I have this ongoing, love-hate relationship with the portrayal of artificial intelligence for the law in the media. Almost without exception the articles are accompanied by a futuristic image, like a robot carrying a briefcase saying “Is this going to be the lawyer of the future?”. This leaves people either with a dystopian, scared vision of the future, or an overexcited sense of where the world is going. And it completely falls short of the reality.

The reality of AI in the legal industry is not going to be some sort of sci-fi looking robot. It’s going to be happening as part of the work our teams are doing each day and embedded in the software and hardware we use to do that work. Here’s a summary of some early observations about AI that I presented during Legal Innovation & Tech Fest 2017 that are based on the experiences we have had in experimenting with and implementing elements of AI at Gilbert + Tobin.

1. AI Is Cheap to Get Started but Expensive to Keep Going

When it comes to innovation in the legal industry, a major plus is that cloud-based technology and SaaS systems make it cheaper to get started with AI. It’s not like the old days of having to invest in expensive servers and long-term software licences before even getting to experiment – you can now purchase a subscription and get started right away.

Law firms (and probably in-house teams too) should at a minimum have accounts with the major machine learning document review players at their entry level plans to start experimenting with AI.  Reading about it or even getting a demo is not enough – you need to put a few documents in and play with it. I’ve learned a lot from starting to do the training on a set of documents myself, to see what happens. You don’t need a big investment or a business case to do this much.

But beware. The total cost of ownership of these experiments goes beyond just the software, and often once you start to scale it up, you find that it costs quite a bit more than you originally thought. The catch is that it’s the volume of documents that makes machine learning more accurate. If you begin to avoid putting volume through to avoid a per document or per gigabyte fee, you may hold back the benefits and ultimately the business case.

Eventually, you will want to get to the point where you’ve bought some sort of enterprise license on an all you can eat plan – so that there’s no disincentive to build the volume of learning. But that’s much more of a substantial investment for a lot of places, particularly smaller law firms, and will generally require a business case.

So getting started with AI is not that hard, and certainly not cost prohibitive, but making it a part of your core business requires a more substantial investment.

2. You Need to be Willing to Experiment

The logic of technology and everything you read about it says “Don’t just throw technology around like a hammer looking for a nail; find a business problem to solve and implement your solution to make that happen”. That is the logic, and our instinct says that’s the way we need to go.

On the other hand, it’s difficult at this stage to know what’s going to work in the AI space and where the value is going to be. So the other option you have is to experiment and iterate. But you often end up in the conundrum of not being able to make a business case because you can’t necessarily define the problem about where you’re going to save money, yet you know there’s something in there to be solved. So you have to be prepared to take a bit of time and spend some money to “fail fast”, as we heard a lot about during Legal Innovation & Tech Fest, to experiment with AI.

3. AI Training Requires Skills and Time

Machine learning, particularly in the document review area, generally requires a form of supervised learning –  which needs humans, and it needs humans who know a fair bit about what they’re doing. Sure, you could use some lower cost resources, but if you want to train a machine well, it needs to have the right answers.

Taking people offline, particularly in a busy law firm, is expensive; it has not only actual cost but opportunity cost associated with it. To run a matter in a traditional way and also run a parallel matter using AI processes is 100% the right way to do it, but it’s also really expensive to double up on the costs.

That is not to say it’s prohibitive, because the learning gets better each time – the first few times are going to take some real effort from people, but then it will start to gain momentum. So when you’re thinking about the cost of getting started with AI, take into account what it would cost to pull some people off certain projects to focus on AI.

4. Don’t Underestimate Change Management in AI

The way in which AI is manifesting in the legal sector is taking part of an existing process and finding ways to do it differently. So it’s essential that the people who are involved in that end-to-end process understand how the AI piece fits in, how it will make a difference, and also how the rest of the process still needs to be performed.

People have quite a strong muscle memory with ingrained ways of working, and changing those behaviours can not only be difficult, it can also cause a productivity fall-back. A piece of technology will make the process much easier the third, fourth, or fifth time we do it, but the first and second time is hard. It’s like changing from using a PC to a Mac – your hand doesn’t quite work for the first day or two because everything is different, but then you get used to it. We need to help our people through that difficult process and accept that there will be an initial decline in productivity, which is not always easy in a client service environment.

It’s important to factor this in to your project planning and resourcing and most importantly into your lawyers’ expectations of results.

5. The Hard Part of AI Is Not the Technology

So much of the hard work in getting started in AI isn’t about the technology; the struggle is how you integrate it into your broader processes. From our firm’s part, we’ve rented technology, we’ve built some, and we’re working with partners on some of it. And while the technologists’ role is incredibly important, it’s actually the people who are supporting the implementation of our AI initiatives – our process and knowledge teams and the lawyers themselves – that are helping us reap the actual benefits.

These 5 things include some reality checks on getting value from AI. I believe these underscore how important it is to be moving quickly and looking to address these challenges.

About the Author

Sam Nickless is the COO and a Partner of Gilbert + Tobin, which he joined in 2015. He qualified as a lawyer but has never practised – having been a partner at McKinsey & Company, and then held executive roles at NAB, Aristocrat and GPT Group. Sam leads G+T’s operational teams as well as driving the firm’s strategy and innovation agenda. Sam is also on the Board of tech-driven legal startup.