When you have been in eDiscovery as long as we have, you expect to see, year on year, improvements and new features and the industry rarely disappoints. When I think back many years ago to being excited seeing a solution “tally” documents by custodian and I look at what has developed since it is really quite astonishing. As I say, we expect, and corporate clients expect, constant changes which react to the ever growing volumes and increasing complexity of electronic data, all aimed at saving time and driving down cost. However, just once in a while, a new feature occurs which is truly game changing, and I cannot remember being quite as excited as I am now to see Relativity announce the launch of aiR for Review, at their recent Relativity Fest.
What is it? What does it mean? Why am I so excited?
I have said many times that in the world of eDiscovery, AI is not new as we have had AI in the software for more than decade. Yet, in the last year or more all we read about is AI, AI and more AI and of course more recently, generative AI such as Chat GPT-4. We have seen numerous posts, examples, some train wrecks in cases in the USA, and yes, in South Africa, as a result of the misuse of generative AI. I wrote a few months ago that we should all calm down and take stock and I made the point that the AI developments within eDiscovery software providers such as Relativity over the years, had been properly researched and tested by scientists, technologists and lawyers before being released, thereby ensuring, the magic word defensibility. aiR for Review is Relativity’s first GPT-4 enabled product. Founded on Microsoft Azure OpenAI (and that in itself pleases me as MS and Relativity have a deep longstanding global relationship and MS Azure hosts RelativityOne globally), aiR for Review will analyse documents and highlight the most important (relevant, privileged etc.) information to aid human reviewers. It goes without saying that it will provide faster, higher quality, more reliable results than manual review workflows and although we know nothing yet about pricing, I am sure it will be markedly less expensive than manual reviews. It also means that lawyers will be able to get into the real facts and issues in their cases faster than they ever have before.
As I mentioned we have seen bad instances where generative AI has been used in cases by lawyers and incorrect or false citations etc. have resulted, and I have cautioned about checks and balances. Relativity aiR for Review will explain its recommendations and rationale allowing for human validation of its findings. This is why I am excited because I trust Relativity to do it right. They do not react to all of the hype and competitor pressure. They take their time on research and development and they use their clients and partners within the Relativity community to beta test again and again. I trust their 6 declared AI principles
1. We build AI with purpose that delivers value to our customers
2. We empower our customers with clarity and control
3. We ensure fairness is front and centre in our AI developments
4. We champion privacy throughout the AI product development lifecycle
5. We place the security of our customers’ data at the heart of everything that we do
6. We act with a high standard of accountability
Here in SA I have seen cases with months’ worth of T1 review costing millions of Rand and still human errors all of which cost more time and money to rectify. I already know of some SA corporations who will cry out for aiR for Review as the cost of human review is becoming disproportionate and unsustainable. I don’t want to consider the negativity of review lawyers saying that jobs will be at risk, because technology brings opportunities as well as casualties. Humans will still be required to do the checks and balances and therefore there will be different types of lawyers. This is exactly what Prof Richard Susskind has been preaching for two decades or as Senior Counsel friend of mine here in SA says, “Adapt or die”. Nor do I want to consider lawyers saying that this will take money out of their pockets as this will take them to the next level and ensure more successful work.
Waseema and I, at Harrisons eDiscovery Consuting, cannot wait, as RelOne Partners in SA to see aiR for Review and in practical terms I understand there will be limited general availability by the end of 2023 with full roll out in Q1/Q2 2024. In the intervening period further testing will continue and one USA law firm and technology team involved in the testing has already said that at the moment the only limit seems to be their own imagination.
I am still reeling from the potential of aiR for Review because it is not just a new feature it is a game changer in handling cases, that I cannot channel my brain to also comment on Relativity’s two other new launches of AI based initiatives PI Detect and Data Breach Response. Let me take a breath and I will write soon.