Companies are increasingly turning to artificial intelligence (AI) bots to provide customer support. These bots use natural language processing (NLP) and machine learning to understand and interpret user queries, providing relevant information and solutions instantly. With the rise of AI support bots, it's become more important than ever to write a clear, comprehensive and optimized knowledge base and FAQ content to train these bots. 

Below, we will explain how you can optimize your knowledge base and FAQs for AI bots. We will discuss tips and tricks for writing clear and concise content, using structured data, and providing context as well as examples to help AI bots interpret user queries more accurately and resolve as many questions as possible without human intervention.

‍

Write in full sentences or paragraphs

When writing content for your knowledge base, it's important to use full sentences rather than relying on question and answer-style responses. AI bots rely on natural language processing to identify and interpret user queries, and they often use context clues from the preceding sentence to generate accurate responses. 

‍

If your knowledge base is structured in Q&A style and you cannot change this, it’s important to repeat the question within the answer. 

‍

DO ✅

Q: Is Discord free?

A: Discord is free to use

‍

OR, simply write a sentence, without a question preceding it:

‍

DO ✅ “Discord is free to use”

‍

When encountering related questions, such as "Does Discord require payment?" or "How much does Discord cost?", the AI can typically understand and extract the answer even if the question is not explicitly stated in your knowledge base.

‍

DON’T ❌

Q: Is Discord free?

A: Yes

‍

The way the data is ingested and scraped in Mava and many other AI support systems, means it won’t always link questions and answers automatically. Rather, it reads the document as a whole to understand your business and therefore it is not recommended to simply write “correct” or “yes”.

‍

Tip: In Mava, if you want to map a specific question to answer, you can do so in the “Template Answers” section, found within Mava’s Dashboard settings. 

‍

Write clear and concise language and define any jargon

It's important to use clear and concise language for your knowledge base because AI bots rely on natural language processing (NLP) to understand and interpret user queries. If the language used in the knowledge base is too complex or unclear, the AI bot may have difficulty understanding it.

‍

Sometimes you may have to use jargon, acronyms or product names. In this case, it’s important to explain the term and also explicitly define synonyms; normally the AI can pick up synonymous terms by itself but it may not be able to if it relates to something very niche or specific to your project.

‍

DO ✅ “Our Vaults, also referred to as savings account, are only accessible via our mobile app”

‍

DON’T ❌ ”Our savings account is only accessible via our mobile app”

‍

Since Vaults is the name of a product, specific to your business, the AI would not be able to understand that it’s the same product as a “savings account”. Therefore you will have to explicitly define this in the content in order for the AI bot to pick up user queries using language interchangeably.

‍

Make sure to explicitly refer to products and topics

If you have multiple products or are selling to multiple customer types, it’s important to define and refer to that in your content to make sure the AI doesn’t get confused. 

‍

DO ✅ “The fees for our premium plan are $20 / month”

‍

DON’T ❌ “Our premium plan has tons of benefits such as (...). It also allows you to (...). It costs $20 / month“

‍

If the content is grouped together and discussed in the same paragraph, it is likely the AI will still understand the context, but to increase accuracy, it is best to avoid terms like “they” and “it” if referring to something discussed earlier in the text. Instead, be specific about what topic or product you are referring to.

‍

Use headings to structure your text

In order for the AI to more easily and quickly scan your knowledge base, it is helpful to structure your content. Typically, the AI won’t go through the entirety of your content to answer a question, but rather it picks out the parts of the text where the question is most likely to be addressed, and then focuses its search there.

‍

DO ✅ Add headings for core themes to separate and organize your content.
For example: “Technical Documentation”; “Pricing”; “Security”

‍

DON’T ❌ write one long block of text without any paragraphs, headings or structure

‍

Add text and descriptions to your images

If your knowledge base contains any images or videos, for example a diagram, make sure to also add a text explanation so the AI understands what’s going on.

‍

DO ✅ write out and explain content in an image.

Below are all the wallets supported on our platform: MetaMask, Coinbase Wallet, WalletConnect, Ledger, Phantom, Bitkeep. 

‍

DON’T ❌ simply upload an image.

Below are all the wallets supported on our platform. 

‍

‍

Frequently update your knowledge base

Frequently update your knowledge base to make sure the information is up-to-date. Keep track of common questions and topics by tagging questions handled by your support team, so that you can create additional knowledge base content on common issues and topics not yet addressed.

Moreover, make sure that the AI bot you are using is fetching the latest information. With Mava, you can simply press the “refresh” button in your AI settings, to re-upload your website and knowledge base after you’ve updated any content. 

In conclusion, using AI bots for customer support is a great way to provide fast, efficient, and cost-effective support to your users. By optimizing your knowledge base for AI bots, you can improve the accuracy of your bot’s answers, save time for your agents and provide a better overall user experience. 

At Mava, we enable you to upload a link to any existing knowledge base and instantly turn this into an AI-enabled support bot that works across Discord, Web and soon Telegram. Try it out for free here: https://dashboard.mava.app/

‍