sending new details to start an automation.
An on-screen animated face that speaks for a voice agent during web calls.
How a voice agent manages who speaks when: how quickly it replies, whether callers can interrupt and when it ends the call.
The event that starts an agent or automation, such as a missed call, a new order, a schedule or a webhook.
A connection to an app, such as Shopify or Google Calendar, that lets an AI take action instead of only talking.
A to-do item tied to a customer for anything that can't be solved in one conversation, such as a refund to process or a call-back to make.
Technology that turns written text into a spoken voice. Voice agents use it to talk to callers.
A ready-made assistant, agent or automation for a common job that you install and adjust. A bundle installs several that work together.
The written instructions that tell an AI who it is, what to do and what not to do.
How quickly a business contacts a new lead after they get in touch. Faster contact usually means a better chance of winning the customer.
Technology that turns spoken words into written text. Voice agents use it to understand callers.
Running software in your own cloud or on your own servers instead of the vendor's, for tighter control over data and security.
Short for retrieval-augmented generation. The AI first searches your knowledge for relevant passages, then writes its answer from them, which keeps replies grounded in your facts.
The share of conversations an AI finishes without handing over to a person. A rising rate usually means your knowledge is getting better.
A voice technology where the AI listens and speaks almost instantly, making conversations feel natural instead of stop-start.
A tool that writes or rewrites an AI's instructions for you from a short plain-language request.
Automatically hiding sensitive details, such as card numbers, emails or phone numbers, so they don't appear where they shouldn't.
Serving customers across many channels with one connected system, so the same answers and history follow them from chat to phone to WhatsApp.
An AI assistant that understands and replies in more than one language, and can switch when the customer does.
An automatic text or WhatsApp message sent to someone whose call you missed, so they don't call a competitor instead. Also called missed-call text-back.
An open standard that lets AI models connect to outside tools and data in a consistent way. It makes adding new integrations simpler.
The AI model that reads and writes language. It's the "brain" behind an assistant's replies.
A named group of knowledge sources, such as "Returns policy" or "Product catalogue", that you attach to one or more assistants.
The documents and written answers an AI searches before it replies: policies, product details, opening hours. Better knowledge means better answers.
An attempt to trick an AI into ignoring its rules, such as telling it to "forget your instructions". Guardrails help an AI resist these.
One place where your team sees every chat, call and message from every channel, and can step in at any time.
A ready-made web page for your AI assistant with its own public link, so customers can chat without you adding anything to your website.
When an AI passes a conversation to a person on your team, for example because the customer asked for someone or the question needs judgment.
When an AI states something false with confidence. Grounding answers in your own knowledge and letting the AI say "I don't know" reduces it.
Safety rules that control what an AI may say and do. They can block harmful messages, hide personal details, resist jailbreak tricks and keep the AI on topic.
Training a custom AI model on your own example conversations so it picks up your tone and preferred answers. Most businesses start with good knowledge and instructions first.
A short phrase a voice agent says while it looks something up, such as "One moment, let me check that." It avoids awkward silence on a call.
Keeping data stored in a specific country or region, often to meet legal or company rules. Self-hosting is one way to control it.
A connection your developer sets up so an AI can use your own systems, by calling an API or an MCP server.
AI that holds a back-and-forth conversation in text or voice, understanding what people mean rather than matching keywords or menu options.
The chat bubble that sits in the corner of a website. Visitors click it to ask questions, and an AI assistant answers.
A place where customers reach you, such as website chat, a phone number, WhatsApp, email, Slack or Telegram.
Using your own accounts with AI or phone providers instead of the platform's. Provider usage is billed to you directly.
When a caller interrupts a voice agent mid-sentence. A good voice agent stops talking and listens, just as a person would.
A step-by-step workflow that runs the same way every time: when something happens, check a condition, ask a person if needed, then act. Each step is visible and can be tested first.
A record of who did what and when, such as who changed a setting or published an automation. Admins use it to keep track of changes.
A point in an automation where the run pauses until a person approves or rejects it. Useful for refunds, large orders or any message that should be checked first.
A secret code that lets one piece of software use another, such as a voice provider. Keys stay private and are stored encrypted.
A company of human operators who answer calls in your business name and take messages. An AI receptionist does a similar job with software instead of people.
An AI assistant that talks out loud. It listens to the caller, understands them and replies in a natural voice, on a phone line or in the browser.
An AI assistant that answers your business phone and messages, books appointments and takes details, the way a front desk would. It works at any hour and answers several callers at once.
An AI that talks directly to your customers on channels like website chat, phone and WhatsApp. It answers from your own information and hands over to a person when needed.
Software that uses AI to understand a request, decide what to do and take action, such as looking up an order or booking a slot. It goes beyond answering questions by doing real work in other systems.
One complete job done by an agent, from the moment it starts to its result. Each run is logged with its outcome, so you can see what worked and why a run failed.
An AI worker that runs in the background. It starts on a schedule, a webhook or an incoming message, looks things up in your tools, does the job and reports back to your team. Unlike an assistant, it doesn't wait for a customer to start a chat.
