Augmenting CX and KM with ChatGPT technology

With the virality of generative and conversational AI in recent months, organizations are looking toward this highly popular technology to optimize their business operations, increase productivity, and generate positive business outcomes. In the realm of CX, AX (agent experience), and EX (employee experience), generative AI poses a wealth of opportunities for competitive business advantage.

eGain thought leaders and practitioners joined KMWorld’s webinar, “ChatGPT and Generative AI for CX: Hype, Reality, and What to Know and Do,” to unveil the ways in which generative AI can revolutionize experience areas for businesses.

Ashu Roy, CEO of eGain Corporation, boiled down what ChatGPT really is: an evolution of ongoing trends, particularly those related to ML, that delivers “tipping point” class conversational experience and ease of use.

The excitement surrounding this evolved trend is due to the wide variety of use cases applicable, ranging from content generation to content summary, even reaching into specialized areas like solving math problems, writing code, and writing music. 

Ian Jones, head of strategic solutions at eGain Corporation, dove into how resources like ChatGPT will affect CX. Firstly, on the customer end, many of them are aware of ChatGPT and its capabilities—particularly leveraging the tool for search and correspondence (such as emails, letters, etc.).

With the level of convincing answers and generation of information that ChatGPT provides, everyday consumers of the tool may begin to look at it as a source of truth, neglecting to fact-check its generations. This presents a breeding ground for misinformation or spam—a challenge of ChatGPT that would also impact businesses attempting to integrate with it.

For CX content teams, the effects of ChatGPT and other generative AI look slightly different; those generating content can find solace in leveraging the tool as a research resource, a content drafter, a style editor, a translation tool, or a tool for training and education. Once again, Jones emphasized, the need to validate the information generated by ChatGPT is critical in implementing it successfully.

Jones then addressed the concerns regarding customer-facing systems and the impact of ChatGPT workflows. For those who already work with a chatbot thoroughly trained on their content or a self-service knowledge base, the impact should be relatively minimal.

An area where ChatGPT may affect, however, is the content available for its responses. Curated content, documented content, and generated content must be consistent with certain attributes—including accurate, up-to-date, explainable, compliant, and brand-aligned—while still being consumable and accessible within the system.

Content generated by ChatGPT has certain implications depending on the context in which it has been requested; for an area like medical or legal advice where information is high risk and public content, Jones emphasized avoiding GPT, and instead directing the question to an individual with appropriate expertise. Another high-risk area—private, core content refresh for re-styling or reformatting corporate, customer-facing knowledge bases—can benefit from GPT generation but with human approval due to its high risk.

Additionally, low risk areas—user education and value-add additional information—can both benefit from GPT. Value-added additional information involving public content, such as recipes, travel advice, and general troubleshooting, can offer GPT content with a disclaimer that refers to its generation by a GPT tool. User education, which deals with private content, benefits from GPT with an auto-validate feature that ensures accuracy.

By looking at the bigger picture of how GPT fits within an enterprise stack and CX workflows, Jones explained that the tool fits within feedback and suggestions (similarly in how one might use a search or browsing tool), as well as in customer interaction content. It particularly helps the authors, Jones argued, when reformatting content or to help flesh out queries.

On the knowledge management front, ChatGPT can help envision the purpose of a knowledge system before the content is fed into it. Understanding what the KM system is trying to achieve and how it represents a brand persona can be discovered with a GPT tool; additionally identifying personas seeking information, deriving style guides, generating curated content, and designing content workflows are areas where generative AI can excel.

“ChatGPT and GPT products can enhance your knowledge management on the creation and formatting side, but also critically,” said Jones. “Strong knowledge management can really enhance what you get out of ChatGPT.”

Ultimately, the balance of GPT and KM requires knowledge on how the two services can augment each other. KM tools and processes can leverage GPT to:

• Surface content gaps
• Surface underperforming content, such as content that generates poor ratings, little reader retention, or repeated searches
• Drive content management review/approval workflows

On the other hand, for KM systems, generative tools can:

• Identify likely questions
• Provide feedback on existing content
• Generate new content from articles and PDFs

Jones then shifted the lens onto how GPT can augment automated responses, using eGain Instant Answers, eGain’s search experience for enterprise knowledge, as an example. After semantically searching the knowledge base, eGain Instant Answers applies generative AI to generate a valid answer, then checks it for tone and accuracy. Through a series of ML operations, GPT generates responses for eGain Instant Answers while the platform auto validates those same responses.

The discussion concluded with a few lessons learned and things to bear in mind when incorporating GPT technology within enterprise knowledge bases. Jones explained that prompts are key; clear, short, concise answers that lack jargon and acronyms will provide the most optimal responses. Additionally, while correctness is critical, it’s certainly a difficult requirement to meet effectively.

Ultimately, generative technologies are likely to rejuvenate knowledge management because trusted content is key, according to Jones. Furthermore, enterprise controls are essential to operationalize new capabilities, and the GPT landscape is changing rapidly; adopting a composable platform to experiment is crucial in staying agile.

For an in-depth discussion of how GPT technologies can augment KM, you can view an archived version of the webinar here.