Can Legal Tech Overcome Its Own Inefficiency?

Legal tech is going to vanquish inefficiency in the legal world…….or is it? To do that legal tech needs to overcome one of the paradoxical impacts of innovation, namely that improvements in technology result in demand expansion or new input needs, which then jam up the system again with inefficiency. You then need new solutions to solve the problem.

A classic example here is the way that contracts are easier to create now compared to, for example the 1980s, with digital templating, word processing, email or online negotiation spaces, and then DMS and KM systems.

Today, most lawyers in their 60s will tell you that contracts, at least in the US and UK, are far longer and more complex today than in the past. One reason, (other than an expansion in regulation, which is itself often connected to technological advances), is simply that being able to make contracts more easily tends to make them longer.

As Neil Peretz, an experienced lawyer and founder of Contract Wrangler (which has just been bought by Conga) told Artificial Lawyer: ‘If I had to write a contract with a typewriter it would be a lot shorter than the contracts of today.’ And our chat inspired this article.

If all this technology had come to bear on a world of relatively short contracts that never grew much, the efficiency gains would have been far greater. But, that isn’t what happened. Tech both helped and made things more complicated at the same time.

Then look at eDiscovery. As electronic communications have expanded – because 1) they are easy to produce, and effortless and cheap to send, and 2) because of a proliferation of formats, from email, to Slack, to Facebook Messenger, to a dozen other channels, and 3) because of the now global adoption of computers and smart phones because they are so useful – the quantity of material that needs to be examined is gargantuan.

If eDiscovery’s capabilities had been able to work in a world frozen at early 1990s levels of material, before the surge in data really got going, then the efficiency gains would have been far greater.

Another analogy, and this time from outside of the legal world, is often called in the UK ‘the extra lane on the M25 problem’.

The M25 is the main motorway that circles London. Many people would like to see motorways like this expand to make car travel easier. However, planners who have been around the block a few times comment from experience that expanding roads next to high population centres can result in more people using their cars, rather than relying on public transport. Soon enough the newly expanded roads clog up again and you are back where you started – or at least they did before the pandemic. And then you need to build another new lane on the motorway. (Or work hard to get people back onto public transport).

In short, the expansion of capability operates in a world in flux – not a static one – and its expansion, or one might say the impact of innovation, in turn creates new and sometimes unexpected demands.

The Inefficiency Cycle

The Irony of NLP Tools

Another example is the irony of NLP tools, in that they need a lot of training, i.e. manual labour. (At least if you want very good results).

Either the vendor has to do this for you, or you do, and in most cases both will do it. So, a tool that is designed to remove inefficiency from the legal system, in fact can in many cases introduce new types of inefficiency and labour demands on the customer.

This is why some of the law firms that make a lot of use of NLP tools, for M&A reviews for example, in turn need to build dedicated teams to help with training the tool and managing its use.

The reason why these scenarios don’t get out of hand is usually because the vendor has done a lot of pre-training, i.e. the NLP software was been heavily trained on the language you are likely to be showing it. But, that work had to be done somewhere.

And in fact, one way to avoid the inefficiency cycle here is not to really use the tools that much, e.g. in the case of one vendor this site knows, some clients only use it to break up documents into specific sets (which it can do easily and without training), rather than get into the details of the clauses during a due diligence project – which needs on-the-job training

Then take legal research. The giant legal publishers have tried to defeat (at least in part) the inefficiency cycle when it comes to the use of NLP and case law research. They did this by conducting all the training up-front, so the users didn’t have to experience it.

Of course, you could argue that more legal data – and easier access to it – means more to sift through, which sucks up more time. And perhaps, even with better search methods, there is still a ‘new lane on the M25’ kind of problem. I.e. more data is accessible, and the new NLP search tools are an improvement on what we had before….but still, here we are with a massive and growing volume. Plus we still have teams of people having to soak up the inefficiency ‘behind the curtain’, so users don’t feel it.

