Inside the Eigen / Sirion Deal: Multi-Modal + Multi-Model

Why did CLM company Sirion buy Eigen? Was it like Icertis and AI-focused Evisort, or DocuSign and Lexion? The short answer is: no, Sirion was already heavily into AI. In this case Eigen’s addition has helped to form a larger, multi-model and multi-modal platform for the clients.

To explain more about the deal, which was announced earlier this month, Artificial Lawyer spoke to Ajay Agrawal, Founder and CEO of Sirion, and Lewis Liu, Co-Founder and CEO of Eigen Technologies (pictured above).

First some background on the two companies. Sirion, with about 800 staff, is nine years old, has raised $170m and has about 300 customers. It’s also very much focused on CLM and as noted has been working with AI for some time. In fact, Agrawal stated: ‘We are AI native, if our AI stops then the company stops.’ And they leverage the AI with legacy contracts and new ones coming into the inhouse legal team.

Eigen, with about 70 staff, can trace its roots back to 2015, but probably was first visible in the legal world a couple of years after that. It is very much an AI-first company, helping lawyers, banks and insurance companies extract key information from documents.

Now, you may say: ‘Hold on, so this is not about an AI bolt-on. They both have good AI capabilities. Then why do the deal?’  

As Agrawal explained: ‘Our clients asked us about performance data related to contracts that is in other documents, such as orders, invoices, and service level reports, as they want to know if they have got value from the contract.’

But, segments of these documents are not in regular text. There are numerical sections and tables, and other hard to handle information that natural language systems can find difficult.

‘We worked with a lot of the same customers as Eigen, such as ING and Morgan Stanley. At the same time their software is so versatile it can take information from any type of document. So, it became obvious to us that ‘contract intelligence’ needs to expand into ‘document intelligence’,‘ Agrawal explained.

I.e. Sirion’s value would be greater if it could handle information not just from the contracts the inhouse lawyers worked with, but with all the related documents and information streams that connected to those contracts, and that meant dealing with numbers and tables in addition to more text.

Agrawal added that despite the coincidence in timing their M&A deal had no connection to what has been a small wave of CLM combinations this year.

Liu then explained that ahead of the deal, his company was just in the process of gaining new financing and even had the term sheet all ready when the Sirion move was proposed. He added that other CLM companies had also made approaches around the same time.

After considering Sirion’s AI credentials in the CLM space, Liu said that he saw that this ‘is the path with the maximum outcome and it was the obvious option’.

The Tech

So, where are the companies on genAI? Agrawal stated: ‘LLMs are not a panacea for extraction from contracts. They are not always accurate and they can be too expensive because of token costs.’

Meanwhile, Eigen had been exploring something that Liu terms ‘small data AI’, which is smaller models for specific information. With this smaller, more narrow approach ‘you can still have natural language Q&A’ and gain the accuracy you need.

Liu explained that they had their first customer in 2017, when the Eigen system was running on ‘NLP and maths’. Then they moved to a more ‘probabilistic model, using small models that were tuned to domains such as finance and law, hence the term ‘small data AI’’.

When it came to transformer models, Liu – back in 2018 and long before many companies were exploring this approach – decided that for extraction of key data from contracts that LLMs, i.e. over 1 billion parameters, would not work well.

They went with BERT, a smaller model, and were one of the first to deploy it in this way. Then, as time went by and in part because Eigen had always been close to the banks and investment world – after all Goldman Sachs and Temasek were investors – they developed expertise in handling much more than text.

And so we get to the end result of the combination. As Agrawal stated: ‘This is an important moment to come together because the future will be multi-modal (i.e. images that contain tables, and text) and multi-model (i.e. ML/NLP as well as different approaches to genAI).’

He added that a strategy built on just one LLM is ‘flawed’ and that ‘there will always be new models’.

Liu concluded: ‘There is text; there is an image layer for tables and graphs; and then there is the human understanding level (e.g. via the natural language Q&A) where it all comes together.’

And via the combination they can now offer this to the clients, with all of Sirion’s CLM workflows in addition.

What Next?

The Eigen brand will live on inside Sirion because of its name recognition among major banks. Liu is now Chief AI Officer and will work on both tech and client relationships.

And the future of the CLM market, which has seen so much change? Agrawal doesn’t believe there can be so many companies in one space. There will be a major shakeout for sure.

But, for Sirion and Eigen, now combined, they firmly believe they are on the right trajectory together.

Or as Agrawal put it: ‘The wings of Eigen will let the bird of Sirion fly in the field of document intelligence.