Affording GenAI: High Costs, Changing Workflows, and Client Impact
November 5, 2024
I recently had the opportunity to attend the annual Relativity Fest conference held in Chicago, and the star of the conference, as with many legal technology conferences, was Generative AI (GenAI). Specifically, Relativity was excited to share news of newly developed tools that allow users to use GenAI for document review, case strategy, and privilege checks.
While it is not difficult to imagine how GenAI technologies like Relativity’s have huge potential to assist lawyers with completing their tasks faster and more accurately than ever before, conversations about the potential of these technologies would not be complete without also discussing the massive costs.
High Costs
While most, if not all, legal GenAI products like Relativity’s new aiR for Review do not publicly disclose the prices of their products, the prices are incredibly high. Relativity’s aiR for Review has been tested by only approximately 25 users prior to release – likely because of the extremely high cost of the product.
The high cost of tools like aiR for Review reflect costs of both development and use. GenAI technologies are expensive to train – the process involves labor costs, information gathering, content filtering, and extensive testing. Companies that train on licensed content, as legal GenAI often does, spend even more on development. Once developed, these technologies require costly processing and power in order to run, as well as large amounts of fresh water to cool processors and generators. By one calculation, by 2027 the global AI sector could consume about the same amount of annual energy as the entire country of the Netherlands.
The immense cost of GenAI tools will have several implications for the practice of law, as well as how we teach new attorneys. Since the cost of legal GenAI tools is often very, very high, only limited firms with the means and willingness to spend very large sums will be able to use these tools. Even for firms that have the means to purchase tools like Relativity’s aiR for Review, these technologies are still very new, and there will likely be hesitation from most firms to invest so heavily in a relatively unproven product.
Changing Workflows
The counter-argument to the huge cost of GenAI is that lawyers will be able to work much more efficiently and accurately – thus saving the firm money in the long run. While results from initial tests of products such as aiR for Review do have promising results in time, efficiency, and accuracy, there are still considerable time expenditures from senior attorneys to use the product as envisioned by the developers.
Namely, the need for subject matter experts, or senior attorneys who are well-versed in both the case at hand and the relevant law surrounding the issue, was stressed during the introduction of Relativity’s GenAI tools. It is essential that a subject matter expert be involved in the prompt criteria writing process, which asks users to define things like key terms and players. During a traditional non-AI document review, a senior attorney would likely also be involved in defining what a relevant document may be. However, it is essential that these senior attorneys continue to hone the criteria writing process when using GenAI tools, like reviewing results on small sets of documents and editing their criteria as appropriate. While products like aiR for Review promise quick and accurate results with sufficient prompt criteria, the billable hours spent by senior attorneys writing these criteria are likely to remain the same, if not greater, than the time spent writing guidelines for traditional document review.
While the senior attorney’s job may not change dramatically with the introduction of GenAI products such as aiR for Review, the same cannot be said for the traditional document reviewers. These technologies open up the possibility that human first-pass reviewers may be a thing of the past. First year associates may in the future spend far less time on first-pass reviews, leaving them with the ability to spend their time tackling other tasks – potentially other tasks that require higher-level legal thinking.
Client Impact
Changing traditional case workflows raises questions about what impact may be found in client bills. A judge in New York rejected an award amount calculated using GenAI – criticizing the original GenAI-made calculation as an “aggressive fee” that was “utterly and unusually unpersuasive.”
In addition to issues with GenAI-created bills, there are concerns that the use of GenAI may lead to overbilling. The California State Bar included billing in their Practical Guidance For the Use of Generative Artificial Intelligence in the Practice of Law and the Florida Bar issued an Ethics Opinion on the use of GenAI that includes questions of billing. Both resources warn attorneys that they must not falsely inflate hours or charge clients for time saved due to the use of GenAI.
Considering the excessive cost of GenAI tools, paired with limitations on overbilling, there remains a question as to how the cost of GenAI tools may be passed on to clients. It is possible that we may see a growing use of flat-fee agreements, allowing attorneys to build the cost of GenAI tools into their agreements, and allowing attorneys to utilize GenAI without worrying about overbilling for time saved. Additionally, some legal GenAI products are offered “a la carte”, meaning that attorneys may look to pass on the cost of running GenAI searches or GenAI review by charging clients per task – considering it a cost of representation.
Finally, the cost of legal GenAI products such as those offered by Relativity raise concerns about how pro se litigants, litigants helped by legal aid organizations or nonprofits, or even litigants represented by smaller firms may be placed at a (further) disadvantage. It is likely that these groups will be unable to afford legal GenAI tools any time soon, further expanding the justice gap.