Druckansicht | © 2002 – 2024 tcworld GmbH | Seite drucken

Is your content ready for GenAI-based agents?

Unlike human users, GenAI requires elaborate, text-rich information from which to generate precise responses. Here is how to prepare your content for a new era.

Text by Selvaraaju Murugesan

Inhaltsübersicht

Image: © Kirillm/stock.adobe.com

The emergence of Generative Artificial Intelligence (GenAI) has had an impact on self-service. Customers are no longer interested in typing search keywords and browsing a few articles to find what they are looking for. Instead, they utilize ChatGPT-like assistive search providing them with accurate responses to their questions. The underlying content needs to be trustworthy to ensure the responses meet customer expectations and prevent GenAI from hallucinating.

There are a lot of companies around the world that are working on enabling GenAI assistive search to latch onto their knowledge bases. However, the underlying content in these knowledge bases is still written for human consumption and focuses on meeting human requirements such as:

  • Content is usually concise, catering to the limitations of the human attention span.
  • Content is often complemented by rich media artifacts such as animated GIFs, images, videos, and so on to ensure that content is easy to comprehend.
  • Hyperlinks to various related articles are added for additional references.

However, writing content for GenAI-based agents has to cater to different requirements, such as:

  • GenAI-based agents are text-hungry; thus, underlying content must be as explanatory as it can be.
  • Given the nature of the chat-like interface, the underlying content must be written in a conversational style in a more generic persona.
  • A good business glossary is needed to ensure that GenAI offers accurate responses rather than hallucinations.

If your customers are utilizing assistive search such as ChatGPT and your existing content is not tailored to accommodate the requirements of GenAI-based agents, it is high time to conduct a content audit. The underlying content must be GenAI-friendly to ensure it serves your customers with trustworthy responses. This article provides a set of assessment criteria to make your content ready for GenAI agents to build upon. The content audit will focus on content quality issues such as content freshness, content structure, and business glossary, among others.

Content freshness

For a GenAI-based agent to provide accurate information to your customers, the underlying content should always be up to date. Thus, it is imperative to ensure that all the content in your knowledge base has the “last modified”/“last updated” date added to its metadata. This helps GenAI-based agents to provide timely and accurate responses to customer questions. A good measure to check content freshness is categorizing content into a few buckets based on time, e.g., one-month-old content, three-month-old content, and more than three-month-old content. Then prioritize reviewing your content to ensure that your knowledge base remains up to date. Focus on the content that needs to be fresh all the time, as new information needs to be added frequently. Apply good processes within your technical writing team.

Content structure

The content structure is crucial for GenAI-based agents to understand the importance of each subsection inside your articles. The article content hierarchy can be adhered to with the help of H1 – H6 tags. This semantic structuring of content helps GenAI-based agents retrieve relevant information inside appropriate subsections of your article to answer customers' questions. During the content audit, check for instances where article content deviates from semantic rules. After the audit, technical writers can focus on restructuring each section of the article so that the information flow is streamlined and the transition between different sections is clearer.

Article content

The article content needs to be revised so that all content is elaborate. Add more text content to make your content as explanatory as possible. Given the nature of GenAI to harness the power of natural language, it needs more content to understand the semantics and domain of your knowledge base content. When you review your content, ensure that the purpose of the article is clearly articulated and the scope is clearly defined. Depending on the nature of the article content, the procedural steps can be added in the following subsection of your article. Any other relevant content can be added to the article following good content structuring.

The pronoun “it” can be used within a paragraph, but not to reference anything in the following paragraph. Repeat the subject in new paragraphs. The rationale behind this is that GenAI-based agents use a Retrieval Augmented Generation (RAG) framework that works by chunking content to generate apt contexts. If “it” is used in the chunked content, it might lose relevance! After the audit, technical writers can focus on adding more content to existing articles so GenAI-based agents get a holistic perspective on your content. Figure 1 shows an example of an elaborate “Getting Started Guide” from Airtable. GenAI toolkits also have techniques to summarize article content if customers wish to interact with your knowledge base content directly instead of GenAI agents.

Figure 1: A “Getting Started Guide” from Airtable
Source: support.airtable.com

FAQs

While auditing content, check whether your article content has FAQs. If not, it is important to add these for each article in your knowledge base. The content for creating FAQs can be sourced from your customer support channels, so that GenAI can be better suited to answer common questions. Also, FAQs on existing content help GenAI-based agents understand the nuances of your article content. Technical writers should update FAQs at regular time intervals to cater to emerging customer questions. This new knowledge addition helps GenAI-based agents to generate reliable and consistent responses. Figure 2 shows an example of FAQs on article content.

Figure 2: FAQs need to be updated regularly to answer emerging customer requests.
Source: docs.document360.com

Business glossary usage

During the audit, assess any business terms in your entire knowledge base content. Using consistent business terms across your knowledge base is indispensable in eliminating ambiguity and confusion. For example, if you are using terms such as “clients”, “customers”, “users”, and “stakeholders” synonymously, GenAI might become confused, as the “sentence similarity” between those terms is very close, yet they might have different business definitions. GenAI-based agents might therefore produce inconsistent responses depending on which term is used in the customer’s questions. During the audit, create a business glossary that contains the list of business terms along with their definitions. Then revisit all the content to ensure that business terms are used consistently. This helps GenAI-based agents provide clear answers to your customers. 

Other aspects

Technical writers need to be aware of the preprocessing steps involved with the knowledge base content while deploying any GenAI-based agents. The vendor/developer implementing the GenAI-based agent can provide you with the list of preprocessing steps for your content. It is vital to understand how variables, snippets, tables, code snippets, and content labels are preprocessed before they are sent for further processing in the RAG framework. Technical writers should also seek information from vendors/developers on the content chunking strategies of their tool.

Next steps

The shift in customer behavior to accessing your content via GenAI-based agents calls for a content revamp. Your customers can now choose the format of the response they get from the GenAI, such as bullet points, tables, language preferences, and so on. Thus, writing content in a generic format and conversational style will help GenAI-based agents produce appropriate responses based on your customer persona and their personal preferences. After the content audit, the technical writing team should evaluate and prioritize efforts to strengthen their knowledge base content to cater to the needs of GenAI-based agents. If this requires a major revamp of your entire knowledge base, it is advisable to involve information architects.

Conclusion

A content audit is a must before deploying GenAI-based agents on your knowledge base. The existing knowledge base content must be revamped to cater to the characteristics of your GenAI-based agent. Having timely and accurate content builds trust in your knowledge base, which enables GenAI to provide reliable and consistent responses. Utilizing consistent business terms across your knowledge base will empower GenAI to generate responses with utmost clarity and reduce hallucination.

Let’s get your content ready for GenAI-based agents to engage with your customers for a richer knowledge experience.