By Ron Friedmann and Oz Benamaram (the then Practice Resources Attorney for Morrison & Foerster LLP where he provided the firm strategic direction in the use of technology and knowledge management to support the substantive practice of law)
This article (in shorter form) was originally published as IT @ Morrison & Foerster: Lessons Learned from Retail in the December 2005 issue of Law Technology News (article at LTN; article as PDF on Morrison & Foerster LLP web site here.)
Lawyers As Shoppers:
It’s All About Finding Information
Knowledge Management Innovation at Morrison & Foerster LLP
“Is there anything else relevant to our case?” was the question I dreaded most as a law clerk. “Very likely, your honor, but I haven’t checked all of Australian and British case law” did not seem an acceptable answer. With growing access to information, how do you find the relevant needle in the ever-expanding haystack?
Or the perfect shoe in the shoe store? Unlikely as it may seem, lawyers can learn from retailers, who turn out to be pretty good at figuring out what shoppers want. When we walk into a shoe store, regardless of how specific or vague our request, the sales clerk typically suggests two or three pairs that are good choices. And when we shop online, single-word searches plus a couple of mouse clicks often yield a short list of attractive possibilities. Retailers enable customers to select quickly. To do so, they have to zero in on what’s relevant.
A similar principle guided Morrison & Foerster’s knowledge management and systems development: provide relevant search results in a fast and easy way, to present our attorneys with information they can act on.
This article describes how the firm designed its next-generation knowledge management approach by linking information from multiple sources to provide more context about our people, the clients we serve, and the documents we write. It also explains how the firm is executing this approach and how it has culminated in an innovative form of enterprise search.
The principle of “relevant, fast, and easy” emerged from studying how our attorneys tap existing information sources and process. In particular, we focused on use of email, mailing lists, our Portal, and “Knowledge Exchange.”
Knowledge Exchange is Morrison & Foerster’s collection of best practice forms and exemplars created by our attorneys. Each practice group assigns an attorney on a rotating basis to identify, categorize, and designate selected documents as high-value precedents. Although manually collecting and maintaining documents is inherently time-consuming and does miss some useful precedents, the Knowledge Exchange has largely succeeded. Attorneys can easily query it for precedents, organized by a taxonomy, for reuse. Observing attorney use of Knowledge Exchange precedents yielded three insights:
Insight: Attorneys need context to use precedents
An attorney needs to know the context in which a document was created to understand it fully and reuse it properly. For example, it helps to know what the litigation was about, the industry of the client, the type of deal, and who had the bargaining power.
Implication: “Connect people to information”
Improve attorneys’ access to existing information by linking various information sources to provide context.
Insight: Attorneys use precedents to identify expertise in the firm
Attorneys find it valuable to talk to experienced colleagues. Often they ask attorneys they already know because they do not know whom else to ask. We found that sometimes they use documents in Knowledge Exchange as “pointers” to lawyers with expertise. The guidance of an experienced attorney often turns out to be much more valuable than document contents alone.
Implication: “Connect people to people”
Enable attorneys to locate colleagues with relevant experience quickly and then make collaboration among them easier.
Insight: Attorneys use only simple and relevant systems
Too many sophisticated systems with too many ways to find information turn attorneys off. Yet even the most tech-phobic knows how to perform Google or Yahoo searches. Lawyers are much more likely to use internal systems that are as simple as what they already use on the web.
Implication: “Make it simple”
Keep the interface as simple as possible and off-load tasks where possible from attorneys to automation or staff.
After analyzing our needs and available technology, we realized that, like better retailers, we could make good predictions to help our shoppers (that is, lawyers) find the information they seek. The user interface would offer Google-like simplicity and power. Behind-the-scenes, the enterprise search system would identify hidden and explicit contextual information from multiple sources and automatically make inferences to create non-obvious meaning for users.
We evaluated “federated search” engines to find those that would access multiple databases, including documents, e-mail, portal contents, matters database, time and billing system, contacts, conflicts database, and ethical walls sources. But that was not enough. The software also had to “make sense” of the morass of data by applying logic and semantic analysis to suggest the most appropriate resources.
After reviewing dozens of vendor responses to our request for information and performing initial evaluations, few met our requirements. Though we had a long list, many focusing on back-end issues, the key differences among the vendors were their ability to:
- Identify experienced attorneys based on the documents they wrote and the clients they served.
- Order search results by likely relevance:
- Give extra weight to more valuable documents, e.g., a legal memo is assumed more relevant than a fax cover sheet; documents tagged as precedent are assumed to be of higher importance.
- Adjust relevance ranking by cross-referencing information, e.g., raise the relevance of a document whose author’s biography, or the description of the matter it was written for, includes the search term.
- Personalize rankings by inferring a user’s interests based on what we already know about that user from our internal systems, e.g. their HR affiliation, personal biography, documents they drafted and matters worked on — similar to “people who bought this product also bought these”
- Scale to handle large amounts of data (terabits of information).
