Specialized handheld devices deliver situation-specific advice on demand

Expert systems—applications that mimic the human decision-making process—have proved their worth in many situations, from IT help desks to online sales and process control.

One promising area for expert systems is field service, but as of late, a lack of mobility has inhibited use, with field-service workers often constrained from taking a laptop computer to the work site.

But a software vendor called Exsys claims to have solved that problem by making its expert systems compatible with handheld devices. "Laptops can't be used everywhere," says Dustin Huntington, president. "Many organizations require the use of small, compact devices for expertise access and retrieval."

Exsys offers its Corvid 3.4 package on Hewlett-Packard iPAQ pocket PCs. The solution actually is a tool kit used to develop expert systems that help workers diagnose equipment problems, and control production processes.

Differences between traditional and handheld expert systems are slight. With handhelds, the screen size is smaller, so no scrolling is required; and image maps with hotspots can be used instead of explanations, Huntington says.

Whether on the Web or handhelds, Exsys expert systems use a Java applet for system deployment across platforms and browsers, and a Corvid Inference Engine that generates interface screens dynamically in HTML.

Exsys offers a three-day training class to give users the tools they need to create their own expert systems. "A group of experts, who collectively have one hundred years experience, collaborate to build all these rules," says Huntington.

Rules can be applied via decision trees, which follow each branch of logic to the root of the problem; or probabilistically, which looks at the various characteristics of a defect.

Rules are defined as "if, then" statements, such as "If the power light is on, then the machine is getting power." Then, Huntington explains, "The inference engines takes these rules and use them in a logical manner. If you say the power light is on, it won't ask if the machine is powered."

With the rules defined, users can ask the system a question, and in return, the system will pose multiple-choice questions to determine the solution. "The interaction is dynamic," Huntington explains. "It's the same series of questions you'd expect if you were talking to a human."

Apart from eliminating errors and increasing efficiency, an expert system can capture the knowledge of a company's best operators, which allows engineers—no matter their level of expertise—to do the work of an expert.

As Huntington says, "A little knowledge can be worth a lot in the right place." And with a handheld, the right place is right where you are.