top of page

Legal tech: Beyond the myths #4 - Are robots taking over our jobs?

By Arnoud Engelfriet

Did you notice? Whenever you read an article on robotization or artificial intelligence in law, there’s always a stock picture of a robot wearing a gown or wig, or tapping a judge’s gavel. An easy way to illustrate “robots doing legal work” of course. But it shows an underlying assumption, namely that robotization or AI in law means that a robot takes over the work we human lawyers do. That’s not what’s going to happen, but the idea is stubborn. Until we manage to take it away, however, robots aren’t going to be much of help.

Mere automation

The main promise of legal tech is that technology will help lawyers improve their work. This of course isn’t a new promise. For decades legal providers have put efforts in automating aspects of their work. Timekeeping, file management and formatting documents, to name a few. This provided a huge efficiency gain, which allowed lawyers to focus more on substantive work.

Automation did not, however, fundamentally change the way of working of the legal professional. It merely meant the same work could be done faster, with fewer distractions caused by limitations of the tools. Writing with a word processor is faster than a typewriter because you can eliminate mistakes more easily. But it does not make the actual thought process faster: the contract won’t be written any easier, and a legal argument won’t sound more convincing because of a better word processor.

Work transformation

A second stage in legal automation is to automate parts of the substantive work itself. This goes a step beyond automating the way of working. Let’s take an example from legal practice: writing legal opinions. Automating the tools would mean using a word processor, and searching a case law database with keywords and filters, then typing in (or maybe copy-pasting) the relevant passages and the case citation. One could imagine a tool inside the word processor, inserting citations with passages with one click.

But still, choosing which cases (and which passages) is truly the choice of the human lawyer. A tool like IBM’s ROSS is a good example that takes this one step further. Given a legal case, the tool retrieves relevant citations and provides the skeleton of a legal argument. This does more than just saving time: it takes drudge work out of the lawyers’ hands.

In contract drafting practice, a common example is the document assembly tool that prompts the user with questions: who are the parties, what is the price, should confidentiality be one-sided or mutual, how long should the term be, and so on. Based on the answer, the tool selects relevant clauses and builds a contract. Clever and useful as this may be, in the end this is still a mere automation of the old model agreement that said “insert contract term here (in years)”.

Substantive transformation

The third stage in legal automation, which is what’s usually referred to as legal tech today, is the transformation of how substantive work is performed. True transformation means doing things differently because technology allows you to.

For instance, in the contract assembly example, a true transformation would be to come up with a different template based on the client or the type of deal, or more generally to go beyond the standard clauses from the model for answers A through D. Or choosing to omit certain articles in their entirety because the client’s history reveals they are not useful. Or re-drafting a limitation of liability after reviewing the other party’s professional liability insurance. This is different from merely answering questions and always getting clause 23.A or 23.B for a force majeure statement.

In contract review, an example of automation would be a tool that recognizes legal clauses and flags them for the lawyer to review. This helps the lawyer do his or her work better, but does not change it. A transformation would be a tool that not only recognizes clauses, but also determines their impact and makes a decision whether the lawyer should even see them. If the contract is low-value and the clause is only a small deviation from the company’s policy, why bother?

(If you want an example of such transformation that’s closer to home, consider a legal liability insurer. Often, when a small claim is received, the insurer may make the decision to pay the insured out of its own pockets and forego the traditional route to court to recover damages from the party that caused it. This is of course based on financial considerations, but still it is very different from the traditional approach where you recover damages from the party that caused them.)

The challenge, of course, is getting such a transformation actually adopted. And this is hard. Change is always hard, but for some reasons especially in the legal sector. This despite the fact that this sector deals with new stuff all the time.

Four hundred years of experience

The legal sector has an image of little change. For a large part, this is only logical: the work is fundamentally the same as, say, 400 years ago. Back then as now there were conflicts that needed legal arguments to be settled and agreements to be put on paper. While the subjects and the law may have been different, the principle remained the same.

At the same time, this is strange. Most lawyers are very much open to new developments, from the latest gadgets to large-scale developments in society. And again this is only logical, as keeping up with the world is needed to do legal work. If a lawyer can’t operate a fax machine, how can he provide legal advice on the status of fax messages, to name just one example. Novelties are part of the work. So why did the legal work remain the same for so long?

