By Carolyn Duncan, Sr. AI/ML Specialist – AWS
By David Shute, Sr. AI/ML Specialist – AWS
By Pat Higbie, Co-Founder and CEO – XAPP AI
Imagine if prospects, customers, and employees could get the information they need by asking a question through voice or text. Everyone could get what they asked for quickly and efficiently—anytime, anywhere, and from any device.
This natural “ask and get” style of self-service experience would reduce the time spent digging through a website, waiting on hold, or answering 20 questions.
In this post, we’ll highlight a conversational self-service (CSS) solution created by XAPP AI and powered by AWS artificial intelligence (AI) services.
XAPP AI’s CSS solution gives you the ability to provide this level of self-service experience by allowing users to ask natural language questions and get accurate answers from existing corporate knowledge bases.
XAPP AI is an AWS Machine Learning Competency Partner that provides conversational AI solutions empowering the next generation of customer service and deliver rapid return on investment (ROI).
You can try out a working demo of a CSS chatbot to answer consumer financial questions. The demo is powered by Amazon Lex and Amazon Kendra, and was developed in under one hour using the concepts detailed in this post.
The Challenge of Implementing CSS
The customer experience and operational efficiency benefits of conversational self-service are compelling. They’re best achieved using the latest machine learning (ML) technology to avoid the pain points that many enterprises contend with during implementation.
Typical pain points include lack of ML and cloud computing expertise and scarce IT resources. Additionally, organizations that attempt to capture content specifically for CSS manually find it can be a daunting task. It’s challenging not only to capture the content but also to run it through the required legal and compliance approvals and keep it up to date.
Similarly, many enterprises attempt to move into transactional use cases. These include functionality such as ordering products from a large catalog without understanding the complexity. Slow CSS implementations can lead to a loss of organizational support that can jeopardize the business outcomes that CSS is meant to achieve.
All of these pain points can slow the implementation of CSS projects.
Making it Easier
You can avoid these pain points and quickly achieve your goals with Conversational Self Service for Contact Center Intelligence (CSS4CCI) from XAPP AI. It delivers powerful self-service customer experiences with speed and simplicity by combining the power of Amazon Lex and Amazon Kendra.
Amazon Lex is a service for building conversational interfaces into any application using voice and text. It provides advanced deep learning functionalities of automatic speech recognition (ASR) for converting speech to text. It uses natural language understanding (NLU) to recognize the intent of the text in applications with highly engaging user experiences and lifelike conversational interactions.
Amazon Kendra is an intelligent search service powered by ML. It reimagines enterprise search for websites and applications so employees and customers can find the content they’re looking for more easily. The content may even be scattered across multiple locations and content repositories within an organization.
CSS4CCI starts the implementation process by automatically deploying and securing the necessary Amazon Web Services (AWS) infrastructure into an enterprise customer’s AWS account (see the diagram in Figure 1).
This includes automatically creating the Amazon Lex model needed to understand natural language questions from customers. It also creates the Amazon Kendra index for answering those questions from document repositories of unstructured content.
Figure 1 – CSS4CCI architecture showing automatically deployed AWS infrastructure.
The next two steps highlight unique features of CSS4CCI and are essential to achieving CSS success quickly and with measurable results:
- Automate knowledge capture: CSS4CCI’s intelligent web crawler can crawl any publicly accessible website. It scans the HTML for embedded metadata and automatically creates FAQs in Amazon Kendra. In addition, it creates an accompanying Amazon Simple Storage Service (Amazon S3) metadata file containing source URLs for “Read More” functionality. This is important for curious customers that want to do more research on a topic after reading a high-level answer.
The intelligent web crawler can be set up for any URL and provides for ongoing content management. It can also set a cadence for re-crawling any or all of the URLs included in the Amazon Kendra index. Automating knowledge capture and keeping it current with existing content is critical for conversational self-service.
- Optimize knowledge curation: XAPP AI’s automated human-in-the-loop workflow portal supports non-technical subject matter experts (SMEs). For example, it enables customer service representatives to more easily identify, prioritize, and improve on inadequate system responses.
SMEs then optimize the virtual assistant by adding information into the portal. This helps the virtual assistant either understand the user request or provide appropriate system responses. This trains the ML models to respond accurately to customer requests and further enhances the intelligence of the self-service virtual assistant.
Increasing the Value of Corporate Knowledge Bases
Content is essential in conversational self-service and is often plentiful in enterprises. CSS4CCI uses Amazon Kendra to ingest unstructured data to process and comprehend, and then deliver intelligent responses.
CSS4CCI can easily derive value from many content sources, including the following:
- Publicly available web content: Large websites can be difficult for customers to navigate. However, even sites with thousands of pages can be crawled and indexed for CSS in a matter of minutes.
- FAQs: These represent the types of questions customers ask regularly, but it’s often time-consuming to find the ones that apply to their needs. CSS4CCI uses existing FAQs to give customers the ability to ask questions and get instant responses, saving them time and improving their experience.
Amazon Lex plays an important role by first determining if the query can be handled in the interaction model before calling Amazon Kendra. This is essential to long-term conversational AI success because it enables virtual assistants to understand both informational and transactional use cases.
