AGENT INTELLIGENT SEARCH WITH AMAZON KENDRA & CHATGPT

In today's fast-paced world, contact center agents are expected to deliver prompt and accurate responses to customer inquiries, regardless of the complexity of the questions or the volume of the queries they receive. However, the challenge lies in finding relevant answers quickly when searching across multiple repositories containing unstructured content. This is especially true in the healthcare industry, where contact center agents deal with a very large volume of member and provider calls and they are responsible for providing vital information in a timely and efficient manner. According to industry reports, a center agent spends on average from 2-5 minutes to find an answer for a health benefit question due to the unstructured and complex nature of plan and benefit information. This time frame may be longer for more complex or technical inquiries or for agents who are less experienced or less familiar with the content being searched. These challenges result in longer response times, lower customer satisfaction, and higher operational costs for healthcare organizations.

Brillio’s “Agent Intelligent Search” solution is a cognitive search powered by GPT to help call center agents quickly and effortlessly find the direct answers to any questions. It utilizes natural language processing and machine learning algorithms to parse through unstructured data and retrieve relevant information from multiple content repositories. The service supports the following key features:

• Support of Natural Language / Semantic Search: Unlike keyword search, semantic search considers the semantic context of search queries and content to better understand meaning. Semantic search engines also consider the relationships between entities for returning search results. In contrast, purely keyword-based search systems only work because of a keyword-text match.
• Support for Investigative Search (Google Like Results ) Experience: Business can decide if they want results to be returned in from one or more documents with the ability to open the source document and get more context (just like searching Google) and to what level of confidence the results needs to be restricted to be showed on the response screen along with the deep linking to the document and specific section of the document.
• Support for Generative response (ChatGPT like Result): Business can decide to provide generative responses to the agents (and in future can be extended as self-search for members / providers.“Generative Model Transparency and Control”
• Data Lineage: Offers the full flexibility by supporting multiple different data formats, languages, and different types of CMS. Supports & extracts data from different file formats such as HTML files, MS (PPT) presentations, MS WORD documents, Plain text documents, PDFs, Comma Separated Values (CSV) files, Microsoft Excel (MS EXCEL) files, XML files, JSON files, Markdown Documentation (MD) files, Rich Text Format (RTF) files, Extensible Stylesheet Language Transformation (XSLT) files.
• Extensible connectors (Built in & extendable): Directly connecting your document repositories such as MS SharePoint, Salesforce knowledge base, internal or external websites, file share drive, etc. This gives you the flexibility to also not have to manage document updates and versions. You can configure your connectors to update the knowledge base if a new version of the document is published or overrides it.
• API Layer driven: The cognitive contextualized search response for the query is available as a service API for integration with any upstream / downstream systems. The Generative response using ChatGPT is context controlled via API and exposed as API service for any upstream / downstream system needs.
• Document / Content additions / updates / removals: The decoupled and scalable architecture allows the document and/or content inclusions, modifications, and deletion to be seamlessly performed without impacting the end-to-end solution performance as the models’ self-train themselves and adapt over time.
• Feedback Loop: Ability for users to provide feedback about answer relevancy through a click event (user clicking on the document link) or thumb up/down so the system can learn and get better over time.
• IAM / Access controlled: Ability to restrict the entry via IAM / Ad authentication.

Major Challenges:

• Agents' performance is bad if they are not empowered with the right technology, tools, or automated training systems.
• Searching and navigating across multiple data sources to extract relevant information requires a lot of time and results in customer dissatisfaction.
• Increasing call volumes are difficult to handle with the insufficient number of agents present at the contact centers.

Please contact jayadev.b@brillio.com  for additional information.