Internet Buzz with Richard Wiggins | 16

So you'd be a sort of service bureau for them?

Yep, it's a total turnkey system. Sign up with us, and we'll deliver the results for you.

Where would the database reside?

The software would reside at the customer's site. We'd deliver working software and the question base for them install on their servers.

Does it help for the researchers to have expertise in the subject area for which you're building a question base? For instance, would you assign PC experts to build the question base for the PC maker?

Some do, yes. For one thing, [when building a system for a major PC vendor] we assigned people who happened to own that brand of computer. But that's actually not required. You have the researcher go through the site in a systematic way. It's a methodology – you actually survey the entire site to find the answers first, then identify the questions – and that applies to subjects the researcher knows very little about.

That's brings up an interesting point. The process is driven by what they actually have on the site. But doesn't that assume that the site will have answers to the questions people actually ask?

What happens is, you build the system, and then you have a testing period where we try all the questions we can think of, to try to see what's missing. Then, when we put the system live, we may find that there are a few other missing things. Maybe we're missing a synonym we need. Every day, you'll get a list of the questions that weren't answered – for which there is no answer on the site – and you can add it to the site. The system actually gets smarter over time. Over time, GateDellPaq's Web site becomes more complete, and Jeeves becomes the smartest customer support rep at GateDellPaq.

What if a company had an intranet application, and wanted to set up a team of their own researchers to build a question base for proprietary information available only behind a firewall?

We'd talk to someone who wanted to do something like that. I must say, though, that there is a bit of an art to building a question base. Our researchers get quite good at it, but it takes training and practice to get there. We might suggest that we build the first set of questions together, so that they could learn by doing.

You know, most companies are busy doing other things, and they love the fact that we offer a turnkey system. It's actually a very efficient process, and a very cost effective solution.

From what I've heard about company intranets, one of the great advantages is the ability to discover who else in a company is working on a similar problem, so that you can get together and combine efforts. For instance, a researcher at Ford might do a search on "electrophoreses" to find out who else in Ford worldwide is working on the same thing. How might AskJeeves help there?

We see Jeeves in general, as well as company-specific applications of Jeeves, as very much an "80-20 rule" kind of thing. We use humans to identify the 80% of questions that people are asking, and in so doing provide a quick path to the answers to those questions. The final 20% of questions people ask, or the final 2% of questions people ask, probably aren't going to be handled by Jeeves. It isn't worth the effort involved in having human researcher locate the answer and add the question to the database.

Does this mean that there's a need for search engines after all?

Absolutely. For very specific kinds of questions, involving very arcane areas of knowledge, or very specialized search terms, users may find it necessary at times to use search engines. That's true on the Internet in general, and it's true for company-specific or intranet applications.