Evaluate Your Agents

In-chat Evaluation

Users can rate each response from the Domain Expert Agents (DXAs) as "Helpful" or "Not Helpful." Admins receive notifications to review and improve unhelpful answers, ensuring the continuous refinement of agent responses.

Rate as an helpful answer

Rate as an unhelpful answer

Contributors will be notified when an agent’s answer is rated as unhelpful

Management Evaluation

Experts can upload datasets, run evaluations, and analyze results to systematically enhance agent performance. This structured evaluation process allows for targeted improvements and knowledge updates.

Step 1: Upload dataset

Choose a dataset to evaluate the agent. This dataset should be representative of the types of queries the agent will handle.

Step 2: Review accuracy result

View the results of the evaluation to identify incorrect answers and understand why these answers failed. This analysis helps in pinpointing specific areas where the agent needs improvement.

Evaluation results

Analyze unsatisfactory answers

Step 3: Re-evaluate the agent

Contribute knowledge to the agent in bulk to massively improve the quality of the answers. After adding new knowledge or updating existing information, re-evaluate the agent to see the improvements in its performance.

Adding knowledge

Re-evaluate agent

Final Results