Skip to main content
Most bot platforms show you how much the bot does. Ciarem’s Control center (under AI configuration) answers a different question: is it doing its job well? One screen shows whether your setup is ready, how much confidence it has earned, and what your live conversations look like. The Evaluation center is one click away when you want to judge actual replies.

​
Readiness: before you launch

Three cards at the top of the Control center track the pieces your agent needs, each with its own health bar and a link to the screen that fixes it:
  • Agent team: you have active agents with something to work with (“Setup complete” when everything passes).
  • Knowledge base: your business information is indexed and connected to your agents. If agents keep coming up short on answers, this card warns “Insufficient sources”.
  • AI training: you have rated enough test conversations for Ciarem to judge how your agents perform. The card links straight to the playground, where every rated reply counts.
Next to the AI status toggle, the Confidence level meter (low, medium, high) summarizes how complete the configuration behind your agents is. If confidence is low, Ciarem asks you to confirm before publishing. You can still publish. The step-by-step version of all this is the launch checklist. The Control center: the three readiness cards, the confidence level, and the live conversation numbers

​
Live numbers: once you are live

Below the cards, the Control center shows your real activity over the range you pick (the last 12 months by default):
  • Conversations handled: how many conversations your workspace took on
  • Awaiting reply: how many customers are waiting right now. Keep this at zero.
  • Conversations transferred: how many were handed to a human
  • Response time (median): how fast customers get a first answer
The chart underneath breaks conversations down by channel and month. For a deeper cut (how many AI replies were answered outright, transferred, or out of scope), follow Go to general metrics and read the AI agent outcomes card on the Metrics screen.

​
Beyond the numbers: review real replies

The Control center shows the numbers. The Evaluation center shows which replies to fix. It collects the replies your agent actually sent, groups them into themes, and turns your approvals and corrections into lasting behavior and transfer rules. When there is a fresh batch of replies worth reviewing, the Control center itself invites you to run an evaluation. Look at your live numbers weekly and run an evaluation now and then.

​
The first weeks

With low conversation volume, numbers move a lot. Do not chase a percentage in week one. Focus on keeping Awaiting reply at zero, skimming transferred conversations in the Inbox to see why they left the AI, and running the playground with real customer questions.