How Panelbot Works
How Panelbot Works explains the trust model, the safety boundary between the AI and your servers, and where the assistant stops and a human starts. The section is written for owners and technical leads who want to understand exactly what the AI is allowed to do before they turn anything on. You will see how customer chat is validated against ownership records from your billing portal, how each operation is graded by impact, how read-only diagnostics run automatically while state-changing actions are gated, and how the highest-impact class of change is never performed by the AI at all. It covers the honest boundaries: platforms and problems the assistant handles well, cases it hands off, and the reasons behind those choices. Read this section before configuring the widget so you know what you are switching on.
What this section covers
- How Panelbot works end to end
See how you, your customer, and your servers connect in Panelbot, and how identity, ownership, and every action are checked at each layer on their own.
- What Panelbot can and cannot do
An honest look at what Panelbot handles automatically, what needs your approval, and the limits it will not cross, so you know exactly what to expect.
- Safety and permissions
The Panelbot trust model in plain terms: the AI only translates, actions come from a pre-approved list, ownership is checked in code, all logged.
- When a human takes over
Panelbot escalates to your team when it is unsure, a change is risky, an issue is out of scope, or a customer asks, so nothing risky happens unattended.
- What the AI does and does not do
The AI in Panelbot only reads the customer's message, pulls out details, and words replies. It never chooses actions, picks servers, or sees others' data.
Common questions
What does the AI actually do?
It reads the customer's message, pulls out the details that matter, and words a reply. It does not choose actions, pick servers, or make authorization calls. Code does that.
How are actions kept safe?
Every action is a pre-approved operation with a fixed impact grade. Read-only diagnostics run automatically, reversible fixes are opt-in, higher-impact changes need a person, and destructive actions are never automated.
When does a human take over?
When the AI is not confident, when a fix would exceed its permission tier, when the topic is out of scope, or when the customer asks for a person. Your team picks up with the full transcript in front of them.
What are the limits Panelbot will not cross?
It will not touch accounts a customer does not own, run anything outside the approved catalog, or perform the top impact tier of changes on its own. Those always need a human.
Start here