1Preface

With the rise of ChatGPT, enterprises are actively exploring the potential advantages of AI in their daily operations. Our organization harnesses the capabilities of ChatGPT and other LLMs(Large Language Models) to provide round-the-clock, multilingual support and exceptional semantic comprehension, aiding not only our internal staff but also external customers. However, to overcome these LLMs' inability to address company-specific inquiries, we created ChatInsight, a solution that enables us to access company-internal knowledge and generate tailored responses accordingly.

2What is ChatInsight?

ChatInsight is a knowledge-based AI chatbot. It leverages the powerful semantic comprehension of AI and is capable of answering questions despite being phrased differently. Based on the Azure OpenAI GPT 3.5 and 4.0 model as well as LLMs, ChatInsight allows users to ask questions in a natural and intuitive way and always reply with accurate responses. With its 24/7 service and multilingual support, ChatInsight can become your best consultant, handling internal questions for your staff or external service queries for your customers.

3ChatInsight for Corporation

ChatInsight has many advantages and can help corporations improve efficiency, reduce costs, provide intelligent services. Additionally, it ensures secure sharing, compliance, and seamless integration. Many enterprises have already benefited from using ChatInsight.

  • Company-owned knowledge base management: Enterprises can establish their own knowledge base according to their different business needs. This enables chatbots to be utilized by internal employees for inquiries about the company's internal information, as well as by external users for inquiries about product knowledge.
  • Cost Reduction: The deployment of ChatInsight can result in significant cost savings for businesses, which include a decrease in labor costs, a reduction in error rates, and an enhancement in productivity.
  • Secure Sharing: Corporations can regulate access to ChatInsight by setting up passwords, IP restrictions, and allowed domains. This strategy ensures that only authorized people have access to the chatbot, enhancing collaboration and efficiency.
  • Compliance: ChatInsight is powered by the Microsoft Azure OpenAI model, which is bolstered by Azure Service Level Agreements for enhanced reliability. Furthermore, Microsoft Azure conforms to an extensive array of international and industry-specific standards such as GDPR, ISO 27001, and HIPAA, among others. This ensures that the usage of ChatInsight is in line with relevant regulations.
  • Integration and Automation: ChatInsight can integrate with AirDroid Business or enterprises' information systems. Users can directly call its functions by interacting with the bot created by ChatInsight, for instance, querying device reports, executing predefined workflows, checking customer orders or logistics information on e-commerce platforms, etc. This means enterprises can build an intelligent assistant integrated with AirDroid Business or the company's information systems via ChatInsight. Enterprises can understand the device information and status and manage these devices through ChatInsight, without having to go through AirDroid Business for everything, simplifying the process and difficulty of device management.
  • Embed: Corporations can embed the bots created by ChatInsight into your website via a widget or iframe. Alternatively, you can deploy them to your enterprise-managed devices or BYOD devices through AirDroid Business, or directly embed them into your app's in-app browser via a URL.


4ChatInsight for End User

ChatInsight can greatly improve end-user satisfaction in many aspects.

  • Multilingual Support: ChatInsight offers support in multiple languages, thereby meeting the needs of users in different regions.
  • 24/7 Availability: Chatbots created by ChatInsight can provide services at any time and anywhere without rest.
  • Immediate response: With ChatInsight, end users can receive real-time responses and feedback, eliminating unnecessary wait times. This significantly enhances problem-solving efficiency.
  • Human Interaction: ChatInsight can understand and simulate human language and emotions to provide users with a more human service experience. Whether answering questions or performing complex tasks, chatbots can communicate with users in a more friendly and patient way.

5Use Cases of ChatInsight

Around-the-clock AI assistants integrate with enterprise information systems and fully support your business in the aspects of such as, but not limited to, sales, marketing, customer service, and technical support. Here below are some example use cases.

  • Online Customer Service: ChatInsight can provide instant and professional responses on product-related information to satisfy customers' needs.
  • Marketing Conversational Chatbot: ChatInsight can understand what your customer wants while optimizing and customizing service support according to your needs.
  • Company Wiki: ChatInsight allows teams to quickly ask for general company information like workplace policies, guidelines, workflows, and the code of conduct.
  • Employee Education Training: ChatInsight can serve as an internal consultant to respond to all questions for employees, especially newly onboard ones, and free up your HR/IT teams to focus on growth.

