by Michel Klompmaker
We recently spoke with Arnoud Engelfriet in the heart of Amsterdam about Lynn Legal’s legal tech solution. Lynn Legal is actually a kind of cross between a modern IT company and a legal consultancy. Founders Arnoud Engelfriet and Steven Ras (see photo from left to right) are convinced that legal processes can be organised more intuitively and efficiently thanks to the application of modern technology. Their lawyer bots have already reviewed more than 14,000 NDAs and more than 1,000 data processing agreements since the beginning of 2021. In total, more than 50,000 legal problems have been spotted and resolved. We spoke with Arnoud about the background, the present and the future.
Lynn is trained on thousands of documents and recognises many variations with a margin of error of less than five percent. The customers themselves retain an overview because all reviewed documents are available via the control panel, including the review findings. And it goes fast too, on average the full review takes one minute and 20 seconds. Customers themselves determine the degree of rigor or indulgence. They themselves are in control via a playbook, in which various workflows are possible, whereby the link with the Contract Management System of the customer is also possible.
First, please tell us a little bit more about your background and who you are.
Arnoud Engelfriet: “Lynn Legal bv is a sister organisation of the cloud platform for Legal Tech applications JuriBlox, which we founded from the legal consultancy ICTRecht. We have been involved in ICT and internet law for more than ten years. Fortunately, thanks to our own software for document creation, we have a lot of recognition from the market. We have organised it so that JuriBlox and Lynn Legal are independent entities and they operate separately from ICTRecht. We ourselves, that is to say Steven Ras and I, are lawyers by nature, but with a special interest in IT, and that cannot be said of every lawyer.”,
Please tell us more about the mission statement of the new company, Lynn Legal.
Arnoud Engelfriet: “We have been convinced for some time that legal processes can be set up more intuitively and efficiently. With that conviction, we have developed the Legal Tech platform. This platform solves various problems, but above all ensures that both lawyers and non-lawyers can keep the necessary legal documents and processes in order. In other words, we are real lawyers with an understanding of technology. We recognise better than anyone where the bottlenecks, and therefore also where the pains are when it comes to legal documentation, and how we can technically remove those pains. In addition, we think it is important that, in principle, everyone should be able to draw up a legal document. That is why we provide understandable software, without difficult terms. We therefore follow the motto “Let Lynn do the standard work, then you can move on to the real legal challenges”. Our motivation is to innovate the legal market, to automatically generate contracts and to organise contract processes more efficiently and more securely.”
Do you have some concrete examples of your approach, and for whom you do it for?
Arnoud Engelfriet: “Let’s take a look at two common agreements in business, the nondisclosure agreement and the processor agreement. It is of course crazy that a lawyer is going to send a fat bill to his client for this, as if he / she would have to figure it out from scratch, which of course is not the case. Because how does one nondisclosure agreement differ from another? Right, there are details, because in a nutshell, they are all the same. So are the processing agreements. So why spend too much precious time on it? Why spend a lot of time looking for standard problems when we can do it much faster and just as thoroughly? Automating legal processes leads to lower legal costs and higher labor productivity.”
That sounds logical and understandable, but could you name something more concrete?
Arnoud Engelfriet: “I would like to give the example of the processing agreement. This should always be ready yesterday. When a Data Protection Officer (DPO) has to be engaged for this, it is not always possible, or it is postponed for a while. But thanks to DPALynn, we can automatically check whether such a processing agreement meets the requirements of the GDPR or not. And also, moreover, with suggestions that appear in passages that are incorrect or missing. And all this with a degree of urgency. Almost too good to be true, but through us this is done in a matter of minutes. That is why many FGs are very happy with our tool, because it saves them a lot of time.”
How does this work in terms of technology?
Arnoud Engelfriet: “Our customers can make Lynn work the way they want. Would you like to accept only Dutch law, or extend liability far and wide, whereby penalty clauses are always included as standard? No problem: the customer makes the choices and Lynn implements them every time. Accurate and legally correct. You can set up our workflow the way you want. Documents can be uploaded via the web interface, or the client can forward an emailed contract to Lynn, and you will receive the review back in minutes. An API connection is also possible. If a contract does not meet the client’s requirements, Lynn will identify it and make a counter offer.
We also use a broad database of best practice clauses, with the aim of getting the counter-proposal quickly accepted. So, don’t always impose one standard text. We really look for an adjustment that fits the current text. In technical terms, Lynn is a boosted trees machine learning ensemble. It analyses word groups for comparable meaning, and can compare clauses found in this way with a database with thousands of alternatives and variants. These have been manually screened and classified by our legal and technical experts: very strict, very long, and so on. This matching of texts is known as natural language processing as a field, which has gained a lot of attention in recent years.
And due to major innovations in hardware and software, very high reliability can be achieved in identifying texts and extracting relevant information. Errors are of course possible in computer analysis of texts. Just like a human analysis, you will think, but a computer analyses differently. At its core, our software looks for a suitable label to put on a clause. This can go wrong in two ways: the wrong label, or no label at all because there is too little information in the clause. Various measures have been built against these types of errors, such as omitting recognition with poor reliability and double-checking ‘suspicious’ labels by means of keywords. All reviewed documents are also periodically checked to make recognition even better.”