After last’s week’s announcement from SAP about their Enterprise Search offering, I’ve been tossing around several ideas about enterprise search. Here’s one:
Unstructured Applications
High value is in searching “Unstructured Applications.” While searching “unstructured content” comes up often and is the key first step for enterprise search, I’ve begun to think that searching “unstructured applications” will (or is) the real high value in enterprise search. By “unstructured application” I mean taking an existing, legacy system, and someone how indexing it’s content; even better, allowing users to click on “actions” in the search result.
For example, at a previous company we used an ancient copy of Vantive, a bug/issue tracking system. It was classic 90’s client/server and, by today’s standards, it was painful. The company tried to upgrade to a modern system (their own, embarrassingly enough), but they’d sunk so much customization and data into Vantive that it was impossible given the resource spend allocated.
But, an enterprising (pun!) employee had written a very thin search client layer on-top of Vantive’s ancient, but functional API. You could even click on results to edit the issues/bugs. Of course, this client was completely grass-roots and passed around via email instead of being blessed as a company-wide product. But: aren’t all improvements to info-workers life such?
Thus, without the company spending time and money on a new tracking system, there was a way to get more blood out of the Vantive stone. Now, imagine if instead of a custom client for all that, it just integrated with your Enterprise Search results. Sounds like useful SOA on the cheap, huh?
More Blood from the Stone
Now, every enterprise is full of all sorts of applications jam packed with valuable content. You can’t throw a rock in enterprise software pitch-land without encountering that idea. These are “unstructured applications” in so much as they’re not designed to be content stores that easily interoperate with mashup/composites. And yet, the vast majority of them have APIs that allow you to interface with them. To me — putting on my programmer hat — this means they’re just waiting for a thin, cheap to produce and maintain layer on top of them that will interface with search.
Now, if I were one who sold enterprise software, I’d deploy my FUD-bombs of compliance, auditing, and managing access to all of that information. Those are indeed, important, but the difficultly of taking care of those concerns shouldn’t distract from the core idea. Indeed, a little COA is probably the answer enough.
What makes me excited about this line of thinking is that I’m always trying to think of how enterprises can get more value from software they already have without having to spend the time and money to upgrade. Interestingly enough, the SDN DemoJam winners this year provide a great example along these lines: they layers a thin Rails based web-app on-top of SAP. As Dan put it in his talk, “when my boss saw it, he was floored.”
Disclaimer: SAP paid my T&E for TechEd. And extremely indirectly, through a long chain of acquisitions, Oracle (who owns Vantive now) is a client.
Tags: search, sap, sapteched2006
The problem with most enterprise search is that is just a web front end to a command line interface, like "Find Customer John Smith" if you use something like "What is John Smith's address" it won't work. I think this is the defining issue in this space. Until someone can figure out a way not to just wrap the APIs but actually infer meaning from the object those APIs represent your Enterprise search will be nothing more then a faux-command line interface. People want Google.
I think it is possible to give some of that meta information about data in the BI arena, which is how i think you could then link back to the "unstructured" data that you are referring to…..however, that only exists in my head and therefore doesn't count as possible.
Although i make reference to BI being the starting point for this meta data the idea of Natural Language Processing extends here too, i would love to be able to ask my Business Intelligence system, "Graph top 10 customers by net sales." I'm no analyst but I don't think anyone has even tried such a thing yet.
-d
In a word: Yuh!