Your single source for new lessons on legal technology, e-discovery, and the people innovating behind the scenes.

AI Visionary Matthew Golab on the Leap from Hard Copies, to Predictive Coding, to Natural Language Processing

Ryan Docherty

In his initial interview, Matthew Golab—director of legal informatics and R&D at Gilbert + Tobin—said that “AI systems don’t sleep, or forget, or make mistakes.” His firm knows that the strategic use of AI frees up time and energy that will enable lawyers to work more efficiently and more thoughtfully for their clients. And his team is focused on finding, testing, and implementing evermore ways of making that benefit a reality for their collagues.

Ryan Docherty: Please describe your role in your organization and how technology plays a part in it.

Matthew Golab: I joined Gilbert + Tobin in 2011 with a mandate to establish a litigation support function. Eleven years later, we have a Legal Informatics team of 20+ and utilise our deep technical skills to assist clients in disputes and investigations, competition and regulatory matters, Royal Commissions and Inquiries. In addition to e-discovery, we also specialise in digital forensics, cartography, data analytics, and M&A data rooms. I also have an R&D role where we consider a variety of hypothesis and run pilots using a range of technologies.

What were your interests early on and what drew you to your line of work?

I entered this field in about 2000, and have had the privilege of watching a very labour-intensive, hardcopy process morph into an almost wholly digital process.

Each client and each matter is different; however, a key advantage that we have in the digital realm that we didn’t have in hardcopy is the ability to scale using technology. Hardcopy was a very linear process with little shortcuts, whereas in digital there are many ways in which you are able to use the analytical tools and techniques available to us to significantly reduce the effort required in a large-scale document review.

I have an aptitude for the abstract concepts in using technology and data structures and I enjoy dreaming up novel ways of using technology to assist in solving problems. I very much enjoy problem solving and using a broad range of techniques and technologies to solve problems.

Any advice for those who are interested in following your career path?

Love what you do; focus on following challenging work; have a lifelong thirst for knowledge and learning.

Stay humble; listen first rather than speak first.

Exercise and eat healthy food.

Understand that your role is an ancillary function to a complex legal process.

Strive for excellence in everything you do.

Be open with others; share knowledge and ask questions.

Ask for help. It is all too easy to become overloaded with work pressures and stress and to eventually become depressed.

When you’re not working, how do you like to decompress?

I’m a lifelong reader, so my Kindle is always at my side. I enjoy listening to music, as well as endurance and cross country mountain bike riding. I regularly walk around Sydney as a way to reset—both on the way to and from work, and also during the lunch break.

When you think about your business or your industry in five or ten years, how will AI have affected it?

Machine learning and document clustering have been in our industry for at least the last 10 years, if not longer, and we have very established workflows and techniques for us to use these tools in our day-to-day work. I am excited about the potential for us to scale our efforts for larger tasks, and achieve significant efficiencies for our clients in the time it takes for us to assist them and the cost of our assistance.

I’m particularly interested in the ability to build models to predict risk for clients. For instance, they could be used to predict legal privilege, and we could have these models active in real time so that they are able to give a risk prediction immediately after ingestion of client data (as opposed to our current scenario, where the model is typically trained from scratch for each new matter).

Another future area is for AI to be able to comprehend language, e.g., natural language understanding (NLU).

How can AI enthusiasts pursue professional development in this area?

Many scientific papers on the subject are freely available, as are blogs, social media feeds, and podcasts from AI luminaries such as Yann LeCun; Geoffrey Hinton; Yoshua Bengio; Andrew Ng; Ian Goodfellow; Chris Manning; Andrej Karpathy; and so on. There are also numerous free online courses available, as well as an amazing range of free Python libraries.

It is never too late to start learning.

What AI-powered innovations are you looking forward to implementing in the future?

With the recent boom in generative AI, it is exciting to see the rapid progress in natural language. A very significant amount of content within law is text based, and there are many ways that generative AI could augment legal work: drafting contracts or advice, reviewing contracts, and, of course, reviewing documents during e-discovery. It is exciting to think about heavily customised AI models that are trained on your corpus, or even models trained for particular clients.

Our focus at Gilbert + Tobin is to look for best-in-class AI and work out how to integrate these into standard work practices and augment the efforts of our lawyers, so that we can continue to improve our efficiency and automate (or at least reduce the effort required by) low-level administrative tasks. By doing this, we can focus on high-level strategic work for clients, as well as be more efficient at these tasks so that we achieve cost savings for our clients.

Are you seeing an AI skills gap in your industry or organization, and if so, what will help to close it? 

Yes, there are significant skill gaps as well as capacity gaps in the APAC region as the demand for skilled workers exceeds supply. Our market is also in transition as we move from work practices that have generally been unchanged since the mid-2000s, and move toward practices that are heavily dependent on the use of predictive coding and machine learning.

There is also a broad range of different skills in our market, which broadly fall into either a project management role or a technical role. Not everyone who is experienced in these roles has the necessary aptitude to master AI tools due to the abstract concepts. However, as the tools and underlying systems that we use become more sophisticated, they will also become easier to use.

As to what will help close the skills gap: partly it is an adjustment in recruiting priorities to look for a proportion of new employees who have a data science or computer science background, and partly it is up to each employer to place emphasis on educating and promoting the use of AI.

What is one small thing someone could do today to move toward an AI-enabled future?

Understand your data, curate your data, have high quality metadata, and consider how you can build high quality data stores (and regularly assess and update these data stores).

Consider the current challenges or inefficiencies in your business, then undertake process mapping and interview experts to understand more about these hurdles. Although it is very exciting to look at AI as a solution to everything, it is a lot better in the long run to carefully consider problems and then their particular solutions, rather than trying to have a solution first and then find problems that fit the solution.

Look for like-minded AI enthusiasts and ways in which you can share knowledge and “geek out,” as well as share tips. This shouldn’t necessarily be limited to your business unit or even your organisation, though obviously you do have to be careful to avoid sharing actual business data or your business strategy or business secrets with peers outside of your company.

To move toward an AI-enabled future, ensure you understand your data, curate your data, have high quality metadata, and consider how you can build high quality data stores (and regularly assess and update these data stores).

What is unique about adopting and implementing AI in your region?

In 2023, I would say that at Gilbert + Tobin we use predictive coding in about 75 to 80 percent of matters and it is not something that we need to continue to promote or encourage lawyers to use. Instead, it is widely acknowledged within the firm that we use a range of technologies and workflows to augment and enhance the work of lawyers, and to ideally reduce the effort required in large-scale document review. We also use these tools to ensure consistency of review, and as risk detectors—such as custom models that have been developed for privilege or high-risk documents such as board papers.

We’ve seen a lot of unique innovations coming out of the Australian legal community in recent years. How has your team participated in these efforts?

We have been using AI within Gilbert + Tobin for the past 10 years. We use it in several ways, such as:

  • e-Discovery and large scale document review
  • Due diligence for M&A matters (e.g., contract review)
  • Custom, one-off data analytics and data tasks for clients
  • Data analytics and data tasks within the firm to assist in our operations
  • Building internal apps and tools to automate tasks for lawyers, using natural language processing
  • R&D sercies to regularly test out hypothesis and prototype things
  • Building privilege and risk detection models for large and repeat financial services clients
  • Running pilots with AI and legal tech startups

We have a high level of awareness of AI’s potential within the firm, and a willingness to embrace and explore how AI could assist us in solving large-scale problems.

Ryan Docherty is a senior account executive at Relativity.