Harnessing the potential of LLMs and ChatGPT with your own AI Assistant

ChatGPT and Google Bard are currently breaking all user number records. With these large language models (LLMs), AI is increasingly entering our professional and private lives and fundamentally changing how we work.

Our AI Assistant is an intelligent chatbot that accesses all your company data and – thanks to ChatGPT’s existing computational power and intelligence – makes many tedious and tiresome administrative tasks obsolete. Enhance the quality of your customer service, streamline processes, and relieve employees of time-consuming tasks.

Our AI Assistant use cases

Organizations and companies from all industries have already implemented the AI Assistant and leveraged the potential of ChatGPT and LLMs for their business.

Deploying an HR chatbot in the public sector

The HR team of an administrative body in the public sector serves 2,000 employees, ranging from office staff to police and firefighting personnel. Due to different laws and regulations, HR policies vary significantly depending on the role or seniority level. The HR chatbot has access to a wide range of non-sensitive data, including regulations, legal texts, and pay scales. This means it can handle queries up to a certain level of complexity in a fully automated process. HR employees receive the response generated, which they can review before approval. Where necessary, they can add additional information.

Giving customers detailed answers

A manufacturer of highly complex laser devices has enhanced its customer support chat system with the AI Assistant. The new NLP chatbot provides customers with detailed answers to their questions by referencing information from documents and videos. For instance, the chatbot provides the relevant video and even indicates the time stamp within the video where the requested information can be found.

Streamlining compliance checks

One of the tasks of the Liechtenstein Financial Market Authority (FMA) is to conduct compliance checks on exposed individuals, such as board members and compliance officers. This involves searching public databases for critical reports on the respective individual. Thanks to the AI Assistant, the FMA has been able to automate this process: The AI Assistant performs the search, downloads the search results, and checks them for controversial content. The results are then verified by experts, thereby continuously improving the compliance process.

Check out the blog post!

Transcribing voice recordings

An association allows its members to send questions as voice recordings in various languages via WhatsApp. These questions are transcribed and answered by the AI Assistant. Combining AI-generated responses and the convenient voice recording feature creates a user-friendly customer experience.

How our AI Assistant for companies works

The AI Assistant is a natural language processing chatbot, or an NLP chatbot for short. “Natural language” refers to languages spoken by humans, as opposed to other languages, such as programming languages. The AI Assistant searches your company data and provides a coherent, precise, and linguistically appealing response thanks to its integration with ChatGPT. To enhance transparency, the sources on which the response is based can be indicated.

Basically, the procedure of the AI Assistant can be divided into the following four process steps.

Step 1

Define the data pool and index data

You decide which internal company data is made available to the AI Assistant. The data can take the form of individual documents, such as PDFs or Word files, be stored in a system like SharePoint, in a database, or even online, for example, on your company’s website or YouTube channel. Once the data pool has been defined, the data provided is indexed.

Step 2

Create a search index

The indexed documents are now stored in a search index. This index consists of vectors known as “embeddings” that represent the meaning of an indexed document.

Step 3

Retrieve relevant information

When a user asks a question, the most relevant sections of all the indexed documents are identified and retrieved. To ensure these pieces of information are quickly retrievable, the retriever model used for this purpose must be executed locally.

Step 4

GPT-4 computes the answer

Along with the question, the most relevant sections of the documents are sent to GPT-4. Based on the information provided, GPT-4 generates a coherent, linguistically appealing response for the user.

The new power of NLP chatbots compared to ordinary chatbots

The AI Assistant combines information from various sources and formats to provide coherent responses in a chat dialog. Complex indexing of the data is no longer necessary. The AI Assistant considers information from your connected data sources and delivers answers not only in natural language but also in other formats, such as Excel spreadsheets. Find out what makes the AI Assistant different from ordinary chatbots.

Search method

Traditional search functions provide a large and incomplete number of unstructured documents, which the user must search through to find the correct answer. By contrast, NLP chatbots combine information from various sources and deliver summarized, coherent responses in a conversation.

Results presentation

Until now, users had to search for the required information again in all the available data repositories, evaluate it, and manually transfer it into new documents each time. NLP chatbots, on the other hand, search all defined sources and not only generate responses in natural language but also prepare them in various formats, such as Excel.

Data management

Until now, maintaining data repositories to help employees navigate them was a cumbersome task. That’s now a thing of the past, as the AI Assistant accesses all the indexed data, whether it’s on your intranet or website, in business applications, or stored elsewhere. This significantly reduces your data management workload.

Do you want to learn even more about innovation and AI? Download our free magazine now!

Having already gone through two “AI winters” in the last 50 years, AI is now back on our doorsteps – younger, stronger, and more dynamic than ever. Download our latest magazine on AI and innovation for free to find out more about this technology.

More of ti&m special

Why ti&m?

We take an integrated four-step approach to connect your company with AI. Our experts will support you in harnessing the full potential of AI so that you can take your business to the next level. You decide where to start and how much support you want from us.

Learn more

Head of Machine Learning

Ursin Brunner

Do you have any questions about the ti&m AI Assistant, ChatGPT, or LLMs, or would you like to work with us? Then get in touch – we’d love to hear from you!