Artificial Intelligence and Delivery of Unbundled Legal Services to Corporate and Under-Represented Parties

Charlie von Simson
March 18, 2019
AI and The Law

Unbundling is a model for the delivery of legal services in which the lawyer and client agree, generally at the beginning of the representation, that the lawyer will provide some but not all of the services to address a legal matter, while the client will address the remaining issues. The unbundled model has been common for corporate or institutional legal services where outside law firms tend to team with in-house counsel to address legal matters. Unbundling is also recognized as an emerging model for providing legal services for under-represented parties, often in areas such as family law, consumer, housing and small claims matters.

Artificial intelligence will speed the development and adoption of unbundled services for both corporate and under-represented clients. At the corporate level, artificial intelligence will be brought to bear on project and success based billing. The shift from hourly billing will allow in-house counsel to segment legal assignments according to the price and expertise of outside counsel. Technology will catalyze the trend toward unbundling the delivery services to under-represented parties by automating more and more discrete aspects of overall legal representation.

I. New AI-Based Pricing Models for Unbundled Services Delivered to Corporate Clients

The legal industry is grappling with disgruntled clients and upstart competitors. Start-ups and some incumbents are exploring the possibility of using predictive technology and big data analytics to measure value rather than billable hours. Although most firms are still wed to traditional billing models, they are now at least open to the possibility of following other professional services in the move towards new outcome-based price models rather than time and materials billing.

Corporate general counsel long ago adopted the process of disaggregating traditional law firms, taking advantage of new competitors such as Axiom that reduce costs and increase efficiency through technology, streamlined workflow, and alternative staffing models. Those emerging businesses are helping general counsel move beyond cost and brand as proxies for quality toward the “Yelpification of law.”

As transparency about the quality of legal services grows, clients are also reducing their reliance on solution-shop providers. They can assess the jobs they need done and funnel work to the firms that are most appropriate for those jobs. This disaggregation may mark the beginning of a shift in Big Law’s competitive landscape from the dominance of integrated firms, which are designed to conduct all aspects of the client engagement, to firms that specialize in one substantive area.

The shift toward unbundling will accelerate as new pricing models show clients that they are paying too much for features they don’t value while depriving them of efficiency, responsiveness, and control in the handling of their legal matters.

Disaggregation will place increased pressure on law firm margins, but it isn’t all bad news for well-managed firms that apply the right mix of technology to the project fee puzzle. At its core, unbundling can be a form of price optimization. And thanks to the growing availability of internal and external data, advances in machine learning, and increases in computing speed, price optimization can be applied more broadly to the law firm’s advantage.

The key challenge for the application of algorithm-based project fees is finding the right mix of machine learning strategies and sources of data. Most unbundled legal services are, from a machine learning standpoint, “first-exposure” services that have never been priced before. One approach to predictive price modeling for “first exposure” professional services is to employ a clustering technique that mines data to look for similar projects that can be isolated based on historical data. Those projects can be clustered with other matters having a similar structure.

Once the clustering technique is applied, algorithms can be developed to estimate the hourly input for an unbundled service and the cost of that input to the firm. Regression trees, a machine learning technique developed in the 1960s, have proven to be the best tool for predicting optimum pricing for two reasons. First, it can successfully cluster projects completed in the past and use only the relevant projects to predict demand for the current project being analyzed. Second, it works in special pricing situations, such as the inverted price relationship common in luxury goods, where demand actually goes up when the price rises. That condition is present in unbundled services because clients are willing to pay higher rates for isolated projects or targeted outcomes. Traditional linear regression techniques can’t handle this special situation, but it is critical for pricing specialized services, particularly where price is often interpreted as a signal of quality.

The development of unbundled legal services pricing algorithms would be divided into two stages: learning and fee optimization. During the learning stage, analysts would observe client decisions about whether or not to undertake an unbundled project based on certain fees and fee structures. At the end of the learning period, it should be possible to set fees for particular episodes in a deal or litigation or for specialized advice. If firms are able to run the test parameters for a long time (or test them against a large corpus of anonymized data across firms) firms will develop a good understanding of the proper pricing for unbundled services.

