Last month, IBM's (IBM) Watson supercomputer won $1 million on Jeopardy. But the computer's defeat of Ken Jennings and Brad Rutter was more than just a publicity stunt for International Business Machines. At the core of Watson's success was something called Semantic Analysis Technology (SAT) -- a way for computers to do much smarter searches of large amounts of data.
Companies are using SAT from IBM and its competitors to boost their profitability through a wide range of applications, including better customer service, much more efficient analysis of documents used in lawsuits, and uncovering and shutting down financial fraud before it costs banks money.
Companies are getting big returns on investment in SAT, and so can you -- by investing in those that supply this technology: IBM and two other companies you may not have heard of: BAE Systems (BA) and Autonomy (AU). There are also privately held companies that could go public if they can take advantage of the growing market for SAT: Keep an eye out for Cataphora and Clearwell Systems among those potential IPOs.
How Watson Wins for Businesses
But what exactly is SAT? IBM describes it as a technology that enables clients to search a complex data store on the basis of a cluster of terms sharing a common meaning, rather than a simple key word.
For Jeopardy, Watson downloaded about 200 million pages of material -- including encyclopedias and movie scripts. When Watson got a question, say, in the category "Church and State," about Saturday Night Live's famous "Isn't that special!" comment by Dana Carvey, it searched those pages -- known as its corpus -- and came up with hundreds of possible answers that fit some of the clues.
Then, according to IBM researcher David Gondek, Watson applied an algorithm to assign a confidence number to each answer, based in part on the number of connections to the clue, and weighed those numbers -- picking the best one. In that case, Watson's guess -- "The Church Lady" -- was correct.
While the application of SAT is still emerging in different business arenas, three areas seem to be most promising for the technology:
IBM has used SAT to help insurance companies to convert dissatisfied customers who were getting ready to switch insurance providers into long-term customers who renew their policies. Through SAT, IBM has helped insurers such as Japan's Dai-Ichi Life and U.S.-based Assurant Solutions, a unit of Assurant (AIZ), to boost their revenues by an average of 37% from customers who renew their policies: The number of those rose 119%, according to my interview with IBM's Christian Bieck.
SAT revealed that specific customer service agents are better at satisfying specific customers, based on matches between certain traits of the agents -- such as where they grew up -- and characteristics of the customers -- such as education or income according to my interview with Assurant Systems Vice President, Cameron Hurst. By matching up agents with customers they are likely to perform best with, IBM helped the companies boost revenues and profits.
2. Boosting litigator productivity
During the discovery phase of a lawsuit, lawyers may need to pore through as many as a million documents to find the relatively few -- say, 50,000 -- that will be relevant to their case. Three or four attorneys can spend months -- costing clients $500,000 or more -- to come up with a short list of documents, only 60% of which are relevant to the case, according to Clearwell Systems marketing executive Kamal Shah.
But thanks to e-discovery software from companies such as Autonomy, Cataphora and Clearwell Systems that uses SAT, litigators can spend about $100,000 to get 95% accuracy in just a couple of days, said former Cataphora client Bill Herr, who headed e-discovery for a large chemical company. The resulting boost in litigation productivity is creating a fast-growing market for e-discovery software.
3. Nipping financial fraud in the bud
Fraud against banks, insurance companies and tax authorities is rampant. I interviewed Imam Hoque, Managing Director of Detica NetReveal, whose parent, Detica Limited, was acquired by defense contractor BAE in 2008. According to Hoque, a particularly prominent form of financial fraud is called bust-out.
In a bust-out fraud, the perpetrator deposits $10,000 a month into, say, 10 accounts whose identifying information the perpetrator has stolen from unsuspecting foreign students with no credit histories. After a few years of regular deposits from the perpetrator, the bank decides that these 10 account holders ought to get loans due to their apparently excellent cash flow. The break-out perpetrator gets the loans, shuts down the accounts, and skips town.
Hoque estimates that a large bank might lose $100 million from break-out fraud -- but if that bank spends what I estimate to be about $2.5 million and several months installing Detica's software, it can map out the links in those account holders' financial networks and nip the fraud in the bud. The SAT software will recognize and flag, for example, that those 10 account holders aren't making regular payments out of their accounts for things like rent or credit card bills. That will cue the bank's fraud department to dig deeper, determine that the accounts are fraudulent, and shut them down -- before making loans to the fraudster.
How You Can Profit
In the longer term, the best way for individual investors to profit from the boost SAT brings to business productivity may be to invest in the IPOs of e-discovery companies -- should they happen.
Of the three companies mentioned, Autonomy is the closest thing to an SAT pure-play. It bought e-discovery company Zantaz in 2007, when it had sales of $106 million and the e-discovery market was growing at 47% a year. Assuming that those sales kept growing at 40% a year, in 2010, Autonomy may have garnered about $291 million in 2010 sales from its Protect e-discovery software, representing about 33% of its 2010 sales of $870 million.
Beyond that, SAT represents only a small portion of the revenues of the other public companies that provide it. For example, Detica's NetReveal revenues are probably around $40 million to $50 million -- compared to parent BAE's 2010 sales of about $22 billion. And IBM's SAT revenues are just a tiny fraction of its $100 billion in 2010 revenues.
So if you want to bet on the future business benefits of Watson's technology, buying stock in Autonomy may be your best bet for now.