6Beginner's Guide

The steps below are the basic guidelines for you to complete the process of implementing the chatbot in your business with minimized uploaded documentation. To fully optimize the chatbot and knowledge base will take some effort. Just remember, an AI also needs proper training, just like one of your employees, and we can provide some guidance if you are willing to share your thoughts with us.

Step 1: Organize Your Knowledge Base

The first step of experiencing ChatInsight is not the bot but the Knowledge Base. With a proper knowledge base, ChatInsight will be able to provide the information you designed to your end user. Here is the most basic principle an article should have in ChatInsight's Knowledge Base.

  • Title of the article - The heading format is a must for any Title. With the heading format, ChatInsight can know where is the start of an article. We also suggest you use multi-layers of headings for better article structure if you have a long article.
  • Instruction of how and when to use the article - Although the AI model ChatInsight is using has an incredible capability of comprehending different types of information and answering in multilingual, we still suggest you add some instruction before the content of the article to increase the overall quality of answers.
  • Translate the terms or important sentences for ChatInsight - If your target audience uses a different language and there are specific terms in your knowledge base, such as industry jargon, AI may not provide the correct translation for these terms in different languages. Clarifying some terms or directly explaining the correct translation of a term in a certain language can greatly enhance the accuracy of translation. Moreover, artificial intelligence models can understand information well and only need a little extra support without the need for perfect translation in all languages. For instance, when discussing solutions like AirDroid Business, which is an MDM/EMM Solution, the following explanation would be helpful:

Please note : The phrase "attended and unattended Android devices" refers to "Android devices with or without human supervision."

Here is the most basic example of what an article should look like with the principles above. Just a reminder, you can put multiple articles in one document with proper use of heading format:

example-chatinsight

After you complete your first article, you can click "Save and train" and let's start to build your first chatbot.

With the existing articles you used to use, the AI bot can provide decent and accurate answers to most of the questions from those articles. However, if you are seeking better quality services, such as less robotic answers or interactive conversations, you will need to further tailor your knowledge base with some advanced tips. If you are interested in those tips, you are welcome to book a meeting with us.

Step 2: Create Your Bot

Once your knowledge base is prepared, it's time to create your first chatbot. Here are the basic steps to create your first chatbot:

  1. Give a unique name to the bot. This name is an internal for your manager to distinguish bots.

  2. Set a displayed title for your bot. This is the displayed title when the end users open the bot to chat and it can be the same as other bots.

  3. Set an example query. We suggest you ask the question your first article is meant to respond to.

  4. Select the Data Source for your bot.

  5. We suggest you keep the rest of the settings as default and click "OK" to complete the process.

From here, your first bot has been created, and you can start to experience your first AI chatbot provided by ChatInsight.

We provide lots of options for you to tailor the Bot to fit your business operation. If you are interested in how to set up and need some consultation from our specialists, please don't hesitate to book a meeting with us.

create-new-bot

Step 3: Test Your Bot

Everything released by a company to end users must go through a proper testing process; there is no exception for AI chatbots. As a result, we have a testing ground for you to test the bot, and you can also use this testing ground to replicate the answer when an end user is not satisfied.

  1. Click the name of the Bot, and you will be led to the testing page.

  2. You could click on the example query you just set or type in the question. Please note that with default settings, the Chatbot will only refer to your knowledge base to generate the answer.

  3. If there is no reference under the answer, it means the relevance of the question and the knowledge base is too low. In this case, you should edit your articles to make them more relevant.

  4. There will be references under the answer once the bot generates the response from your knowledge base, and you can check the original article by clicking those references. If you want to edit your articles, you can also do it from there.

test-chatinsight

After a couple of tests, you may want to know what methods we provide for you to implement the chatbot into your business operation. Let us introduce the methods to you so you can decide on the next implementation step.
We understand that article optimization and AI training processes may be frustrating. As a result, we provide consultation services; if you need some suggestions on these processes, feel free to book a demo meeting with us for further assistance.

Step 4: Share Your Bot

The last step of creating a chatbot is releasing it to your target audience. This step will be as traditional as it is; we provide three common methods for you to share your bots.

