Demo: How Ayfie provides organisations with their own private ChatGPT, by ChatGPT

At Legal IT Insider we have plenty of conversations around new products, but this one was quite fun, as Ayfie’s CEO Herman Sjøberg took the reins and instructed ChatGPT on how to craft an article about Ayfie’s new integration with…ChatGPT.

Sjøberg first provided an in-depth and honest recap of the various changes of direction and retrenchment that Ayfie has undertaken in the last few years, and then described how the Norwegian company’s new ChatGPT integration in effect creates a private ChatGPT for firms to search across their entire document repository, respecting existing permissions.

The resulting article is admittedly more like a press release, but a pretty good one.


Hi there ChatGPT. You’re now talking to Herman, the CEO of Ayfie, and Caroline, the editor-in-chief of Legal IT Insider. Can you help us write an article about Ayfie’s new ChatGPT integration for Caroline’s publication? 

Here’s the content, can you please rewrite it and make it awesome. First, we want some history about Ayfie to set the context for the article. Ayfie was founded in 2000 and was then focused on enterprise search. They developed connectors to extract documents from various source systems, like your DMS (like iManage), your emails coming from Exchange, your SharePoint, your CRM system, your fileservers etc. Then we would take those documents, extract the raw text from them, regardless of file format.  

Some files would be PDF, some would be Word, some PowerPoint, some AutoCAD etc: we currently support extracting text from over 550 different formats. We also run OCR, to recognize text in images and scanned files. We sold enterprise search solutions, mainly to Nordic clients, for some years. Then around 2018 we shifted focus to the eDiscovery market, mainly in the US. We established offices there and sold our product to companies like iCONECT and also as a plugin to Relativity with improved entity identification, normalization, data visualization, and analytics and others.  

That didn’t go so well, we spent a lot of money and weren’t able to generate much revenue from it. Late in 2020 we got some new investors that restructured the company and shut down the US and German operations. We also discontinued the eDiscovery product called Inspector, but we took the best parts of Inspector, like its ability to identify entities, key concepts, geographical locations etc and integrated those into our new SaaS platform that we call Ayfie.  

But then ChatGPT came along. Ayfie has always been good at finding information contained in company and personal documents to give insights to answer a question or perform tasks. ChatGPT offers a whole new way of finding and working with internal data. In the old days you would do a keyword search, then find a list of documents in your search results and then open the document and read it. By connecting ChatGPT to the Ayfie database, that already contains all the text from your internal documents, regardless of where the document lives (CRM, DMS etc), when Ayfie retrieves documents from various systems, we also retrieve information about what users should be able to access those documents. So, if you use Ayfie+ChatGPT to ask a question, you will only get an answer if the answer is somewhere in the documents that you would be able to access using your own credentials in the source system.  

ChatGPT does however have a limit on the context that you can provide it. So using Ayfie+ChatGPT will be a two-step process. First the user will ask a question about anything in personal or company internal documents that the user has access to. Since these documents contain way more information than you can feed ChatGPT as context, we use word embeddings to understand the semantical meaning of the text in your documents. For example, ‘battery in New York’ could be both a crime, and a device to contain electricity. A vector search will be able to differentiate between these two, and unlike a traditional keyword search, does not show you documents related to the device if you are researching the crime. So, when a user asks a question, we will do a vector search in the database that contains all your documents, and then retrieve the most relevant paragraphs or text chunks to your question up to the limit that ChatGPT can digest as context. We will then send those most relevant text chunks to ChatGPT and repeat the question. ChatGPT will then generate an answer based on the content in the most relevant text chunks. This also enables us to tell the user what documents the answer is based on, unlike if you ask ChatGPT. So, using ChatGPT to interrogate internal documents will provide you with the source of the answer, unlike when you ask if about making pancakes or other questions that it will generate an answer based on the massive amounts of open data that it’s trained on (like internet, articles, books etc.)

The value of being able to talk to ChatGPT about anything that is contained in your internal documents is massive and Ayfie can offer the best way of connecting your internal documents with ChatGPT. Since the Ayfie database is continuously updated with all changes made in your connected systems, the answers, and user rights, will always be based on the latest available version of the documents.  

