Digital workplace

Field report: From idea to AI-based chatbot with Gemini on the Google Cloud

Liam Ormond
Liam Ormond
-
Published on
13.05.2024
Field report: From idea to AI-based chatbot with Gemini on the Google Cloud

AI-based chatbot with Gemini on the Google Cloud

We recently held an internal workshop on the topic of "AI-based chatbots". The aim was to delve deeper into the topic and at the same time make our employees' work easier and increase efficiency using an AI-based chatbot. 

What do we mean by an AI chatbot and how does an AI chatbot differ from chatbots? 

We distinguish between two types of chatbots: the intent-flow chatbot and the AI chatbot.

The intent-flow chatbot acts like a friendly tour guide, helping to efficiently achieve the goals of the asking user. It works like a well-organized decision diagram, identifying keywords or phrases and navigating along a predefined path to provide appropriate answers - similar to driving along a fixed route. 

The AI chatbot, on the other hand, behaves like an intelligent assistant that strives to understand the user's needs in depth and provide comprehensive support. It continuously learns from interactions with users, applies complex algorithms to understand the context of questions and offers personalized answers. It can also recognize subtle nuances in language and adapt accordingly, much like an experienced interpreter.

In summary, intent-flow chatbots are based on pre-programmed paths and use specific keywords to generate responses. AI chatbots, on the other hand, use machine learning and natural language processing to react dynamically to needs and enable a more natural conversational experience. 

The workshop as a starting signal for the creation of our AI chatbot

Our goal was clear: we wanted to develop an intelligent chatbot that would provide our employees and auditors with quick access to information by accessing our intranet. The AI chatbot was to use the intranet as a source of information based on Confluence and generate an interactive component in the chat using Gemini Pro, Google's new model. 

To do this, we delved deep into the world of Retrieval Augmented Generation (RAG) and Large Language Models (LLM)

Retrieval Augmented Generation (RAG): Retrieval Augmented Generation uses current or company-specific information from a knowledge database. In contrast to traditional methods, a semantic search enables topic-specific information to be found even without knowing the exact search term. This makes it possible to search for relevant information without knowing exactly what you are looking for. 

Large Language Models (LLM): A Large Language Model (LLM) is an advanced AI program that is trained with huge amounts of data to understand and generate language. This technology makes it possible to write texts, translate language and answer questions.

Effectively query and search for information with the AI-based chatbot


Quick and accurate answers without long searches‍

The decision to use Confluence as the primary source of information has proven to be advantageous. The chatbot developed pulls up-to-date information directly from our Confluence intranet, making access to answers from the associated sources quick and straightforward for all employees, new hires and external audit bodies. 

Employees can now retrieve information efficiently and in a targeted manner. This innovation has simplified onboarding and internal training and significantly reduced communication costs.

AI-based chatbot with Google Gemini on the Google Cloud

Further possible applications for AI-based chatbots

In addition to internal use, we see a number of other potential applications for AI-based chatbots:

  • Customer support: Fast and efficient response to customer inquiries around the clock.
  • HR assistance: Automation of standard requests for vacation days, payroll accounting and other HR-related topics.
  • Project management: Support in the coordination of projects by providing relevant documents and schedules on request.

Conclusion

The strong interest and positive feedback from the workshop confirmed to us that AI-based chatbots are an effective way to improve workflows.

We were excited by the potential that the further development of these technologies opens up. The next steps to integrate further functionalities are already being planned. 

Download now

Table of contents
What is an AI-based chatbot and how does it differ from other chatbots

Digital workplace

Field report: From idea to AI-based chatbot with Gemini on the Google Cloud

Liam Ormond
Liam Ormond
-
Published on
13.05.2024
Field report: From idea to AI-based chatbot with Gemini on the Google Cloud

AI-based chatbot with Gemini on the Google Cloud

We recently held an internal workshop on the topic of "AI-based chatbots". The aim was to delve deeper into the topic and at the same time make our employees' work easier and increase efficiency using an AI-based chatbot. 

What do we mean by an AI chatbot and how does an AI chatbot differ from chatbots? 

We distinguish between two types of chatbots: the intent-flow chatbot and the AI chatbot.

The intent-flow chatbot acts like a friendly tour guide, helping to efficiently achieve the goals of the asking user. It works like a well-organized decision diagram, identifying keywords or phrases and navigating along a predefined path to provide appropriate answers - similar to driving along a fixed route. 

The AI chatbot, on the other hand, behaves like an intelligent assistant that strives to understand the user's needs in depth and provide comprehensive support. It continuously learns from interactions with users, applies complex algorithms to understand the context of questions and offers personalized answers. It can also recognize subtle nuances in language and adapt accordingly, much like an experienced interpreter.

In summary, intent-flow chatbots are based on pre-programmed paths and use specific keywords to generate responses. AI chatbots, on the other hand, use machine learning and natural language processing to react dynamically to needs and enable a more natural conversational experience. 

The workshop as a starting signal for the creation of our AI chatbot

Our goal was clear: we wanted to develop an intelligent chatbot that would provide our employees and auditors with quick access to information by accessing our intranet. The AI chatbot was to use the intranet as a source of information based on Confluence and generate an interactive component in the chat using Gemini Pro, Google's new model. 

To do this, we delved deep into the world of Retrieval Augmented Generation (RAG) and Large Language Models (LLM)

Retrieval Augmented Generation (RAG): Retrieval Augmented Generation uses current or company-specific information from a knowledge database. In contrast to traditional methods, a semantic search enables topic-specific information to be found even without knowing the exact search term. This makes it possible to search for relevant information without knowing exactly what you are looking for. 

Large Language Models (LLM): A Large Language Model (LLM) is an advanced AI program that is trained with huge amounts of data to understand and generate language. This technology makes it possible to write texts, translate language and answer questions.

Effectively query and search for information with the AI-based chatbot


Quick and accurate answers without long searches‍

The decision to use Confluence as the primary source of information has proven to be advantageous. The chatbot developed pulls up-to-date information directly from our Confluence intranet, making access to answers from the associated sources quick and straightforward for all employees, new hires and external audit bodies. 

Employees can now retrieve information efficiently and in a targeted manner. This innovation has simplified onboarding and internal training and significantly reduced communication costs.

AI-based chatbot with Google Gemini on the Google Cloud

Further possible applications for AI-based chatbots

In addition to internal use, we see a number of other potential applications for AI-based chatbots:

  • Customer support: Fast and efficient response to customer inquiries around the clock.
  • HR assistance: Automation of standard requests for vacation days, payroll accounting and other HR-related topics.
  • Project management: Support in the coordination of projects by providing relevant documents and schedules on request.

Conclusion

The strong interest and positive feedback from the workshop confirmed to us that AI-based chatbots are an effective way to improve workflows.

We were excited by the potential that the further development of these technologies opens up. The next steps to integrate further functionalities are already being planned. 

Download now

Icon arrow