On the NLP tools for due diligence side of things, or for the contract review and negotiation stage, that is harder and much depends on how right you want to be. At present no NLP system can continuously get 99.999% accuracy in a ‘one click’ way when it comes to long and complex documents, especially if it has not seen them before.

This is why LawGeex, for example, which pioneered the use of NLP for contract review now has a team of lawyers who make sure the responses they send back to clients are correct. (Because, guess what, the clients like things to be correct and accurate……)

So, whether the new inefficiency is handled by the vendor, or the client, someone somewhere has to take on the newly created inefficiency. And, then of course we have the M25 problem again, i.e. if it were to become really easy to do this review work, then volumes would undoubtedly increase, creating new problems for the market.

The Volume Problem

But, it’s not just NLP doc review and eDiscovery work that face the inefficiency cycle in the legal world. All legal work does.

As the wider world adopts more technology in general, the volume of legal work increases. Clients create more data and more documents, and the regulators and law makers – desperate to stay on top of things – create more rules about how that data is governed. All of this creates even more work for lawyers to handle, and that creates new inefficiency for the clients.

And that is the irony: legal tech is eating up inefficiency, but at the same time technology’s never ending march forwards is creating new challenges around the volume of output, which then needs more legal tech to help.

The end result is that legal tech is always chasing a target that it may never reach, like a greyhound chasing the ‘rabbit’ at a racecourse that it will never, ever catch.

There Is No Alternative

Now, you may read this and conclude: ‘Wow, the Luddites were right.’ Well….not really. Standing still means no innovation and none of the benefits that come with it.

We may have a busier, more complex world, overflowing with data, and too many rules and lots of tools, but it is one that also makes our lives easier in many ways.

The reality is that technological advancement is always portrayed as ‘all good’, but that unsurprisingly is rarely the case. The real world is more complex. Tech helps us, but drives up needs and usage, or creates new challenges we had not expected. Then more tech is needed to come along and deal with the problems we created.

One could say that we are stuck with the world we have and its ineluctable inefficiency cycle, but better a world of moving forwards – with all its challenges – than being stuck forever in a world you cannot change.

P.S. One last word on this relates to the idea that technology will replace lawyers. That is of course a red herring, and the inefficiency cycle is a key reason why it is false. I.e. new tech very often creates new needs, whether that is the expansion of demand, and/or additional labour needed to support the new wave of technological change, or it creates completely new needs we could not have predicted if we only operate from the perspective of a world fixed at a particular point in time. But, as noted, the world is always in flux. So, let’s embrace the change, but be realistic about what it will entail.

By Richard Tromans, Founder, Artificial Lawyer – September 2021


  1. Spot on, it’s TINA: There is no alternative. Stasis, denial and objecting to innovations because they aren’t perfect or are by nature iterative is also a decision, and a bad one. We may always be behind due to the dynamic cycle that you describe. The digital transformation in the front office continues the create the tsunami of demand, that won’t decrease; rather, accelerate. We have to arm up with some of the same tools, re-arm with new or better ones as developed or as the ones first chosen reach their limitations.

  2. Very good points. I do question whether the extra work is a necessary part of legal tech in action, or if it’s a transition cost. Getting data labeled correctly is a lot of work, as is discovering the right way of working with the new tools. But it won’t always be that way. Once your dataset is “good enough”, the amount of work drops dramatically.

    Case in point: my lawyerbot NDA Lynn now has over 14.000 NDA’s analyzed, and mistakes are rare at this stage. So I am finally done with reviewing NDA’s all the time.

    What I further see as a next step is standardization. Once you have AI that can classify dozens of clauses in one category (“security provision”, “venue/Munich,DE” and so on), the next step is to start using the category instead of just a random clause from that category. You don’t need one robot to output a three-paragraph clause specifying a 3-year term with automatic renewal (unless cancelled with one month notice) which another robot then reads using NLP and acts upon. Robot A will say “term/3years+renewal1month” and robot B will interpret that and respond with “term/2years+renewal1month” which robot A accepts according to its playbook.

1 Trackback / Pingback

  1. Smart + Digital Contracts: What Are The Use Cases? – Artificial Lawyer

Comments are closed.