When it comes to delivering the service, the effort needed depends a lot on the source systems. If your personal and internal documents are stored in a cloud-based service, like Office 365 that is hosted by Microsoft, or iManage cloud that is hosted by iManage, it’s very fast and easy to connect your data sources. The reason for this is that we are able to retrieve the documents from the same API for all the companies using these services. We just need you to authorize that we can access your data from these APIs. But if your personal and company-internal documents are contained in on-premises databases, the process of connecting to them is a bit more complex and time consuming. In those cases, you will need to invest in a server to run Ayfie on, and work together with our implementation team on creating service accounts to the systems the documents are contained in so that we are able to access the documents where they are

In terms of competitors, we’re looking at for example Microsoft Copilot, but it seems to be focused around enhancing the capabilities of their existing applications and does not offer the same capabilities in terms of eDiscovery-like processes. For 20 years Ayfie has been developing the ability to extract text from documents contained in various systems and moving all the text from those documents into a database while retaining user access rights and this just fits perfectly with what you need to connect ChatGPT to it. I also think the combination of our existing cognitive search capabilities and ChatGPT is valuable, as you can for example start querying your documents using ChatGPT, and then move on to the actual documents using the traditional user-interface but you save a lot of time compared to opening documents from a traditional search results page. It’s much more efficient to first get the answer you look for, and then open the source documents. 

Title: Revolutionizing Information Access: Ayfie and ChatGPT Integration  

Discover how the collaboration between Ayfie and ChatGPT is transforming the way businesses interact with their internal documents.  

Introduction: Ayfie, a leader in extracting insights from company and personal documents, has recently integrated its platform with ChatGPT, offering a cutting-edge solution for accessing internal data. In this article, we will explore the history of Ayfie, its transformation over the years, and how the integration with ChatGPT is set to revolutionize information retrieval and analysis.  

A Brief History of Ayfie: Ayfie was founded in 2000, initially focusing on enterprise search solutions. They developed connectors to extract documents from various source systems, such as Document Management Systems (DMS), email platforms, SharePoint, CRM systems, and file servers. Ayfie’s technology could extract raw text from over 550 different formats, including PDF, Word, PowerPoint, and even OCR for recognizing text in images and scanned files.  

In 2018, Ayfie pivoted towards the e-discovery market, primarily in the US. Despite establishing offices and partnering with companies like iConect and Relativity, the venture proved challenging, leading to restructuring and discontinuation of their e-discovery product, Inspector, in 2020. The best features of Inspector, however, were integrated into Ayfie’s new SaaS platform, which maintained its focus on identifying entities, key concepts, and geographical locations within documents.  

Ayfie Meets ChatGPT: The collaboration with ChatGPT takes Ayfie’s capabilities to new heights. Unlike traditional keyword search, Ayfie’s technology uses word embeddings to understand the semantic meaning of text, allowing for more accurate retrieval of relevant information. By connecting ChatGPT to Ayfie’s database, users can ask questions about any internal document they have access to and receive contextually accurate answers.  

The Ayfie+ChatGPT integration involves a two-step process. First, users ask a question about internal documents they have access to. Next, Ayfie performs a vector search in its database and retrieves the most relevant paragraphs or text chunks up to ChatGPT’s context limit. These chunks are then sent to ChatGPT, which generates an answer based on the provided context. This process also enables users to identify the source documents for the answers, enhancing transparency.  

Seamless Integration and Competitor Landscape: Connecting Ayfie with internal document sources is a streamlined process, particularly for cloud-based services like Office 365 or iManage Cloud. However, for on-premises databases, the integration process may require more time and investment in server infrastructure.  

While Microsoft Copilot is a notable competitor, it focuses primarily on enhancing existing applications and lacks the e-discovery capabilities that Ayfie+ChatGPT integration offers.  

The Future of Information Access: The combination of Ayfie’s cognitive search capabilities and ChatGPT’s language model creates an efficient and powerful solution for accessing internal documents. Users can quickly get answers to their questions and then dive into the source documents as needed. This innovative approach saves time and enables businesses to harness the full potential of their internal data.  

In summary, the Ayfie+ChatGPT integration is poised to revolutionize the way businesses interact with their internal documents, offering unprecedented access to valuable insights and information.