Since your staff is not going to be expert at every question that comes their way, giving them some help is important.
Most helpdesks start with a trouble tracking system. This software lets incidents be entered and move through states over time -- call open, call assigned to field engineer, call being worked on, call resolved, call escalated to senior engineer, call reassigned to different engineer, etc. These systems do not require any AI, as they are essentially databases with time triggers.
However, most of these systems ask that problem description and problem resolution information be entered into the database. From there, it seems a logical next step would be to leverage all that typing and use the information as a primitive
knowledge base. If the knowledge base and the workflow automation are be combined together, voila! An expert system for your helpdesk so they can walk on water -- at least as far as your customers are concerned. Unfortunately once the problem-resolution overload issue is put to rest, the staff jumps to the next problem:(For an amusing look at the sorts of questions helpdesks get, take a look at
Computer Humor - True Helpdesk and Technical Support Stories.)Unfortunately, while the users are all different, the questions called in seldom are. Wouldn't it be nice if one could have a non-human handle the bulk of these questions, saving the interesting or unusual things for the people?
This happy thought has occurred to more people than just you. In fact, the industry that sprung up around providing expert systems for helpdesk staff is now branching out into
expert systems for end users. This market has become more than just automating existing helpdesks, however. In some cases, the software is a means to provide support and services where before none existed -- at least for ordinary mortals sitting at home in front of their computers. (Happy electronic 'self-service' customers [PCWeek])In our research in and around the topic, we found some nice general-purpose resources in addition to all the helpful things the vendors provide. The white paper
Choosing Troubleshooting Technology has an excellent discussion of how to decide among four different knowledge representation approaches (Text Retrieval, Decision Trees, Case-Based Retrieval, and Augmented Fault Models) when designing a knowledge base. The case study described in the Knowledge Base White Paper outlines potential trials and tribulations from someone who has been there. Unfortunately, he seems to be one of the few currently sharing their tales, as he himself laments in his introduction.Naturally, no section discussing "how-tos" would be complete without giving a look at what one can aspire to! These demos are good examples of end-user helpdesk AI. Some are simply knowledge bases made available over the web, but most offer some sort of natural language processing in their user interface, with a variety of tools inhabiting the back end.
Broderbund - Online Technical Support
LucasArts Tech Support (yoda's helpdesk)
ServiceWare Inc. Demos
Schwab.com Technical Support
Inference's Self-Service Help Desk
AIM Helpdesk Demo (a less-popular vendor)
Duke OIT Help Desk (This website has won several awards, and was begun by the same Philip Verghis who maintains the Helpdesk FAQ.)
Search the NC State Help Database
Support Online from Microsoft Technical Support
Helpdesk applications is a hot and rapidly changing field. Not only are new products coming out all the time, but old competitors keep swallowing each other up. To truly see how big the field has become, take a look at
Philip Verghis' Help Desk FAQ page for software vendors. This is a better resource than ours, as he will maintain his. This, after all, is a snapshot taken in mid-August 1998, and we haven't actually used any of the tools.Disclaimers out of the way, the common names that keep cropping up in our readings are:
Network Associates, Inc
SupportMagic SQL, WebMagic, McAfee HelpDesk ; Bendata Inc. HEAT™ (HelpDesk Expert Automation Tool) ; DSSI Inc. Professional Help Desk ; Remedy Corp. Action Request System ; Inference CBR Content Navigator ; Platinum Apriori ; ServiceSoft Corp. Knowledge Builder, Web AdvisorAnother market is for knowledge bases that snap in to the products listed above. These providers include:
Microsoft TechNet ; ServiceWare Inc. Knowledge Pak ; Knowledge Broker KnowledgeBasesHere's an at-a-glance view of some of the products, cross-matched to the areas that they support. "AI-ness" refers to whether or not there are any expert system tools embedded in the software, like case-based reasoning or natural language searching.
Help-Desk Applications
(top)|
|
(return) | (return) |
"AI-ness" |
|
Y |
Y |
Y |
|
|
Y |
Y |
|
|
|
Y |
|
|
|
|
Y |
Y |
Y |
|
|
Y |
Y |
|
|
|
Y |
|
Y |
|
|
Y |
Y |
Y |
|
|
Y |
|
Y |
|
|
Y |
|
Y |
|
|
Y |
N |
Y |
No discussion of products would be complete without a bunch of clippings. In fact, that's most of what we found in our search -- lots and lots and lots of clippings. Amazingly, we didn't save but the two ServiceSoft reviews. To yield your own fruitful search, try the electronic presences for Computerworld, ZDNet (PC Week, PC Magazine), and InfoWorld and search on "helpdesk." It's that easy!
This reference (
Web-based customer support resources) is actually a list of related Computerworld articles. Another helpful Computerworld compendium is Help desk resources on the World Wide Web. An old article (1997) that offers a matrix for different software features, help desk software features [Federal Computer Week] may be superceded by a more recent PC Magazine Shootout.(
I'd like to continue reading the main page, please)
Expert Systems and the Helpdesk | TCO and the Helpdesk | Top
page generated for
"AI and Expert Systems" LIS 385T.15 at the University of Texas at Austin on August 13, 1998 by Jeff Steely, Kimbol Soques, and G. Marc Turner