“We appreciate the hard work of the Sobot team during our collaboration. The upgraded AI Agent has improved a lot compared to the previous version. Currently, the LLMs can extract knowledge quite accurately, reducing the workload for human agents. In particular, it has greatly facilitated the management of knowledge base, saving a significant amount of manpower, and making it much more convenient to update the knowledge in the future.”
–Customer Success Manager, Leading Ecommerce Brand
Currently, Generative AI is rapidly transforming customer contact center, with various new features emerging. It might seem that having more features equates to being more intelligent. However, Sobot believes that truly helping customers “use” AI Agent is far more important than “blindly piling up functions”. Since Sobot launched the LLM-powered AI Agent, the users’ experience has been deeply embedded in this product’s DNA.
Therefore, to further enhance user experience, Sobot has upgraded our AI Agent again– over 60 adjustments have been made, covering 13 of Sobot’s business modules. In particular, Among these, the ease of use of the AI Agent has been significantly optimized.
Significantly Enhanced Knowledge Center
Efficient Knowledge Upload and Precise Knowledge Optimization
1. Support Uploading New Content Formats like URLs and Sheets
Previously, despite its strong autonomous learning capabilities, AI Agent could only address materials in certain formats. Materials like webpages and Excel sheets couldn’t be directly upload to the knowledge center. So companies had to transform those materials to acceptable formats first and then upload them for AI Agent to learn, which was inconvenient.
After the upgrade, AI Agent can now directly process files in formats like webpage URLs and Excel sheets. This improvement is particularly valuable for businesses whose products are diverse in categories and information, and large in content volumes, greatly reducing the time required to organize learning materials.
2. Support Uploading 100 Files At One Time
Previously, due to limitations of the LLMs, operators had to upload knowledge materials one by one. For companies with multiple product lines, numerous product versions, or complex documents, uploading hundreds of documents manually at a time was far from easy.
After this upgrade, Sobot system supports “bulk uploads of up to 100 documents at a time”. This means companies can even skip extensive pre-filtering and directly “throw” all documents to the AI Agent for rapid knowledge preparation. It significantly improves efficiency for businesses handling large volumes of documents or frequent updates.
3. Quickly Recognize Abnormal Performance for Precise Optimization
Sobot has added a “Customer Questions” dashboard in the operation backend. As the AI Agent works, all customers’ questions will be automatically clustered and presented together with key indicators such as chat volume, resolution rate, and transfer-to-agent rate. Operators can supplement more knowledge in specific directions based on the performance of different question types, gradually improving the independent resolution rate.
Drilling down to each chat, operators can filter out chats that had been transferred to humans or received poor ratings in the backend. After review, they can directly upload “supplementary knowledge” for questions the AI Agent answered incorrectly. In subsequent service, the AI Agent will automatically correct its answers to similar questions, improving answering accuracy.
Comprehensive Data Statistics of AI Agent Performance
Not just output, but also “feedback”– a more complete statistical system
1. Customers can give feedback on every answer from the AI Agent
After the upgrade, “Like/Dislike” buttons have been added to every answer given by the AI Agent. Customers can “like” or “dislike” the accuracy of the answer, allowing operators to quickly filter out questions with high “dislike” counts for precise optimization.
2. Every independent AI Agent chat can trigger a “feedback invitation”
Businesses can set up in Sobot’s system and automatically send a “satisfaction feedback invitation” after every chat of the AI Agent. Customers can provide an overall rating for the AI chat, allowing operators to more intuitively understand the AI Agent’s comprehensive performance and focus on chats with low satisfaction.
3. Provide a comprehensive statistical system with “Sankey Diagram” and various indicators
In the statistical reports, besides some basic data, this time’s update also includes more analytical indicators such as satisfaction ratings, chat details, answering like/dislike counts, answering details. These are presented in a clearer Sankey Diagram to show the AI Agent’s performance. This allows operators to quickly identify issues through the reports, enhancing optimization efficiency.
Conclusion
As a customer contact centers provider that embraces “Generative AI”, Sobot is rapidly leading the industry into the “AI Agent” era. Besides the upgrading mentioned earlier, Sobot has also optimized various aspects of the AI Agent, including task workflow, task information collection, intelligent replies, channel adaptation, test operations, and more. While meticulously refining to enhance AI Agent capabilities, Sobot focuses even more on making it truly “easy to use” for business customers. If you wish to learn more detailed information about the upgraded AI Agent, please feel free to contact Sobot anytime.
