IBM Watson Assistant has introduced conversational search, utilizing generative AI to produce conversational answers that are based on specific information from an enterprise’s content. This capability allows the virtual assistant to respond accurately to queries from customers and employees without the need for human authors to manually write or update the answers. By leveraging this technology, Watson Assistant enables businesses to accelerate their time to value and reduce the total cost of ownership of virtual assistants.

Watson Assistant integrates with Watson, IBM’s AI and data platform, which facilitates the training, deployment, and management of foundation models. This integration allows business users to automate precise conversational question-answering by using customized Watson large language models.

Since 2020, IBM Watson Assistant has been utilizing foundation models to enhance text processing and understanding, including customer conversations. Now, with its integration with Watson, Assistant implements Retrieval-Augmented Generation (RAG), which is a generative AI framework that generates contextual answers from enterprise-specific information when responding to natural language questions.

Retrieval-Augmented Generation (RAG)

RAG is an AI framework that combines search with generative AI capabilities. It retrieves enterprise-specific information from a search tool or vector database and uses that information to generate a conversational answer grounded in the retrieved content.

Retrieval Phase

First, Watson Assistant retrieves relevant information from an organization’s content, which can be stored in a knowledge base or content management system. Assistant can connect to this content using various patterns, including no-code, low-code, or custom configuration. It also supports an out-of-the-box, no-code integration with Watson Discovery for search. Clients can also take advantage of Watson Assistant’s starter kits to connect to popular search tools like Coveo, Google Custom Search, Magnolia, and Zendesk Support.

Answer Generation Phase

Once relevant information is retrieved, Watson Assistant passes it to a Watson large language model (LLM) for generating a conversational answer grounded in the content. By providing accurate and up-to-date content to the LLM, Watson Assistant ensures that the generated answers are specific to the enterprise’s closed domain rather than being based on open-domain internet-scale data. This approach reduces the risk of generating incorrect or misleading information and allows for easy traceability of answers back to their source content.

Watson Assistant has partnered with IBM Research and Watson to develop customized Watson LLMs that specialize in generating answers based on enterprise-specific content. Clients can connect Watson Assistant to these customized LLMs or third-party LLMs using the custom extensions framework, enabling retrieval-augmented generation and other generative use cases.

Benefits of Conversational Search with Watson Assistant

Conversational search, powered by the retrieval-augmented generation framework, offers several advantages for building, deploying, and maintaining virtual assistants:

• Easier building and deployment of virtual assistants, as Watson Assistant can accurately answer a wide range of questions without the need for manual authoring by non-technical business users.
• Effortless maintenance of virtual assistants, as Watson Assistant automatically retrieves updated information from the connected knowledge base to inform its answers.
• Accelerated time to value and reduced effort required for building and deploying conversational experiences with Watson Assistant.

Why Choose Conversational Search with Watson Assistant?

In addition to conversational search, Watson Assistant offers prebuilt integrations, a low-code integrations framework, and a no-code authoring experience. This comprehensive solution allows developers and business users to automate question-answering and focus on building higher-value transactional flows and integrated digital experiences with their virtual assistants.

Furthermore, Watson Assistant continues to collaborate with IBM Research and Watson to advance its capabilities in understanding customers and solving various conversational use cases. With large language models, Watson Assistant has achieved significant advancements in accuracy while reducing effort.

To learn more about how you can utilize Watson Assistant’s generative AI capabilities to engage your prospects, customers, and employees with conversational experiences, schedule a consultation or watch the on-demand webinar ‘AI for customer service’.

---

IBM Watson Assistant can now provide conversational answers based on an organization’s proprietary or public-facing content without requiring human authors. This capability allows businesses to rely on their virtual assistant to retrieve up-to-date information and simplify content maintenance across different communication channels and repositories.

With the recent addition of conversational search, IBM Watson Assistant leverages generative AI to generate accurate conversational answers based on enterprise-specific content, freeing non-technical business users from manual authoring and reducing the total cost of virtual assistant ownership.

Frequently Asked Questions (FAQ)

What is generative AI?

Generative AI refers to the use of artificial intelligence techniques to create new and original content. In the context of conversational search, generative AI is employed to generate conversational answers based on enterprise-specific information.

How does Watson Assistant retrieve information from an organization’s content?

Watson Assistant connects to an organization’s content through a search tool, such as Watson Discovery or custom search integrations. It retrieves accurate and up-to-date information from a knowledge base or content management system to inform its generated answers.

Can Watson Assistant generate answers specific to an organization’s closed domain?

Yes, Watson Assistant ensures that the generated answers are based on a closed domain of enterprise-specific content rather than open-domain internet-scale data. This reduces the risk of generating incorrect or misleading information.

What are the benefits of using conversational search with Watson Assistant?

Conversational search with Watson Assistant makes it easier to build and deploy virtual assistants by eliminating the need for manual answer writing by non-technical business users. It also simplifies the maintenance process, as Watson Assistant automatically retrieves updated information from the connected knowledge base. This approach accelerates time to value and reduces the effort required for building and deploying conversational experiences with Watson Assistant.

What other capabilities does Watson Assistant offer?

In addition to conversational search, Watson Assistant provides prebuilt integrations, a low-code integrations framework, and a no-code authoring experience. These features enable developers and business users to automate question-answering and focus on building higher-value transactional flows and integrated digital experiences with their virtual assistants.