A little while ago I opened a discussion about IT ecosystems on a technical mailing list to which I belong. I posed four questions and asked for thoughts on the discussion. What follows is the first question, plus commentary from several other people, and my additional thoughts at the end. I will post the next three questions from the discussion in the following three days. I originally posted the questions  all at once.  I have pulled together the answers by question and edited them for brevity and relevance.

1. What role does enterprise search play in the eco-system? [My previous employer] implemented FAST ESP (www.fastsearch.com) for our main customer web site, but the system is capable of much more. FAST is trying to position their product as a way to consolidate lookups across your business, e.g. if you want to look up a customer record, don't go into your CRM system, go to FAST and search. In part the functionality is database offloading, but it is something else as well. I am trying to find the useful limits of that metaphor. Is anyone else looking at that kind of functionality?

Dave Watts from Fig Leaf Software (www.figleaf.com):

Google has the same kind of approach, using their OneBox functionality. With
a GSA, you can write and upload OneBox modules, which let you integrate
custom application functionality into your standard search interface. For
examples of this, you can do searches on the public Google interface like
this:

movie: xxxxx [zipcode]

and you'll find movie theaters in or near my area code.

I'm working on OneBox modules right now for things like employee directory
search for our Google enterprise clients. This doesn't really offload
anything from your database in many cases, it just gives the user a quick
way to get to data without having to go to the standard interface for that
data. It's quite limited, and often you'd want your OneBox to give you links
to the actual system, rather than trying to fetch all the results you need
to see, or whatever.

But Google is very big on the idea of search being the user's primary
interface to a lot of things which don't normally fall into the common
conception of search. In a way, search becomes analogous to a command prompt
(!), where the user's ability to remember short key strings lets him do lots
of operations without going through the normal GUI. In some cases, the user
might not have to know anything at all; if you can figure out from the data
pattern what OneBox you want to invoke, that's ideal. For example, if you
type your 10-digit US phone number (with or without dashes) into Google, and
your number is publicly listed, you'll see your phonebook listing, and a
link to your address on Google Maps.

Rob Munn:

We're actually moving away from the Google solution. I like what Google does, it just doesn't do some things that we need in the ecommerce arena. The best thing about Google is that the appliances are damn near toasters- plug them in, configure IPs, etc., add URLs and other configuration data, and they just go. FAST is more like a Ferrari- it has 600 hp and more torque than you can believe, but it takes a small team of engineers to keep it running.

Brian Meloche:

For what it's worth, my company bought a Google appliance.  It sits there, basically unused, as our infrastructure team's had all sorts of issues getting it to work with our systems.  They might actually put it up on eBay.  I am hoping that they won't, as it may come in handy during our new intranet project.  I don't know if they've ever tried to use the search against the JDE backend to see what kind of results/performance we get.  I haven't been involved with any of the R&D efforts.

Robi  Sen:

Enterprise search can be huge.  I have been working lately with some organizations that have rather large systems and database and are trying to make access to structured and unstructured data not only easier for users but also trying to figure out what data users need to be aware of before they ask.  A lot of enterprise search now is not simply just indexing and retrieval of data but also machine learning via concepts like Bayesian analysis, link analysis, etc.  I think something I am seeing a lot where I play is not only the concept of being able to search for information I want in a reactive mode but tools that not only guess what else might be useful now but also alert me to data or information that I might need to be more efficient at what I am doing before I ask.  A good place to look into how search is being used in more novel ways is Carnegie Mellon University who has a top flight machine learning program.

Rob Brooks-Bilson:

Like Dave W, we're using the GSA here.  I know you had it, but decided to move away for various reasons.  Google's OneBox and Feeds approach offers a lot more functionality than was there a year ago.  Add to that adapters for various off-the-shelf CMS systems like SharePoint, and the GSA has a lot more to offer.  In our case, it's currently fitting the sweet spot.

Search is huge for the next generation of enterprise IT. The average enterprise today is just scratching the surface when it comes to exposing useful data inside the enterprise.  Various vendors have developed different strategies for dealing with the enterprise data question. Google has leveraged their expertise in Internet search algorithms to create dedicated appliances with additional features to connect to enterprise data repositories. FAST, Autonomy, Endeca, Verity and other players have developed enterprise software packages with features aimed squarely at reaching data embedded in proprietary systems like ERP and CRM packages from Oracle and SAP, among others.

According to Gartner Group, FAST, Endeca, and Autonomy are solidly out in the lead in the enterprise search space. (By way of disclosure, I have experience implementing FAST, Google, and Verity- although not with the newest Verity solutions). Google, of course, has more money than all of the other players combined and could potentially leverage this advantage into a killer app for enterprise search, though none as yet has emerged. As Robi Sen notes, search can go way beyond pure index and retrieval to provide more intelligence for users. Google clearly understands this power on the public Internet, as is reflected in their strategy to record more and more information about users in an attempt to provide more added value for Internet searches. You don’t have to think too far ahead to imagine that Google might just be planning to add such features to their enterprise search offering in the near future.

The other players are hardly sitting still. FAST already has enough power and flexibility to allow a creative architect to build virtually any solution that can be imagined, though the tools are more complex than the less feature-rich Google offering. Autonomy, Endeca, and Verity lean in the direction of FAST. There are also a host of other search appliance players in the market. I have included only Google here because of their market presence in public Internet search and their ability, like Microsoft, to enter virtually any market without regard to profitability.