Jay Budzik, co-founder and CTO of Intellext was the speaker at the June KM Chicago meeting. Jay created Intellext to commercialize his Northwestern University research in a new product, Watson. Watson sits in the background and accesses search engines on your behalf as you work, presenting you with context-sensitive results from both internal and external searches.
Why is this interesting? The promise of corporate knowledge management and search and content management is to drive efficiency in organizations. An oft-quoted IDC research report claims that knowledge workers waste a lot of time searching for information that either can't be found or doesn't exist to begin with (Portals Magazine reference). On the other side of this equation, Jay quoted Kit Sims Taylor as saying that knowledge workers spend time unwittingly recreating existing knowledge (information).
Classically, people do searches with very few terms and the search tools get no benefit of context-sensitivity. In addition, the burden of finding content is on the user, and they frequently have too many places to go and too many results come out of the search.
So, what does Watson do? In essence, it does complex searches to bring back the smartest, most context-sensitive set of results possible. It sits in the background of the MS Office suite, forming search queries based on the content of the item being created. (It also works in the background of your web browser to find related content.) When the user pauses, it fires off queries to the configured search tools: News, Google, premium content, internal databases, etc. When Watson decides the results are significant, it lets the user know that it has results and they can view them if they wish. As the user continues working, Watson modifies and refines the queries, providing as much context to the queries as it possibly can. If you are writing about Jaguar in the context of computer operating systems, Watson should be smart enough to bring back results having to do with the Macintosh OS, not the animal or the car.
Jay tells the story of a couple of trials of the tool in corporate environments that really give a much better picture. In one case, they have been testing Watson with it connected to the corporate CMS. Someone began creating a PowerPoint presentation, and after three slides, Watson was able to inform the user that there were several similar presentations already available in the CMS. In a similar instance, a user was able to begin creating a complex drawing and discover a nearly identical copy sitting in the drawing management system, potentially saving weeks of creation time.
At a more public level, Watson can monitor the activity within a web browser and look for specific types of data or information. For example, if it sees that you are looking at books at an online bookstore, it could let you know that those books are available at your local library, so you needn't buy them.
I've played with Watson a bit and find it interesting (for the free trial). I'm not working in a corporate environment, so I don't have links to unusual databases. In preparing for my KM class, I was able to find additional materials to round out discussions. It would be nice if I could integrate Watson into other applications, such as my weblog editing tool (BlogJet), so that I could find related content as I write my posts.
A number of questions at the end of the session revolved around "how can you sell a desktop application in the corporate environment," where it is assumed corporate IT will only accept server-based applications. Jay's return argument is that Watson isn't a centrally-served application. It needs to "see" what the user is doing and run searches on their behalf, and that is best done at the desktop level. The biggest thing that corporate IT should worry about is the additional search load created by Watson, if it continually refines the search as a user works in a document.