Most of what SuperCognit does is about letting an AI agent answer well. This post is about the opposite: conversations where the right number of AI replies is zero.
There are good reasons for that. A patient writing to a clinic about results wants a person, not a well-phrased guess. A law firm cannot let software say anything that sounds like advice. A private bank's best clients pay, in part, for a name they can write to. And a company's own field technicians, stuck at a site with a question, need a colleague, not a chatbot.
These businesses still want what a good chat gives: one place where messages from the web, WhatsApp and Telegram arrive, teammates who share the work, hours, files and history. That is what a person agent is.
What a person agent is
When you create an agent, there is a third option next to the AI ones: "A person answers". It asks for a name and an intro and nothing else, because there is no prompt to write. Every message on its chat page, widget and channels lands in the Inbox, and your team replies from there.
- No AI ever replies. Not when the team is slow, not outside hours, not when a message looks like something the AI could handle.
- The AI settings are out of the way. Behavior, Knowledge, Media, Tools and Guardrails are hidden from the editor.
- The chat does not call itself an AI. The footer drops the "AI chatbot" wording, the message box says "Write a message" instead of "Ask anything", and there is no "Talk to a person" button, because there is no AI to hand over from.
- The widget shows the team's real status, for example "Team away · we'll reply when back".
- Voice notes and files are kept as they are. Voice notes are not transcribed; the team listens, as they would on their own phone.
AI agent or person agent?
If most questions are routine and a few need a person, an AI agent with hand-over to your team is the better fit. A person agent is for when the answer has to come from a person every time, or when using AI at all is the problem.
Hours and teams
On the agent's Team tab you choose whether conversations are auto-assigned to the least busy teammate who is available, and set business hours. Outside them, messages wait in the Inbox and the sender is told when you are back. You can write your own waiting and out-of-hours messages, or keep the defaults in the visitor's language.
If you use teams, pick the team that gets this agent's conversations — a front desk, a legal intake team, a client service desk — and its hours, chat limits and waiting line apply.
Staff only: a chat for your own people
Some person agents are not for the public at all. A help line for your field technicians, or a desk for your own branch staff, should answer only the people on a list. Under Access → Who can chat, turn on "Staff only" and list who may use it:
- Allowed WhatsApp numbers, with country code.
- Allowed email addresses, on the web chat. People sign in with an email link or Google.
- Allowed email domains, on the web chat: anyone with a proven address at your company's domain. Subdomains need their own line.
The list is checked before anything else runs. With email addresses or domains on the list, the chat page opens behind a sign-in and the sign-up window reads "Use your work email to chat with the team". Someone who signs up with a password has to confirm their email before their messages go through. While Staff only is on, the widget and Telegram are closed, and people who are not on the list get your own message, or a short default, instead of an answer.
Quiet by default
A new person agent starts hidden from search engines. Many of these chats are for existing patients, clients or staff, not for strangers who find it on Google. You can turn search-engine indexing on, on the agent's General tab, if you want it found.
Set it up
Create the agent
On the new-agent page, choose "A person answers", and give it a name and an intro. An existing AI agent can switch too, under General → Who answers.
Set hours and assignment
On the Team tab, turn auto-assign on or off, set business hours, and pick a team if you use them.
Decide who can chat
Leave it open to everyone, or turn on Staff only under Access and list the numbers, emails and domains.
Connect channels and share the link
Connect WhatsApp or Telegram on the Channels tab, add the widget to your site, or share the chat page.
Switching between "AI agent" and "A person answers" keeps every AI setting you had, so you can switch back later and find the prompt, knowledge and tools where you left them. The agent type is also in the API and the platform MCP tools, so you can create a person agent from your own systems.
Frequently asked questions
Can I have a business chat with no AI at all?
Yes. Create an agent with "A person answers". Every message on its chat page, widget, WhatsApp and Telegram goes to your team's Inbox, and no AI ever replies, including outside business hours.
Will visitors see that it is an AI chatbot?
No. A person agent drops the AI wording: the footer does not call it an AI chatbot, the message box says "Write a message", and there is no "Talk to a person" button.
When should I use a person agent instead of an AI agent with hand-over?
When every answer has to come from a person — health, legal, private client service — or when using AI is the problem. If most questions are routine, an AI agent that hands over to your team answers faster.
Can I limit the chat to my own employees?
Yes. Under Access → Who can chat, turn on Staff only and list allowed WhatsApp numbers, email addresses or email domains. People on the web chat sign in with a proven email; everyone else gets your message instead of an answer.
Can I switch an AI agent to a person agent and back?
Yes, under General → Who answers. Your AI settings are kept, so switching back restores the prompt, knowledge and tools.