Many explanations have been proposed. The hourly billing system in particular supposedly blocked innovation: those who work more efficiently, could not claim as many billable hours. Add to this a system where the partners at the top of a law firm receive a percentage for each billable hour, and the result is a very strong stimulus against reducing the number of hours worked. But surely this is not the entire explanation: those who work more efficiently may make fewer hours on one job, but would have time available

to work for other clients.

In my opinion the fundamental reason is the well-known expression “Don’t change a winning team”. Firms that do well, are very busy. Introducing fundamental change in a busy environment is not going to go over well. This takes time and concentration, and both are in strong demand. So the change will have to wait until next year.

But then why now? Customers are asking more vocally for change, for speed and cost savings. And there’s nothing left in terms of small savings, lowering the hourly rate or distributing costs. In addition, more and more firms see options to introduce legal services in innovative ways, threatening the traditional monopoly position law firms have enjoyed. This forces thinking about transformation.

The role of AI in transformation

The rise of artificial intelligence may provide the key to this transformation. Most types of transformation that are currently under consideration, rely heavily on automatically reviewing texts and spotting anomalies. This is is something computers are very good at, and this happens to be an activity often requested from lawyers. But we can do more than just put an AI as a first check to save lawyers a little time.

Artificial intelligence is in particular good in recognizing patterns in huge amounts of data, turning it into actionable intelligence: this is a force majeure clause, this line of reasoning reflects the three prongs from the Sunday Times case, and so on. Such intelligence is useful as direct input to the professional taking the next step – but can also be leveraged to change the next step, change the workflow or process in which the analysis occurs.

As a simple example, consider a company that often receives confidentiality agreements from prospective customers and partners. The traditional process would be to send this to Legal for review, the lawyer would enter into discussions with the other side, and after discussions had led to a mutually acceptable agreement both parties would sign. Applying legal tech would speed up this process: instead of the lawyer reviewing the document, an AI would do the same. The lawyer would review the output, including a marked-up document and open the negotiations.

When transforming this process, one needs to do a step back: what is the intent of this process? Who are the actors that need to operate the process? Even though this is a legal document, the process is by itself not legal: it serves to enable the business to talk with prospects in confidence. There are more ways to address this purpose without having a manual review of each incoming NDA. One simple example: insist on the company’s own NDA, or a well-accepted standard NDA (such as the oneNDA initiative).

More advanced process changes involve putting the AI earlier in the process. For instance, the businessperson sends the NDA to the AI, who reviews and either approves or rejects it, where rejection means “not salvageable, use our own NDA instead”. This would cover two-thirds of the situations, according to statistics of our own product NDA Lynn. In the other one third, negotiations can be started based on a redline prepared by the AI tool. And even here, the lawyer is not (yet) necessary: the redline can be sent back for initial comment, once those are reviewed by the other party, the lawyer can resolve the feedback and negotiate towards mutual agreement.

Managing workflow

Enabling such changes first of all requires a clear workflow. Businesspeople should know to involve the AI, and when to approach the other side with a redline. Technology can help: the review tool can take care of most of the administrative burden, and keep track of the latest actions taken and the next steps to be performed by humans. However, this only works if the humans involved are in agreement on what the next steps are supposed to be. Establishing assent on such matters is very hard, as it may involve changing a company’s culture or even overcoming corporate infighting.

A related matter is to know what the AI tool is to approve or reject. In general terms, most companies have an understanding of what is acceptable in their line of business. For instance, software companies typically want to retain their IP and are concerned with employee poaching, while food suppliers worry more about quality provisions, returns and the confidentiality of price discounts, to name two examples. But zooming in a little more often reveals blind spots: what do we mean with “we typically want to retain IP”? What amount of returns do we accept, exactly? And are there combinations: is a small price discount with a high return right for the customer acceptable just like a high discount with no right of return?

First steps

Successfully deploying legal tech thus, as a first step, forces a company to reflect on its workflows involving the legal department and the underlying assumptions about the business. Getting this on the table in an actionable format is a huge challenge, but at the same time represents 80% of what is needed for a true transformation of such workflows. The tool can then quickly be deployed to fit the new process. As always with technology, it’s not about the tool but about the humans using it.


About the Author

Arnoud Engelfriet is co-founder of the legal tech company JuriBlox, and creator of its AI contract review tool Lynn Legal.

Arnoud has been working as an IT lawyer since 1993. After a career at Royal Philips as IP counsel, he became partner at ICTRecht Legal Services, which has grown from a two-man firm in 2008 to a 80+ person legal consultancy firm.

bottom of page