- Enterprise content repositories: Amazon Kendra connectors make it easier to capture data from multiple content repositories. You can schedule them to sync your index with these data sources automatically. Pre-built connectors are available for Amazon Relational Database Service (Amazon RDS), Amazon S3, Atlassian Confluence, Atlassian Confluence Cloud, Microsoft OneDrive, Microsoft SharePoint Online, Salesforce, ServiceNow, and Google Drive.
- Custom content repositories: Custom connectors allow enterprises to use the Amazon Kendra API to create their own custom connectors. You can push content directly into Amazon Kendra from any content repository. Additionally, service providers have developed more than 50 connectors for a wide variety of repositories.
Amazon Kendra turns unstructured enterprise data into intelligent answers even when the questions aren’t explicitly defined in the content. This advances conversational self-service in a profound way.
To experience this firsthand, use the free trial of CSS4CCI and crawl the Coffee FAQ, hosted by the University of Waterloo, as one of the URL sources. You can use the CSS4CCI chat channel to ask the question, “What is the best coffee?” This question isn’t one of the FAQs, but the answer exists in the web content, so Amazon Kendra finds it. The following screenshot shows the response.
Figure 2 – CSS4CCI configurable chat widget showing user request for “What is the best coffee?”
Deploy to Web, Voice, and Device Channels
CSS comes to life when virtual assistants are deployed to channels where customers are looking for answers. Many customers start with the web, and CSS4CCI supports this channel with a web chat widget you can deploy to any website with one line of code.
The resulting web chatbot automatically invokes a powerful virtual assistant that produces consistently high-quality customer experiences. It’s fully configurable to align aesthetically with any brand, as illustrated in the Coffee FAQ. In addition, virtual assistants created with CSS4CCI integrate with Amazon Lex chatbots directly, or through Amazon Connect’s built-in chatbot.
CSS4CCI also makes it easier to deploy self-service conversational assistants to voice channels via Amazon Connect and Genesys Cloud, in addition to Alexa and Google Assistant.
AI-powered virtual assistants allow enterprises to reinvent customer service for the new realities of business. They can make contact centers a natural first use case for conversational self-service. CSS also lends itself to a wide range of internal-facing use cases, including help desks, HR support, sales support, and enterprise knowledge management.
Website search for customer support is another great use case that can improve the customer experience dramatically.
The Consumer Financial Protection Bureau (CFPB) website has a Consumer Resources section that’s an excellent example of the power of CSS4CCI’s intelligent web crawler combined with Amazon Kendra and Amazon Lex for website search for customer support. CSS4CCI crawled the CFPB Consumer Resources and quickly turned 1,460 webpages into a rich, searchable knowledge base that outperforms the search built into the site.
The following screenshot shows the results for the question: “What happens if I miss a payment on my student loan?” As with many websites, the user is left to find the answer.
Figure 3 – CFPB web search result for “What happens if I miss a payment on my student loan?”
Next, the same question was submitted via the CSS4CCI chat channel widget and it provided a more informative answer. The “Read More” button brings the user directly to a page with more detail and potential solutions, providing a better user experience compared with the website search.
Figure 4 – CSS4CCI chat widget response for “What happens if I miss a payment on my student loan?”
Succeed More Quickly and Scale More Easily
For organizations using machine learning, it’s important to achieve business outcomes quickly in terms of improving the customer experience in addition to ROI. CSS4CCI is a conversational self-service solution that makes it easier for enterprises to create an intelligent virtual assistant in under 60 minutes without AI or cloud experts.
Conversational AI transformation requires a long-term commitment to continuous optimization based on changing customer expectations and constantly improving AI technology.
This complexity increases dramatically for use cases with millions of combinations and permutations of user requests. For example, “I want to order a single shot, medium, skim, no-foam, extra-hot latte” contains over 80,000 drink combinations and more than 2 million ways that customers will ask for them.
CSS4CCI is powered by Optimal Conversation Studio (OC Studio). It gives enterprises the ability to create and maintain advanced Amazon Lex interaction models for understanding users in these complex use cases. It also supports dialog management, content management, and API integrations.
OC Studio enables enterprises to manage conversational self-service as a product. It can be continuously optimized and improved based on changing user behaviors and advances in AWS AI services.
We invite you to check out the free trial of CSS4CCI in AWS Marketplace to see how your enterprise can create a virtual assistant in less than an hour and deploy a CSS chatbot to your website soon thereafter.
You may also want to try the CSS4CCI Workshop to learn a step-by-step process for completing a conversational self-service proof of value. For more information about the AWS AI services that power CSS4CCI, see Amazon Lex, Amazon Kendra, and the AWS Contact Center Intelligence (CCI) program.
XAPP AI – AWS Partner Spotlight
XAPP is an AWS Machine Learning Competency Partner whose Conversational AI solutions empower the next generation of customer service and deliver rapid ROI.
*Already worked with XAPP AI? Rate the Partner
*To review an AWS Partner, you must be a customer that has worked with them directly on a project.