default logo

Meet Your New Lawyer, IBM Watson

Will IBM Watson be your new lawyer? Perhaps someday. For now, I tackle an easier question: what role can Watson play in Big Law? I speculate here so welcome comments and corrections.

Last week at the International Legal Technology Association annual conference  an IBM Watson senior manager presented to an invitation-only meeting of Global 100 law firm CIOs. Some points here stem from second hand reports.

It sounds like IBM intends Legal Watson to replace junior associates or at least perform much of their work (see also The Future of LawAmerican Lawyer, Aug 2014). Legal Watson’s success depends on the answers to three questions:

  1. What information – and how much – does Watson need to ingest and, to the extent necessary, who will tune or train it?
  2. If Watson ingests enough and is tuned, will it exercise good legal judgement? That is, can it actually replace some or all of what junior associates do?
  3. If its judgement is good, what is the market strategy?  That is, who will use it and why?

 

Let’s tackle these in turn. Loading primary legal information seems easy enough. A lot of raw law is public domain, for example, state and federal cases, agency materials, and EDGAR filings. Watson likely also needs more refined lawyer know-how, specifically work product (e.g., memos and briefs) or published treaties. These materials may not be so readily available to power Watson.

Perhaps Legal Watson can work closely with a large law firm: its document store for ingestion and its experienced lawyers for tuning. The motivation of a firm to provide such access depends on market economics (Q3), so more later.

Assessing Legal Watson’s reasoning requires empirical testing (Q2). Jeopardy offered a well-controlled situation with clear rules and outcomes, which made testing easy. Not so for legal reasoning – it will be expensive to compare Watson outputs with that of experienced lawyers.  Cost aside, comparisons may be hard because, unlike Jeopardy or clinical trials for drugs, a legal reasoning gold standard is elusive.

Consider split Supreme Court decisions with multiple concurrences and dissents – even scholars struggle to parse these. Or consider the fuss over e-discovery predictive coding; if that is an indicator, then agreeing Watson works will be hard indeed!

The go to market issue (Q3) seems hardest. Neither law firms nor law departments typically embrace new technology to enhance efficiency and effectiveness.  Simply look at the limited uptake of document assembly over the last 30+ years. It’s not clear why Watson would be different.

Beyond lawyer resistance to change, consider the economics of deploying Legal Watson. Alternative fee talk notwithstanding, large law firms still make much money selling hours. So why would firms invest in a tool that reduces billable hours?  I wish I did not need to ask that but I’m dealing with reality.

So let’s return to asking why Big Law would contribute documents and time to help build Watson. Were clients more demanding or firms more inclined to innovate, then perhaps we would see a taker.  But the client bark is worse than the bite and law firms rush to be 10th, not first.  So I don’t see it happening.

If my assessment is correct, then IBM will need to buy legal analysis and experience at retail – that is, treat building Legal Watson it like any matter. Doing so would be expensive. I leave for another day the privilege question. What about the wholesale route…

IBM should buy a law firm. Owning a law firm, IBM could ingest its work product, tap its lawyers to tune Watson, conduct controlled bake-offs between Watson and lawyers, and have lawyers stand behind the machine’s answers.

Ethics rules, however, forbid IBM from buying a US law firm. And IBM has already eliminated the option of seeking a rule change: IBM General Counsel Robert Weber co-signed a February 29, 2012 letter arguing forcefully against allowing outside ownership. UK and Australian rules do allow buying a law firm; were IBM to do so, it would make a mockery of the letter.

Of course, IBM does, in effect, already own a law firm – its corporate law department. So perhaps the best way to prove out Legal Watson is to deploy it there. Moreover, Mr. Weber, rather than lobbying to preserve archaic rules, could create more value by building a coalition of large law departments to worth together to make Legal Watson succeed.

All of this is idle speculation.  I hope someone who actually knows the story steps forward.

  1. Joel Ginsberg

    Interesting. Maybe someone could make money retailing fully automated online self-help “legal” services directly to consumers, providing lots of disclaimers that the service is not to be construed as legal advice and limiting queries to areas where the law is fairly clear-cut. Or maybe Legal Watson could be the basis for a paid online research tool lawyers and paralegals use to get answers more quickly, again in a constrained and well-structured domain such as the Code of Federal Regulations and the Federal Register. Also seems like Legal Watson could complement Lexis/Nexis or Westlaw, who already possess a lot of intelligence about successful queries and the structure of legal logic and rules…