  • Share URL: You can share the URL of the chatbot to your end users, and they will be able to send the inquiry via the webpage. If you are willing to take a step further, you can even implement the URL to Apps just like us to replace the FAQs, help centers, or feedback, and the user can send their inquiries with the in-app browser without leaving the application.
  • Embed as an iFrame: You can embed the chatbot into your websites or applications if your application supports HTML. Please consult with your IT department on how to implement the codes.
  • Embed as Chat Widget: You can embed the widget code into your websites or applications and let the chat widget float anywhere you want. You can also tailor the widget's style, icon, and instruction. Please consult with your IT department on how to implement the codes. You will be able to release the bot to the enrolled device if you are also using AirDroid Business. This may further save the resources of supporting your end-user with simple device usage questions. Please let us know if this is your most needed function.

sharing-chatbot

7What's Next

From here, you already learned the process of implementing a ChatInsight Bot to your business operation. The next step is to make sure the bots can serve as one of your contact windows and answer the questions from your end users. You can try to expand your knowledge base, play with the advanced options on the bots, or implement some advanced functions we provided to your articles. You are welcome to book a meeting with us for an advance consultation on how to utilize ChatInsight.

8Advance Guide for ChatInsight

This section will provide a basic introduction to advanced settings, functions, and advice. If you want to know or need full guidance on these topics, we suggest you book a meeting with us since including all related information in this document is impossible.

Advance Settings of Chatbot

  • Prompt: In the Prompt section, you can give some detailed instructions to guide the actions and outputs of the bot. For example, you can ask the bot to guide the user to contact your customer service when the bot receives no references for the question. In the System Prompt section, you are suggested to insert the characteristics of how the bot should answer the question, such as being gentle, friendly, rigid, etc.
  • Temperature: This option affects how creative the bot's answer is. The higher the Temperature, the higher the chance the bot will try to use synonyms to generate the answer based on your knowledge base. If the Temperature is too high, the bot may generate some illusion based on the references. If the Temperature is too low, the answer of the bot may be boring.
  • Relevance of References: When references meet the lowest relevance, the bot will use those references as the major information sources and combine them with what the question is asking to generate the information. This means that when the question matches the citation very well, the bot can generate higher-quality answers with relevant sources and vice versa. As a result, finding the balance of how relevant a reference the bot can take as a major source is quite important. It's crucial to note that the relevance between queries and a knowledge base is not constant. The same question may have varying degrees of relevance to the knowledge base in different languages. Therefore, in order to enhance the performance of the knowledge base across multiple language environments, we can enrich it by adding keywords or key phrases in various languages associated with these queries. This way, the AI can reference information across diverse languages, thereby improving its adaptability.

Don't hesitate to meet with us if you are not sure how to adjust these options to fit the bots to your operation better.

Advanced Tips of Optimizing your Knowledge Base

  • Avoid Using Plain Text Files: Just like humans, the AI chatbot needs structures and formats to know what information is more important and valuable for referring. As a result, the AI will comprehend well-structured content better than plain text files.
  • Distinguishable Content: Even humans can be confused when the content is ambiguous, and AI is no exception. As a result, even if it is just a Yes or No question, you should add some explanation so that the AI can know the reason for the answer and provide a proper explanation to the asker.
  • Additional Rules in Article: Some rules may only apply to some of the articles, so adding them to prompts may not be the best solution. In this case, you can add the rules in your article for the bots to follow when referring to these articles.

A better-performing bot is everyone's dream, so meet with us as soon as possible if you are willing to fully optimize your knowledge base for the bots.

Enhance End-user's Experience

  • Prepare Answers for Miscellaneous Questions: You can prepare a knowledge base that is dedicated to some information that all your employees should know but not the main targeted questions you want the bot to answer. This may seem unimportant, but it can tremendously decrease the possibility that the users think the chatbot is just a bot when interacting with it.
  • Including Various Ways of Asking A Question: By adding various ways of asking one question to the article, the AI chatbot will have a better capability to locate the related information when the question's wording or structure is insufficient.

Creating content from nothing is hard. Thus, book a meeting with us for instructions on how to write these documents and templates.

ChatInsight Beginner's Guide

Unlock the power of your AI co-pilot for business by discovering how to use ChatInsight.AI!

Try ChatInsight.AI for FREE Now

Still need help? Submit a Request >>