Morrison & Foerster's "AnswerBase" KM Initiative: Learning from Wal-Mart

When I wrote about the Baker-Robbins/LegalWorks KM conference, I purposely left out the most impressive application/presentation of them all:  Morrison & Foerster's Oz Benamram discussing "AnswerBase," the firm's new KM system which will be rolling out next month.

AnswerBase is the fruit of over two years of labor, and is, in my humble opinion, a revolutionary approach to KM.  Oz was kind enough to give me a one-on-one guided tour in his office two weeks ago, and what I have to say will draw from both his presentation to the KM conference and to our private meeting.   Suffice to say that neither Oz nor I am aware of any other firm taking the Morrison & Foerster approach at the moment, but when I asked Oz who else might adopt it once they see it, his response was "Everyone will, within two years."

 Read on.

At the outset of their redesign of the Morrison & Foerster KM system starting two years ago, Oz and the team went back to first principles.  These were their guiding stars:

• We need "federated search" to search across the many disparate databases which all contain information potentially germane to a lawyer confronting a new task, including:
  • the matter tracking/management system;
  • the client/CRM databases;
  • the financial/accounting/time-keeping and billing databases;
  • personnel information on individuals within the firm;
  • the document management system;
  • the email database; and last but not least
  • the firm's own internal "Knowledge Exchange" system, a continuously-upgraded and dynamic compilation of (manually managed) model documents and templates.
• To use precedents effectively, attorneys need context: who worked on the transaction, what industry was it for, the timing, etc.
• Often the most expeditious way to gain expertise is...by talking to an expert:  This implies that the system must excel at identifying people who have worked on similar matters in the past, and preferably a lot of them.
• Finally, lawyers won't use anything that's not drop-dead simple.  Extremely comprehensive and nuanced search tools may be fine for grad students, but lawyers want something resembling Google or Yahoo.

Perhaps not surprisingly, when they went out into the marketplace of "federated search" vendors to evaluate products, they ran into the realization that while everyone could do 80% (sometimes a different 80%) of what they were seeking, no one could do it all.  Products that excelled at extracting meta-data to identify entities to a transaction, for example, fell down on their relevance-ranking engines, so that the "best" documents did not always appear at or near the top.  Similarly, products that were strong on identifying individuals with relevant experience mis-categorized documents.

At this point, the team was in a bit of a quandary—until Oz happened to attend an "enterprise search" technology conference where some e-commerce vendors were making presentations.

When you or I think of e-commerce, we tend to think of Wal-Mart, Home Depot, Barnes & Noble, not the AmLaw 50.

But Oz's insight was that e-commerce platforms have several built-in capabilities that more conventional engines used to power legal KM systems may lack:

• they are "scalable" beyond belief;
• they make allowance for misspellings, imprecise phraseology, etc.;
• at least with the best-of-breed, they avoid the classic search failure mode I refer to as "all or nothing"—where the answer to your search is either "Search returned no results" or "Showing 1-10 of 2,409"
• they "hate" to come up empty-handed, so are configured to provide near misses and close neighbors rather than "Try Again."  (For example, if you were searching for a 2005 black Honda Accord coupe with a 6-speed manual, and there were none in stock, it might return a 2004 fitting those specs, or a four-door sedan, or a red one, and ask you which criteria were most important to you so it could re-order and refine the results.)
• perhaps most compellingly, they come ready-made with the ability to conduct "faceted search," a term perhaps more readily understood by example than strict definition.  "Faceted search" simply means the ability to categorize the answer set of a search by relevant characteristics.  Endeca, a leading vendor in this area, with clients including Barnes & Noble, Boston Scientific, Circuit City, CompUSA, Home Depot, IBM, the Library of Congress, NASA, Patagonia, Putnam Investments, and Wal-Mart, provides this example after one has searched for "Lego's" at eToys:

Although difficult to make out, you can see that of the "172 results" returned, it invites you to recategorize them (left-hand column) by Age, Gender, Price, Category, Character, etc.  In law-firm-land, the equivalent is offering to recategorize the results of a KM search by client, industry, type of transaction, jurisdiction, office where it was managed, responsible attorneys, date, or even the identity of the law firm on the other side.

Even given the inspiration of Endeca and the e-commerce model, Oz and his team ultimately settled on the proven platform provided by Recommind, which has worked with such name-brand firms as Cleary-Gottlieb, Cooley Godward, DLA Piper, Paul-Hastings, and Shearman & Sterling.

Finally, the Morrison & Foerster system obviously "knows who you are" when you're conducting a search, and adjusts its relevancy rankings accordingly, giving greater prominence to matters arising in your office or your department, or for clients you've worked for. Moreover, it knows how much you've worked on similar matters (say, an antitrust deal) and if you're new, or rusty, it will put training videos higher up in the search-return results.

If Oz is even one-quarter right that "everyone will be doing this in two years," KM professionals have a busy 2006-2007 in front of them.