Testing, Quality, and Credit
On this page
You have the ability to test BLE AI before launching it, measure its quality while it’s live, and track the credits it uses. This article explains preview mode and the Tests, Quality, Insights, Performance, and Credits screens, along with when each one is useful.
- BLE AIOpen the screens in the Measurement and Broadcast folders
- Payment MethodsAdd a Registered Card to Purchase Credits
- Teaching BLE AIClose unanswered questions
1. Preview: Try out the chat
The "Test BLE AI" button in the header of the BLE AI › Settings, Knowledge Base, Flows, and Tools screens opens a "Preview" panel on the right. Here, you can type as if you were a customer and see the agent’s response. The preview works even if the agent is paused; preview conversations do not appear in the inbox and do not send notifications to the team. "Reset Preview" starts a new conversation. If there are unsaved changes in the flow editor, you can test the current version by selecting “Save and Reset Now.”
Answers in the preview also use credits. If your credit balance is zero, BLE AI won’t provide answers even in the preview; a message indicating this will appear at the top of the preview panel.
2. Tests
BLE AI › Measurement › Tests—these allow you to ask the same questions after every change and identify what went wrong before the customer does.
- Create a suite — Click "New Suite" and give it a name.
- Add a question — "Generate from Inbox" adds questions based on real customer inquiries; with "Paste CSV," you can enter "question, expected answer, expected result" on each line, or you can enter questions one by one. A suite can contain up to 50 questions.
- Run — Click "Run Tests." The test runs in the background; results appear on the screen as they come in.
- Read the results — Each answer is marked as Good, Bad, Error, or Not Rated; the summary line also indicates how many questions you answered and how many you skipped.
Each suite contains two built-in questions labeled "Security" that cannot be deleted. These test whether the agent fails to follow instructions such as "forget previous instructions" and does not strictly adhere to internal rules. Test questions also generate real answers and consume credits; if credits run out, the run stops.
3. Quality
BLE AI › Measurement › Measures actual conversations in the quality stream. It has three parts:
- Scoring Cards — What does “Good Answer” mean? It is defined by criteria. Each criterion has a definition (“What does it look for?”) and a weight; an answer that fails to meet a criterion marked as “Critical” is considered a failure regardless of its weight.
- Monitors — Determines which closed calls will be sampled (default 10%) and who will review them. When "Automatic Scoring with Referee" is enabled, scores are automatically assigned using the scorecard you select; if you choose a reviewer, the sampled calls are distributed among them, starting with the one with the lightest workload.
- Your review queue — Conversations assigned to you. Use "Open Conversation" to read them, then mark them as "Passed" or "Pending"; your score will replace the referee’s score.
4. Insights and Performance
BLE AI › Metrics › Insights are populated with tasks that run overnight: “What’s Next” suggestions (content gaps, over-repetitive topics), “Events” showing unusual spikes in volume, “Topics” where customer questions are clustered, and “Weekly Trends.” For a newly created account, it takes a few days for these sections to populate.
BLE AI › Metrics › Displays performance over the last 7, 30, or 90 days: "Resolved by AI," "Escalated to an agent," "Customer satisfaction" (likes/dislikes customers give to answers), and "Unanswered questions." Below, you’ll find a list of “Questions the AI Could Not Answer,” a table showing “Sources Used by the AI,” and a breakdown of “Where Credits Were Spent.” When you fill in a blank by adding content, mark the row as “Resolved.”
5. Credits
A credit is a response provided by BLE AI to a customer. If a response cannot be generated (e.g., when no reliable source is found or an error occurs), the credit is refunded. Previews and test responses also consume credits; when the agent assistant’s Draft and Ask modes consume credits, this is noted below the result. When the balance reaches zero, BLE AI stops providing responses and transfers the conversations to your agents.
On the BLE AI › Publish › Credits screen:
- AI Credit Balance — Remaining credit and approximately how many days it will last at this usage rate.
- Daily Usage (Last 30 Days) — Daily Usage Chart.
- Add Credit — Available packages. To click "Buy," you must have a card registered in the Payment Methods section of your account dashboard. Card verification opens in a separate window; make sure your browser allows pop-ups. If the screen displays a message stating that online payment is not configured, contact the support team to top up via bank transfer.
- Auto-Top-Up — Enable the "Automatically purchase credit when balance is low" option, select the threshold balance and the package to purchase, then save. If payments fail repeatedly, auto-top-up will stop; check your card and re-save the settings.
- Credit Receipts — A receipt for each purchase can be downloaded as a PDF.
A warning will appear on the Credits screen when your balance is low. Keep auto-recharge turned on to avoid interruptions during calls.