Accurate forecasting of fee setting for unbundled services will require firm management committees to create the conditions that are needed to take advantage of the opportunity. To succeed in developing data sets that will allow algorithms to recommend pricing for unbundled services, firms will need to be open to the concept of anonymized data across peer group firms, a practice that is already common in evaluating overall firm profitability.  

Artificial intelligence will also play a significant role as algorithms permit clients to undertake their own costing and pricing analyses using automated tools. Solutions featuring greater predictive technology and automation will only get better with time. What’s more, data analytics and big data radically level the playing field of fields like law that lack transparency in pricing. Their speed and quantifiable output help reduce, and perhaps even negate, brand-based barriers to growth; thus they might accelerate the success of emerging firms.

As technologies for machine learning and artificial intelligence become more advanced and the dimensions of available data expand, the pricing of unbundled legal services will become more sophisticated and dynamic, enabling firms to project cost estimates in real-time. In essence, firms will have the ability to set fixed fees that continuously adjust to changing matter scope while also responding to lawyers’ capacity to take on additional work, profitability requirements and other external pricing influences. As today’s largely unstructured billing data becomes more structured using better tracking tools, firms will be able to marry rich data sets with sophisticated pricing models and apply advanced analytics and machine learning techniques to produce unbundled pricing alternatives across practice areas.

The application of technology alone, however, will obviously be insufficient on its own to enable true unbundled pricing. All pricing actions need to consider the firm’s reputation and areas of expertise, as well as overall firm profitability (as distinct from unbundled matter-level productivity). That analysis will require new thinking and oversight in the form of constraints and outcome objectives. Balance is achieved when the machine models become enablers for the human to apply business judgment to an ever-expanding and changing source of data. The computational power of artificial intelligence in the hand of the human-guided pricing decision will quickly become the new norm.

Of course, many firms still reject the notion of disruption in the legal industry. For one thing, it can be very difficult to get large partnerships to agree on revolutionary strategies, particularly when decision makers have the most to gain from the status quo. They point to the purported impermeability of their brands and reputations. Why try something new, they ask, when what they’ve been doing has worked so well for so long?

II. New Delivery Models for Unrepresented Parties

Lawyers representing clients on a pro bono or reduced fee basis will benefit from leveraging artificial intelligence to deliver unbundled services to their clients. The technology will allow lawyers to provide higher quality service to people who might otherwise have little or no meaningful access to justice.

At the most basic level, automated document-assembly systems will employ natural language processing functions to interact with clients on a conversational level so that responses to questions and automated forms are revised automatically. Suggested changes will be made based on data collected from a large body of users, such as other members of the community accessing specific unbundled services. The systems could also record the editing history of documents to prompt users to make choices on how to complete forms. Through those processes, client libraries of legal forms, instructions and other documents can be assembled. A legal aid office might set up unbundled “packages” of services and post them on a website so that a prospective client may select from various options when searching for online legal assistance. The prospective client can select the desired legal service to be guided through an automated virtual law office from registration through a conflict of interest and jurisdiction check, specific client intake forms and automated responses required of the selected unbundled package.

Artificial intelligence could also be employed to deliver process and decision support to unrepresented clients. Community Legal Education of Ontario (Canada) --headed by the remarkably inventive Julie Matthews--has made a good start at using decision tree systems to provide basic guidance to clients on specific legal problems. Work is now being done on those systems to deploy NLP and machine learning capabilities to improve ease of use and comprehension of resulting guidance.  

Libraries of NLP and machine learning supported documents and decision support systems will be further improved by publication on open source facilities such as Github or through bar associations. That way, in addition to the self-help model, forms and automated advice could be modified and improved by legal aid offices or with volunteer lawyers serving specific communities, including communities speaking regional languages or facing unique legal problems, such as aboriginal communities. Those improvements would then be published and used to enrich the available data set.  

A crowdsourced, open-sourced system contributed to by the entire legal aid community would quickly be able to grow to scale over a period of a few years to amass a large database of unbundled legal assistance that would continually be updated and developed by its user base.

While such a system would not be appropriate to serve many individuals needing full-service and in- person legal assistance, or those who may be served with unbundled deliver but who are less comfortable with technology, this would provide a starting point for a centralized and standardized system of unbundled legal knowledge that could be built upon.

Charlie von Simson

Charlie von Simson is a legal subject matter expert at ROSS. He practiced law for twenty years before running away to join a startup.