Using AI for Time-Entry Analytics to Support Pricing Functions

Topics: Artificial Intelligence, Billing & Pricing, Client Relations, Data Analytics, Efficiency, Law Firm Profitability, Law Firms, Leadership, Midsize Law Firms Blog Posts, Practice Innovations, Staffing & Headcount


The climate of change in legal has given rise to a few cliché phrases du jour. Rhetoric surrounding the evil “robot lawyer” and its relentless siege against the sacred traditional methods of legal practice ranks second on my list of least favorite sensationalistic exaggerations, following only the “death of the billable hour” proclamation and predictions about whether this year will be the year of its demise.

While the image of a terminator-like counselor is indeed intriguing, the reality is that technology is slowly but surely permeating into various facets of the legal services industry. Whether it be quicker, more accurate means of performing research or drafting documents, or more powerful ways of analyzing and leveraging data to inform management decisions, technology is finally finding its way into the operations of legal services firms.

Despite the reluctance to abandon beloved traditions, it is simply becoming impractical not to leverage technology where there are advantages to be gained. Farmers use tractors to plow fields because it is just plain foolish to harness up oxen and head to the fields — why would lawyers not thrive from adopting similar means of increasing productivity, particularly when clients give every indication that it would be not just welcomed, but worthy of reward?

Aside from practice tools which have been making impressive and substantive progress in recent years, emerging financial- and operations-focused technologies seek to harness the power of the almighty time-entry narrative data. For those not immersed on a daily basis in the Rumpelstiltskin-like exercise of trying to spin straw into (data) gold, and in develop pricing models, budgets, and workflow models based on this golden data, a brief explanation may be in order.

The most useful data set in setting prices, analyzing profitability, maximizing efficiency, and evaluating utilization is work breakdown structure of a matter, phase, task, or activity. This blueprint enables lawyers and business professionals to dissect how work is, has been, or ideally should be performed. The ability to look at work patterns in historical matters and segment them out provides a map of the anatomy of a matter, and how work was staffed. Using this analysis, one can determine the optimal mix of resources for each component of a matter and a general expectation of the effort required from these resources to perform the work. One can also identify opportunities to increase efficiencies through process redesign and optimization, integration of technology, and things of that nature. It enables one to take a scalpel instead of a hatchet approach to pricing, budget creation, and workflow mapping. Once completed, the result is a cost-basis, or break-even point, to be referenced when projecting and monitoring profitability.

The most useful data set in setting prices, analyzing profitability, maximizing efficiency, and evaluating utilization is work breakdown structure of a matter, phase, task, or activity. This blueprint enables lawyers and business professionals to dissect how work is, has been, or ideally should be performed.

Here’s the rub: Lawyers are not historically incentivized to painstakingly document their approach, technique, and assumptions for future analytical value. The only point in the process where this is a priority is at time entry, and that is mainly because in most cases, clients ultimately review this, so it must create some testament to the value of the work being performed by each timekeeper. Over a decade ago when I began focusing on strategic pricing and analytics, I quickly learned that data integrity in most law firms is lackluster at best. The only point in the process of data collection where diligence is at all incentivized is time entry. Further, lawyers do not embrace anything that requires behavior change. As a result, I determined that anyone who could exploit time-entry data — including narratives and the billing metrics like rate and hours — would have a significant advantage. This brings us to the focus of this article.

Indeed, today there are a variety of vendors in the market — some start-ups, some more established players — that offer solutions focused on extracting value from time-entry data. The real challenge is the segmentation of activities as captured in the narrative. Because descriptions of work are governed by the individual style and capabilities of the lawyer entering the time, harvesting segmented insights from time-entry narratives can primarily be affected by either i) establishing uniformity, consistency and/or accuracy of time-entry content on the front end, at the point of time entry (generating structured data); or ii) processing the time-entry narrative in a smart, consistent, and accurate way based on the subjective writing style of each lawyer (categorizing unstructured data).

artificial intelligence

Each of these approaches can provide benefits, but the first approach actually improves the accuracy of the second. Because the second option imposes structure to the data on the front end, it makes processing it on the back-end even easier. Be that as it may, we will focus on the front-end solutions first.

Front-End Time-entry Applications

Some vendors have chosen to confront the variety in attorney writing style starting from the source, typically offering a time-entry application that sits on top of the firm’s billing system. This integrated approach enables the application to process the data before it is stored in system, thereby controlling the form and flow of data being captured. The practical magic of these solutions is performed by curating the time-entry narrative such that it clearly, accurately, and consistently articulates what was done, by whom, for what reason, and in connection with what other client or firm team members.

There are a few ways to accomplish the objective of this approach:

  1. Using a pre-defined, standardized time-entry narrative that is developed by lawyers, and sometimes customized, that accurately captures in standardized form what has been done by the timekeeper. The effectiveness of this approach is directly correlated with the level of diligence applied in the process of defining time-entry narratives that are technically perfect. (Perfect in this context can basically be defined using a business school or management consulting term: MECE, which stands for mutually exclusive, collectively exhaustive.) More simply put, every relevant activity is captured and accounted for, and none overlap with each other. This approach is similar to using auto-text or auto-correct on time entries, or sometimes a combination of both. For instance, entering in that WITP might be replaced with Prepared witness for trial.

In a sense one might conceptualize this as a more elaborate substitute for task codes. That comparison alone gives rise to some logical skepticism by virtue of the seemingly endless objections that lawyers have to entering time using task codes. If not done intuitively and applied in a way that can be done with incredible ease, there will always be room for problems. Sometimes these auto-populated, pre-defined narratives allow for free-form editing, and sometimes they do not. The trick is that if you allow for editing, you may have defeated the purpose.

