Printer Friendly

Can your business use artificial intelligence?

Can Your Business Use Artificial Intelligence?

If it can, then you'll have a problem-solving tool for leveraging expertise

Artificial intelligence (AI) can be thought of as the art of making computers do things that would require intelligence or judgment if done by humans. For example, computers approve American Express purchases, diagnose diesel locomotive faults for GE, configure VAXs for Digital, advise on tax matters for Arthur Andersen, schedule production for Westinghouse, and assign rooms for Holiday Inns. Every Fortune 500 company has AI applications on-line, in development, or as the subject of a feasibility study.

Pattern recognition, computer vision, and robotics all employ AI, as do computer programs for chess playing, automated reasoning, and problem solving. Natural language applications of AI range from low-level spelling and grammar checkers to speech recognition and machine translation. There's now an English to Spanish translating application on the market for microcomputers.

AI APPLICATIONS IN BUSINESS

The chief business use of AI is in expert systems, which assist human experts in solving difficult problems. All of the specific examples just mentioned are expert systems. Expert systems embody the knowledge and reasoning of human experts. They aid managers in understanding their businesses and in establishing control of business processes. In addition, they secure valuable intellectual resources for the company and enforce uniform application of policy.

When Campbell Soup's chief engineer was approaching retirement age, his knowledge and experience in keeping the huge cookers running were captured in an expert system and made available to younger engineers. The same scenario was repeated at GE, resulting in the expert system Delta for repairing diesel locomotives. In both of these cases, an important resource was not allowed to retire.

Typically, a firm has many employees upon whose expertise the firm depends. Even the smallest company will depend upon the expertise of at least one person. The very process of making this expert knowledge explicit in such a way that it can be incorporated into an expert system is itself one of the greatest benefits of developing an expert system. During this process, one can become aware of many aspects of the business that had been only vaguely understood previously. This increased understanding can lead to greater control over the business.

The firm of Peat, Marwick, Main (PMM) has implemented an expert system to help their various offices evaluate loan portfolios of lending institutions. PMM had noticed that expertise in evaluating loans varied greatly from office to office, and in some cases the evaluation was even being performed by partners because of the lack of expertise at lower levels. Their problem was to take the expertise of those few persons known as expert evaluators of loan portfolios and make it available, at least in some meaningful part, to employees throughout the company. PMM's success in doing so had the added benefit of establishing uniform standards for loan evaluations at their various offices so there were no longer any significant differences between the way a loan might be evaluated in Minneapolis and the way a similar loan would be evaluated in Memphis.

As a side benefit after a couple of years, employees have tended to consult the system much less frequently, presumably because they have absorbed much of the expertise embodied within the system. In other words, the expert system is proving to be a long-term training tool for employees.

COMPETITIVE ASPECTS OF EXPERT SYSTEMS

The usual considerations of maximizing desirability and feasibility apply equally well to expert systems concerning traditional computer applications. A more distinctive issue is to decide which tasks are appropriate for expert systems and which ones are better handled by other methods. Researchers G. Gorry and M. Scott-Morton proposed a framework for decisions that can be useful in selecting those that can benefit most from expertise, whether human or artificial.

The table on page 11 shows how decisions or tasks can be classified as structured, semi-structured, and unstructured (on the vertical axis) and how levels of managerial action can be analyzed in terms of operational control, management control, and strategic planning (on the horizontal axis). Structured tasks (the first row), of course, usually require very little expertise and can be handled with traditional algorithmic solutions, whereas unstructured tasks (the third row) may very well resist the imposition of any structure, including that of an expert system. Semi-structured tasks seem to be the ones most typically chosen for expert system development.

Some of the classic successes in expert systems have been semi-structured tasks at the level of operational control, such as Digital's XCON system for configuring computer systems. Many expert systems, however, have been developed to aid in decisions at the managerial and strategic levels; and the competitive potential of applications at these levels is, of course, even greater than that of applications at the operational level. Yet there is no reason why even unstructured tasks cannot benefit from expert systems provided that there are human experts who can explain to others what they do.

Finally, potential expert system applications can be evaluated in terms of their position in the value chain (i.e., within a business, the arrangement of tasks that add value). As with other types of technology, the relevant question is whether expert systems are being retrofitted to low value-added applications or being used to exploit opportunities further up the value-added axis.

EXPERT SYSTEM LIFE CYCLE

A variety of issues emerge in the development stage of expert systems. Prototyping, however, is the most important expert system development tool and typically consumes a large share of resources. The two parts of an expert system--the knowledge base (the representation of expert knowledge) and the shell (the remainder of the system)--require separate development and maintenance skills.

It is relatively easy to build an impressive prototype of an expert system in a fairly short time. Completing the system, however, typically takes much longer.

Finally, expert system projects are often seen as risky, since the technology is relatively new and the applications tend to address unstructured problems. This risk needs to be considered in any commitment to development of an AI application. Such risk can be minimized by seeking the counsel of experienced AI consultants.

GETTING STARTED WITH AI

The way to start an AI project is the same as with any other information system application. If a firm is large enough, it will have information system specialists. If not, the firm will probably depend upon an outside consultant for information systems needs. In either case, the initial steps are the same and are the traditional ones--conduct a needs analysis and a feasibility study together with a careful evaluation of costs and benefits. It is important to choose an application that is suited for an expert system--one that will provide cost savings large enough to justify the investment and its risk, or one that gives a firm a strategic advantage in the marketplace.

The evaluation of costs and benefits is particularly important in the case of expert systems, since their relative riskiness and their typically lengthy development time require an unusually careful analysis of expected future returns. The analysis and design of expert systems can depend upon the skills of a knowledge engineer--someone with experience in helping persons with expertise in a firm to articulate that expertise in such a way that it can be incorporated into an expert system. A company's personnel can and should develop the skills of knowledge engineers over time by starting with small, manageable projects. It is more feasible in most cases, however, to call upon consultants who already have such experience.

RISKS AND REWARDS

AI and expert systems provide a wide range of opportunities for automation beyond those of more restricted, traditional computer applications. Yet the risks and applications of new technologies to problems that are difficult to define may deter one from taking advantage of them. At the same time, a wait-and-see attitude may be even more risky, since AI and expert systems can give one's competition an advantage. As with any other project, careful analysis and planning usually translate into tangible benefits.

Broadly speaking, information and expertise are assets of the corporation. Expert systems can be a tool to help solve stubborn problems, a tool for leveraging expertise. Those firms that implement expert systems in their operations, as well as at the managerial and strategic levels, may be able to move ahead of their competition.

Above all, the very experience with AI and expert systems can be leveraged. Those who can invest in such potentially valuable technology are positioned to reap the greatest benefits.

Mr. Byrne is a doctoral candidate in accounting at Memphis State University. Dr. Franklin is a professor in the Mathematical Sciences department and a member of the Institute for Intelligent Systems at MSU.
COPYRIGHT 1990 University of Memphis
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1990 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Byrne, Paul J.; Franklin, Stanley P.
Publication:Business Perspectives
Date:Mar 22, 1990
Words:1456
Previous Article:Computer integrated manufacturing.
Next Article:Telecommunication systems of tomorrow.
Topics:


Related Articles
Artificial intelligence and natural confusion.
Finding the Best Tool for the Job.
Street is not what it was.
Two new journalS from Hindawi Corp.
Business Applications and Computational Intelligence.
Researchers debate utility of autonomous armed robots.
BASF uses AI for color matching.
Business intelligence; artificial intelligence in business, industry and engineering; proceedings.
Artificial Intelligence.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters