Automation Initiatives in Legal Will Require Human Element due to Domain Complexity and Poor Process Discipline
By 2025, 30% of new legal technology automation solutions will combine software with staffing for a “human-in-the-loop” offering, according to Gartner, Inc. Despite growing demand for greater automation and increasing sophistication of tech innovations, machine learning solutions for corporate legal teams will struggle with the high levels of domain expertise required and high rates of exceptions expected.
“There’s little question that growth in legal work is outstripping growth in legal headcount, and on the surface, this can make advanced automation solutions appear very attractive to legal leaders,” said Zack Hutto, director, advisory in the Gartner Legal & Compliance practice. The question is whether the legal department has the capabilities required to customize and configure advanced machine learning systems, which will be needed to enable them to handle the unique scenarios and frequent exceptions that come about in this type of work.
Push Towards Automation
The legal department has lagged many other business units in automation, often deliberately so, but leaders’ attitudes towards automation in legal have softened since the pandemic brought about twin forces of a sharp increase in workload (and with it exhausted lawyers), and a reluctance from CFOs to keep adding headcount.
Vendors, heartened by notable successes with automation in other business functions, are keen to promote their capabilities and open new markets. However, solution providers face headwinds in the complexity of legal workflows, differing risk tolerances across organizations, and inconsistent processes that not only weaken returns but also hamper the capture of information needed for training machine learning solutions.
“Legal departments should not avoid automation,” said Hutto. But the right foundations must be in place. Automation – especially sophisticated AI-driven techniques – should not be seen as a quick fix to old problems.
Solving automation challenges will require greater process discipline in legal teams so that legal data is consistent and comprehensible to machine learning systems. It also requires a careful blend of technical expertise and legal knowledge that can configure and train machine learning solutions in the specific context of an organization.
This kind of in-demand machine learning expertise, coupled with legal understanding, is going to be hard and expensive to find. In reality, hiring for this capability is not going to be very scalable for most corporate legal departments. “There’s also so much complexity to handle in legal work that it seems unlikely there will be any broad-based, effective ‘off the shelf solutions available within three years,” said Hutto.
Solutions attempting to automate legal work right now tend to demonstrate quite high error and exception rates compare to other business functions. Part of the problem is that data assets between different users are quite distinct. A machine learning solution trained at one company will likely be useless when applied to another company.
A hybrid, ‘human-in-the-loop’ model, blending staffing and software, will win out, with the required domain expertise coming from the supply side rather than from within legal departments themselves. An inflection point of productivity in legal automation will come when legal departments have machine learning experts who can truly understand the complexity of the legal problems within the context of their organizations.
This shift represents a marked difference from current market offerings which may provide low- or no-code automation solutions but place the onus on end-users to build upon their platforms. This is a challenging prospect for in-house legal departments given their existing skillsets and already stretched resources. However, increased demand from corporate legal teams is expected to continue, alongside significant acquisitions and venture capital investment in legal tech markets.