Hello everyone,

Admins can fine-tune how the Answer Bot evaluates queries and generates responses by configuring its accuracy threshold settings.

Using predefined confidence levels, Zia compares user questions with the available knowledge base (KB) articles. The selected threshold determines how closely a query must match an article before a response is generated. Higher thresholds focus on precision, while lower thresholds allow broader interpretation.

There are three accuracy levels available:

Exact – Maximum Precision
This setting is ideal when accuracy is critical. The confidence requirement is set very high, meaning the bot responds only when it identifies a near-identical match in the knowledge base.
Under this level, the bot primarily searches for articles with closely matching titles before generating an answer.

For example, if a customer asks, “How do I reset my password?” the bot will respond only if it finds an article with the same or highly similar title addressing password reset instructions.

Confident – Balanced Accuracy
This level offers a moderate confidence setting. The bot responds when it detects a strong contextual relationship between the question and available articles.

Instead of requiring an exact title match, it analyzes intent and related keywords.

For instance, if a customer says, “I forgot my password,” the bot can identify an article covering password recovery and generate an appropriate response.

Possible – Broad Coverage
This option works with a lower confidence threshold. It is designed to assist users even when queries are loosely phrased or indirectly expressed.

Here, the bot prioritizes relevance over strict matching.

For example, if a customer states, “I can’t see my agent under a department,” the bot may connect the query to articles related to teams, departments, or user assignments and provide a helpful answer.

Choosing the right threshold is especially important during the initial setup of the Answer Bot. If the bot’s responses appear too limited, switching to Confident or Possible can expand coverage and improve responsiveness.

By carefully adjusting these threshold levels, teams can strike the right balance between accuracy and accessibility — ensuring users receive consistent, meaningful, and reliable answers.