Posted January 28, 2025

While researching for a blog, it was observed that accessing websites took significantly longer than expected, with each page taking around one to two minutes to load after entering the URL or clicking on a bookmark. This raised questions about whether the issue was related to the performance of an older Mac computer or if the internet speed was unusually slow.

After troubleshooting, I discovered that my internet speed of 100 Mbps was not cutting it, so it was time to upgrade. Since I use Spectrum as my provider, I signed into their website to make the magic happen. After another minute or two, I finally accessed the Spectrum homepage, signed in, and reviewed my account. After confirming my internet plan, it was time to upgrade.

So, a decision to make, call the support desk? Sure, if I had 10-15 minutes of spare time. Click on the upgrade button. Chances are you get the famous 404 page – not available, or it tells you that you need to call the help desk anyways, or do I click on “Chat with Us?” Why not? I will try something new.

Hello Chatbot

After clicking on “Chat with Us?” I received a greeting and was asked, “How can I help you today?” Below, it gave me three choices: pay my bill, report an outage, or view my statement. Or, if I wanted, I could type in a question. After typing my question about available internet speeds, I received an answer in less than five seconds!! Wow, this is great. Can any human type be that fast? No, I was interacting with a chatbot.

What is a Chatbot?

If the above scenario sounds familiar, you are dealing with a chatbot. A chatbot acts the same as a user searching for frequently asked questions on a website. 

According to Oracle, chatbots are computer programs that simulate and process human conversations, allowing us to interact with a digital device, like talking to a natural person.

People are using chatbots more frequently to assist with needs such as upgrading internet service, what the weather will be like at their vacation destination, or the score of the basketball game that took place earlier today.

There are chatbots that can analyze, monitor, and evaluate customer calls. Analysts will use this information to find out what customers are not happy about or if they are satisfied with the solutions that are being offered. 

Today, chatbots are programmed to do many things besides answering basic questions.

Examples of Chatbots

In the past year, OpenAI has developed some very familiar generative AI models, such as Chat GPT4 and DALL-E. These models receive user prompts or instructions and send back AI-generated content. GPT4 understands and produces natural language. DALL-E creates visuals.

Why Chatbots?

Just imagine over 40,000 search queries are made online every second worldwide. According to a recent study, “Blinded by Delight: Why your customer experience should be effortless, not exceptional,” shows that 22% of customers are more likely to repurchase and increase their spending when their queries are answered quickly, or the information is more accessible. However, if a customer must make a lot of effort to make a purchase or get something done, they are 80 times more likely to comment negatively.

Advantages of Chatbots

Besides making life easier for customers, other advantages include:

• Rapid processing of standard questions
• Improved customer support reaction time
• Less work from the customer support team

Disadvantage of Chatbots

Some customers often prefer to avoid the user experience provided by chatbots because they are perceived as cold and distant. However, with DALL-E, the AI team can generate avatars for the virtual assistant or the client’s profile. These may be able to reinforce a more natural conversation flow, and customers would benefit from a more realistic experience.

Preparing the Chatbots

So, how do chatbots get all this information? How does it know that the available internet speeds at Spectrum are 100 Mbps, 200 Mbps, 300 Mbps, 400 Mbps, and a Gigabit? Or that the 40,000 searches are conducted each second worldwide? Most likely It comes from reusing content from technical documentation.

Chatbots are being prepared for primetime right out of the box. They must learn the valuable information needed to answer the customers’ questions. The good news is that the data already exists in the technical documentation. Reusing the content to set up the chatbot saves the organization costs and improves customer interaction.

There is a lot of information from the documentation that can be used to program the chatbot. However, starting off slowly with specific use cases is recommended. Help desk tickets, support hotlines, social media discussions, and FAQs are valuable resources. 

Afterwards, enhance the information with metadata. Metadata assists the chatbot in finding the correct data based on the customer’s inquiry.

So, how does the chatbot understand the customer’s inquiry? It must be able to interpret the question and get the answer. However, the customer’s inquiry may not provide all the information the chatbot needs to answer the question. If that is the case, the chatbot needs to use a dialog system to ask more questions to gain specific information from the user.

The two types of dialog systems are:

1. Task-Oriented: Task-oriented dialogs follow a decision tree for specific tasks, such as customer service or technical support. A good example is updating my internet package with Spectrum.
2. Open-domain dialog: The chatbot uses free-form statements to give answers or inquiries about missing information. These dialogs use computer linguistics and artificial intelligence. A few great examples are Microsoft Copilot, ChatGPT-4, and DALL-E.

Preparing the Documentation for the Chatbot

There is a process to prepare to reuse documentation for the chatbot:

1. Identify the use cases. Use the information from the customer’s point of view.
2. Select the topics from the technical documentation linked to the use cases.
3. Establish the appropriate information for the chatbot to respond to the user’s query.
4. Select a dialog that meets the organization’s requirements for chatbot communication. Only use dialog for the highest use cases.
5. Customize the content by tailoring the information to the chatbot’s target channels and users.
6. Publish the dialog system and contents. Move the topics into a format and link the issues. Use a content delivery system that can help support the chatbot.
7. Protect the information. For example, companies can use Azure Open AI to provide data privacy and internal technical documentation. It is one of the safest options when dealing with confidential data.

Summary

Chatbots aim to make life easier by answering customers’ inquiries such as “Why do my glasses have spots after the dishwashing cycle or who has the most rushing yards in the National Football League this season?” With the creation of Microsoft Copilot, Chat GPT-4, and DALL_E, a user can ask the chatbot to write code, answer emails, review the budget, or plan the following organization’s get-together.

However, as excellent as this may seem, the chatbot is still being prepared to go when it is first created. It needs to have information for it to do its job. That is why it is essential to have up-to-date technical documentation. Start with the most inquired information and then build up. This will get the customers used to and trust the organization’s chatbot.  The more they gain confidence, the happier they will be and the more money they will spend on your product or service.