  1. Employing “smart” time-entry software which processes the text as it is being drafted and either suggests standardized time entries, or takes a page from Google’s playbook and uses a “did you mean?” form of logic that increases the precision and accuracy of the narrative being recorded. Unlike the first option which either does not allow free-form user input or limits it, this approach relies on user-defined input. Provided the logic is accurate and the pool of terms and algorithms to identify them is deep, this approach can be helpful because the behavior change on the part of the lawyers is limited. If it is clumsy, imprecise, or irrelevant, however, a revolt is all but certain.
  2. Applying a “Big Brother” type of application which monitors and stores meta-data from users’ computers and phones, watching and listening in on everything that happens throughout the day. The application then applies logic that attempts to re-create the activities of the day either explicitly as draft time entries, or as a more generic digest that can then be referenced to increase the accuracy of the time entries that will ultimately be entered, sometimes the old-fashioned manual way.

The merits of each of these alternatives are relatively clear, but like anything else, the effectiveness comes down to the execution. Sophisticated algorithms that can capture all variations of tasks, and then effectively correlate them with the most precise and accurate time-entry narrative, are what determine if these approaches are an asset or a liability (or will be adopted).

When all is said and done, standardized, accurate and detailed time-entry data that doesn’t contribute to the “garbage in, garbage out” problem plaguing so many firms serves many purposes, including:

  • reducing write-offs resulting from client bill review of vague or inaccurate time entries;
  • reducing invoice payment cycles as time entries are rejected or challenged less frequently;
  • decreasing the amount of time needed for attorneys to enter time;
  • reducing pre-bill review time required from billing lawyers by increasing the integrity and clarity of the entries from the beginning; and
  • increasing the ability to effectively leverage data analytics during and after work has been completed to aid in optimizing staffing models, developing budgets, and improving price modeling.

As one might imagine, integrating an effective artificial intelligence (AI) component into a front-end solution could hold tremendous value; and as outlined above, some have begun to introduce the associated functionalities into their products. The ability for machine learning to continuously improve the accuracy and versatility of the predictive text based on the user’s input could deliver great benefits. Decreasing the required behavior change alone will catapult a tool to the top of the list of candidates. Further, it is easy to imagine that this capability would enable firms to derive narrative taxonomies custom-tailored to clients’ unique terminology and preferences.

Retroactive Time-entry Narrative Processing Solutions

More recent market entrants provide software applications driven by algorithms which categorize time entries into phases, tasks, and activities using the raw, unfiltered narrative entered by timekeepers. The objective of these tools is to be able to make analytical sense of work being done on a matter or a portfolio of work without requiring the lawyers to change their processes or behaviors. The solutions leverage Natural Language Processing technology — a branch of the AI tree — but one which uses algorithms built by software engineers using input from lawyers, not machine learning.

artificial intelligence

As outlined earlier, effectively segmented matter data has tremendous value to lawyers and law firm management alike, because it enables teams to drill down to see how much time was spent, by whom, performing specific tasks.

To put it simply, it gives you the blueprint of how the work on a matter was performed as a function of each of its component parts. Using this level of detail, it is possible to perform very precise analyses that would otherwise be impossible or prohibitively time-consuming and manual.

Benefits of this functionality include:

  • Pricing new matters using data from historical similar matters can be done with unprecedented precision and customization. Staffing models can be built using specific comparable phase-data from multiple past matters, enabling pricing executives to cherry-pick the most highly-correlated segments of past matters and combine them into a customized new matter staffing plan that closely parallels the unique combination of attributes of the new matter.
  • Once a pricing model is created, different staffing scenarios can be applied to determine if they will maintain the integrity of the work while simultaneously delivering savings and/or cost certainty to the client.
  • Once a matter budget and work breakdown structure are created, they can be used for matter management, both in setting expectations about the effort required of each team member assigned to each task, and enabling budget-to-actual tracking by phase, task, or activity.
  • Matters of a similar type can be compared side-by-side to analyze consistency among their components, which can be used both for establishing pricing certainty as well as for benchmarking workflow patterns to maximize efficiencies.

As transformational as this functionality is from an analytical perspective, these tools fall victim to “garbage in, garbage out” constraints if the time-entry quality is lacking. They also can only tell you how something was done, but they cannot offer meaningful suggestions on how it could be optimized. Users must also have some working knowledge of the principles of statistical analysis in order to avoid drawing improper conclusions about correlations between data sets that are either coincidental or not properly validated.

A Bright Future Ahead

While none of these products have reached full maturity yet, many can already deliver more than enough value to justify their expense. It is worth noting that the ones that have gained more attention in the market have improved dramatically in a short amount of time. This is an encouraging sign. As AI capabilities continue to mature, they can be applied to these products, promising an accelerated rate of evolution and functionality.

Further, there is a natural synergy between the problems each seeks to address, so it is easy to imagine a future where one application ensures the narrative data going in is solid and well-defined so that a counterpart application can process it accurately to help derive a wide spectrum of analytical insights and possibilities. Perhaps down the road these technologies merge into a seamless solution of integrated features.

It is important to recognize that all the best robots started as functional yet scaled down versions of their future selves. If there’s any doubt, one need look no further than the “Terminator” series of movies as validation of this principle.

Practice Innovations” is a quarterly, online newsletter that examines best practices and innovations in law firm information and knowledge management.