Generative AI, in the form of OpenAI’s ChatGPT and other tools, is very much in the public eye at the moment.

As a consumer searching for content or ideas, generative AI offers the promise of a clear, concise response that can be clarified or refined, albeit with a chance that it may be incorrect or out of date.

Business needs, especially for customer service, are different from consumer needs. The customer service department must give consistent responses to a high volume of questions, which must be accurate, up-to-date, compliant with regulations, and understandable to avoid callbacks or escalations. A business will also need to know trends in the questions asked, what responses are working best, and if there are new types of questions.

To assess where generative AI can help with CX, it helps to look at the corporate content landscape, which consists of the following types:

Curated Content: This includes the knowledge base and source documents. Curated content is supposed to be accurate, up-to-date, approved, under change control, etc. This should be the source of truth for the organization.

Documented Content: This can include web sites, intranet, communities etc. The content might have been accurate and even approved when published (apart from opinion pieces or external content) but may not have been looked at since. For example, web sites might contain details of products that are no longer supported or press releases that are out of date. Documented content is often required to fully respond to customer enquiries which means advisors may need to search or browse different sources to augment curated content.

Generated Content: This is information from the web or other sources that forms part of the customer interaction. This is typically helpful ‘human-generated’ content that advisors have picked up and believe to be helpful. Generative AI can augment this type of content generation.

All three types of content are often needed in Customer Service but where they can be used and how they need to be treated, when using generative AI, are different. Below are ten use-cases that not only illustrate how it is done but also explain how to derive tangible benefits from generative AI in the customer contact center.

1. Creating effective prompts from knowledge artifacts

Generated content depends heavily on the prompts used to produce it. Where the knowledge management process has involved documenting artifacts such as the vision for the solution, the desired experience when using the system, the brand values, the tone and the writing style, and the tone of voice for target personas, generative AI can be used to build more effective prompts so that the generated material has the right content and tone of voice for the intended target audience.

2. Identifying likely questions

When a new knowledge base is developed or where there is a need to prepare content for new products or services prior to release, guessing what questions might be asked takes a lot of time and is often inaccurate. This results in creating and approving answers for questions that are never asked. With a suitable prompt, the sorts of questions that the target customer segment is likely to ask about the product or service can be generated. These can be used (possibly together with generated answers) for the first version of a knowledge base which can then be expanded with analytic insights as new questions are detected.

3. Generating draft content for knowledge base articles

Knowledge authors often struggle with the process of turning sprawling compliance-type documents into something that can be used by advisors. With suitable integration, workflows and controls, content can be generated from curated sources and the response checked before the draft is circulated to appropriate experts for approval as part of the knowledge workflows.

As well as creating draft content from curated sources, additional draft content can be generated from documented content and community content. This typically requires some content pre-processing so that it can be passed into a generative tool to constrain the responses and may still need to be checked by an SME rather than getting passed directly to a customer.

4. Augmenting search results and VA responses with generated content

Where a search of available content, either through a VA or web self-service site, does not give a high confidence result, generated content can be used as an additional step before escalation to human-assisted service. Where the question relates to curated content there would need to be a way to pass the necessary content along with the prompt and validate the resulting generated content. When external content is used as a source, the generated content may be used for handling the query with a disclaimer noting that the response has been generated, while pointing to external reference sources.

5. Re-purposing content into different styles or forms

Re-styling an entire knowledge base is a large task, which is often neglected. Generative AI can help automate it when used as part of content workflows. For example, generative AI could perform the bulk of the work in creating customer facing content from advisor-facing content to supplement self-service or a VA.

6. Summarizing feedback for content refinement

User feedback can be summarized from multiple sources into actionable tasks, using generative AI, and content can be refined, based on that feedback. Otherwise, authors would need to read every piece of feedback which might mean poring over hundreds of comments every day. Once the feedback has been summarized it can form part of a prompt that instructs generative AI to revise an article to be in line with the feedback.

7. Creating automatic chat responses from approved content

If the customer query has not been successfully handled by a VA dialogue or auto-suggest, then the query is possibly a hybrid of several questions and so a generated solution would probably be a better fit than a traditional knowledge base article. If the suggestion can be used directly this will be a productivity gain where advisors repeatedly type out similar responses rather than reuse the generated answers.

8. Creating useful additional content to support or augment curated content

The memorable parts of customer service interactions are often the small bits of ‘value add’ that advisors drop into the call after the main issue is resolved. Generated content can be helpful in this respect. Generative AI can augment advisors with the most relevant additional information and uplevel the customer experience to that provided by the most informed advisors.

9. Restructuring content drafted by advisors into specific formats for summary or escalation

After-call work can be a significant part of the agent workload – it is the time when they summarize the call and create notes to help the next person in the chain. Generative AI can take a chat transcript or draft notes and create a concise summary to facilitate escalation or transfer to another department. Having a standard approach to these handoffs can be of benefit to the end-to-end resolution process.

10. Creating follow-up correspondence for advisors / supervisors

Where an advisor needs to compose a follow up note (letter or email) including customer data, generative AI can assist by creating the right tone and format and including real-time data passed from the customer record.

The ability to quickly and easily create complementary customer service content and improve clarity and readability of that content helps organizations improve their customer and advisor experiences, while reducing costs. Whilst generative AI is available to all, it will be organizations that have robust Knowledge Management processes and practices and seamlessly integrate generative AI into those processes that will be able to gain the greatest advantage.

Contact us