What Are the Costs of Implementation?
Software Costs
There are a variety of costs associated with harnessing AI within a law firm, some of which are obvious and others that are inconspicuous enough as to rarely factor into the cost-benefit analysis. According to Anwar Khan, Senior Staff Software Engineer, Enterprise Architecture and AI at John Deere, although licensing fees are the one element of AI implementation that firms always budget for, they usually comprise the smallest cost, with three more notable costs arising later.
It can be incredibly difficult to determine an exact figure for how much your firm is spending on AI.
The first of these relates to the amount of proprietary data that law firms feed into the AI platform. He explains that "many enterprise AI platforms bill against how much material you have loaded into them rather than how many questions anyone asks." As such, a law firm can unnecessarily "raise its own spend considerably by onboarding an archive nobody queries." Second, he warns that the cost and methodology of pricing AI tools can fluctuate significantly when it comes to renewal. Third, and perhaps most important, is the verification burden.
Verification Burden
There have been countless examples reported by the media of "digital hallucinations" (erroneous AI outputs) being
included by law firms in court filings. One
website set up to track fake citations and similar AI errors that make it to court has (at the time of writing) almost reached a count of 2,000 worldwide. As well as potentially impacting litigation success, this kind of exposure inevitably leads to reputational damage for firms. AI hallucinations are increasingly resulting in fines issued by courts, with one federal judge in an
Oregon court imposing $110,000 in fines and attorney fees against two lawyers who filed documents that contained multiple AI fabrications.
In the UK, the Solicitors Regulation Authority has recently issued a warning about this problem and states that lawyers are ultimately responsible for any AI output they use in court filings or any other documentation. This need for legal professionals to check the output of any AI tool they use is known as the "verification burden" which, Khan says, forms "part of the running cost of the tool." Some lawyers argue that the verification burden alone removes any efficiency gains and can even increase the amount of time spent on certain tasks. Brian Inkster, of Scottish firm Inksters Solicitors, says he "would have concerns about the time/cost of the verification burden which, in certain cases, may well outweigh any investment made." He also notes that there is an additional expenditure of training staff to use AI tools properly.
Staff Training and Data Protection
Law firms' learning and development departments have never been so busy, with fast-moving AI products posing a significant drain on resources that need to be directed at helping lawyers and support staff stay up to date with the changes and keep on top of the tools. But along with understanding how to use the various AI tools from a technical standpoint, it's also vital that legal professionals are fully trained in the various responsibilities that come alongside AI use.
This need for legal professionals to check the output of any AI tool they use is known as the "verification burden" which, Khan says, forms "part of the running cost of the tool."
One of the biggest dangers of law firms implementing AI without providing sufficient training relates to data protection. It's crucial that firms ensure that any AI products used by their lawyers do not collect confidential information — particularly pertaining to clients — and later use this data for training new AI models. Ashish Raivadera, Managing Director at Major, Lindsey & Africa, says that, although there are significant sums being spent on buying or developing any particular AI solution, one of the bigger costs is "training staff, both [administrators] and fee-earners, how to use it. Training is a major part of overall implementation costs, but a necessary step to ensure proper integration.”
What Are the Costs of Not Implementing AI?
Although there are clearly a number of different costs associated with adopting AI, firms may also find themselves at a disadvantage if they don’t pursue at least some form of implementation. If the aforementioned efficiency gains exceed costs to achieve a favorable cost-benefit analysis, failure to harness AI could lead to a firm missing out on an improved balance sheet. Perhaps more critically, it could lose out to competitors who are passing those savings onto their clients, with existing clients even demanding reductions in legal fees that are commensurate to what competitors are offering.
"Not using AI effectively will create efficiency challenges that affect price and delivery time for firms, making them less competitive and less favored as a result,” Raivadera says. “Clients are becoming more price sensitive, so one way to deliver a cheaper option is AI. Proper implementation of AI can allow firms to achieve greater profitability without always having to raise prices.”
Another risk of not adopting AI within the workflow is that lawyers and other staff will nevertheless use their own personal AI tools to complete various workplace tasks, a phenomenon known as "shadow AI." The unregulated nature of such an approach can lead to various risks going completely unmanaged and opening the firm up to data protection issues. Khan argues that a firm that "declines to provide a sanctioned AI tool, with training and an audit trail, does not get zero AI. It gets unmanaged AI, on personal accounts, on client matters, invisible to the firm."
It's crucial that firms ensure that any AI products used by their lawyers do not collect confidential information — particularly pertaining to clients — and later use this data for training new AI models.
Former US Secretary of Defense Donald Rumsfeld
famously once said that there are "known knowns, known unknowns and unknown unknowns." Although he was referring to US security threats, the same reasoning can be applied to the future of legal AI. For example, one "known known" is that legal AI platforms are moving from "per-seat" pricing to consumption-based pricing.
Legora has already announced this change and others are likely to follow suit, which will no doubt sunset schemes that encourage
tokenmaxxing, linking bonuses to volumes of AI prompting. Khan says that consumption-based pricing is "fairer" but also "much harder to budget."
A "known unknown" is just how much prices could rise and whether cost structures could change again soon. Inkster warns that "law firms will need to be careful not to get tied into products where the future cost is uncertain."
Although Raivadera does not forsee any dramatic cost reductions, he does "see firms achieving better efficiencies that allow them to justify the continued use of these AI tools," arguing that the "bigger issue at the moment seems to be promoting the actual uptake of AI tools to genuinely achieve these efficiency gains."
Then we have "unknown unknowns." One legal AI analogy to the original defense-centric nature of the aphorism is the view that AI model development is akin to an arms race. Will the most successful law firms of the future be
AI native legal practices, which battle it out using the most expensive cutting-edge AI tools? Inkster is doubtful, saying that, although "law firms should be looking at the best technology to create efficiencies and deliver results for their clients," the concept of an AI native law firm is a "gimmick" rather than the future of the legal industry. Yet another possibility beckons, if we listen to some
AI experts who predict that artificial general intelligence (AGI) could be achieved within the next decade, at which point AI will essentially be able to match and quickly surpass the ability of humans to do all work that relies on the brain.
There is a smorgasbord of costs associated with implementing AI in law firms, some of which are volatile and prone to sudden increases, while others are hidden or at least not immediately obvious. However, there is also a plethora of potential costs that come with the choice to not harness AI tools and the benefits that come with them. Legal administrators will need to think carefully about the cost-benefit analysis of adopting AI for their firms and how best to navigate the uncertain and fast-changing future.