- Guide users quickly to narrow the results in a meaningful way.
To achieve our goal of “fast and easy,” we placed our bet on “enterprise search” as the right supporting technology, but we found we needed to collect more information about matters to provide context and help locate experts.
We crystallized our new understanding in a formally defined “Knowledge MAP,” a set of principles that guide planning our next-generation Knowledge Management infrastructure:
- Matters provide context and connect information from multiple systems
- Attorneys, once identified, have knowledge that they can share
- Precedents such as documents and e-mail can be re-used, especially when contextualized
By linking these three elements together, we are able to amass more useful information on each of them: we can find documents based on the client industry, matters based on types of motions drafted, and attorneys based on the matters they have worked on and documents they created.
Stating our principles succinctly, particularly the value of context and the need to identify expertise was also essential in gaining support for change. We have built a system that collects information as attorneys open matters, revisits descriptions as matters progress, and again as they close. We now capture information such as lawyers for other parties, closing date and deal value, and verdict day and trial outcome.
As we evaluated search engines, we found that our decision to focus more on matters and less on documents was paying off:
- Matters unify information from all of the firm’s discrete systems and thus are a useful way to present information.
- Capturing information about a matter overall is less costly and more practical than doing so for all of the individual documents that belong to it.
- Staff support for gathering detailed matter information is strong, since departments other than KM (e.g., marketing and finance) also need the information.
- Matters, when linked with attorneys and documents, provide a simple and elegant way to identify experienced attorneys.
Auto Profiling Documents
Although we have placed more emphasis on matters, documents are still important. Our analysis of e-commerce engines and recognition of the significance of context drove us to ask a new question. Could we, without manual intervention, provide more context about individual documents as well? If we knew more about each document — for example, party names, dates, jurisdictions, and terms describing deal types — we could improve retrieval and make valuable associations and cross-references.
Asking attorneys to enter this information in document management system profiles is simply unrealistic. Few law firms persuade attorneys to complete more than a minimum of profiling fields. Some attorneys do not even enter meaningful titles, and many document type choices are suspect.
We decided not to try to change our lawyers; instead, we looked for software that could automatically identify “hidden context.” We found the solution in “entity extraction” software that had been designed to strip documents of identifying information in order to address confidentiality issues. Once we realized we could “auto-profile” documents, we expanded our evaluation to include specialized document retrieval systems that have the potential to populate profiles automatically. Either way, the goal is to recognize entities (e.g. party names and jurisdiction) or concepts (e.g., deal type) within documents and automatically populate the document profile. More detailed and accurate profiles are critical for “faceted” search, which we describe next.
Adding Faceted Search
Even with advances such as organizing hits around matters or taxonomies and ranking results with sophisticated inferences, search results were still too extensive. Our shoppers could not make good choices quickly enough. We needed even more “easier and faster”.
We returned to the world of e-commerce where the technology to sort and select quickly already exists. To grow sales, e-retailers need to make it easy to rapidly review and choose products from their large catalogs. To do so, better websites use “faceted search,” which allows consumers to “slice” search results by important attributes such as brand, price range or product features. This dramatically reduces the time needed to find relevant items by recommending relevant ways to find products. In effect, the system makes good guesses about what consumers may want.
We added faceted search to our KM requirements because attorneys, like shoppers, need to choose one or two items from potentially long lists. Attorneys searching for specific matters, people or products can “slice” their results by attributes such as jurisdiction, industry, motion type, party names, governing law, effective date, or law firm on the other side of the deal.
Beta tests with groups of lawyers demonstrated that faceted search is critical. This provided further impetus for manual efforts to collect matter data and automatic efforts to profile documents.
Morrison & Foerster is rolling out its new enterprise knowledge management solution, which we call Answer Base. We designed it by observing and analyzing users’ needs, evaluating off-the-shelf technologies, and working with vendors to add functionality to existing software. For enterprise search, we selected Recommind; we are still evaluating the finalists for document profiling.
We believe that our automated approach, supported by some manual work to describe matters, is bringing Morrison & Foerster much closer to the holy grail of knowledge management — quick, easy, and reliable access to people and information without spending a fortune. We are giving our “shoppers” what they really want.
We are now seeking technologies that will take us to the next phase. Our current solution allows users to find information. Next, we want the system to suggest actions based on the information found. Like shoppers, our attorneys seek information for a reason. For example, “Client X has been sued: press 1 to start a conference call with your most frequent contact in the client’s organization; press 2 to get billing status of the client; press 3 to print the court documents to your default printer; or press 4 to delegate this task.”
Helping our attorneys perform actions will require a deeper inferred understanding of our workflow and may well seem futuristic and even unattainableâ€¦ but only a few years ago that’s what we were told when we set out to create what we have now achieved.