Using expert system technology to foster innovation.

Using Expert System Technology to Foster Innovation

Introduction

In today's highly competitive environment, there is an increasing need for organizations to foster innovation. Consider, for example, American Hospital Supply; it was the first to install on-line order entry terminals in hospitals. As a result, it now dominates the medical supply industry. This is just one of several cases where organizations have used technology for competitive advantage [5]. Expert system technology has the potential for satisfying many needs in management today. In particular, expert system technology can be used to foster innovation.

The objective of this article is to illustrate how expert system technology can be used to support, implement, and stimulate innovation. The article presents a brief introduction to expert system technology and the benefits of combining Lotus worksheets with expert systems. This provides the groundwork for a case study in supporting innovation. To illustrate the implementation of innovations, American Express' Authorization Advisor and Coopers & Lybrand's Exper-Tax are reviewed. The article concludes with a discussion of how expert system technology can be used to stimulate innovation in a business planning environment.

Expert Systems Technology

Since the mid-1970s, specialized computer programs have been developed to aid physicians in treating infections, geologists in locating mineral deposits, and engineers in configuring computers. Because these computer programs solved tasks performed by experts, they have been called expert systems. The generalized techniques employed by these computer programs have been termed expert system technology. Several vendors have developed software packages which utilize this technology. These packages are called expert system tools.

These system tools make it possible to develop an expert system more quickly and inexpensive than if general purpose languages, such as COBOL or FORTRAN, are used [3]. Most expert system tools have a generic structure consisting of at least two parts - the knowledge base and the inference engine. The knowledge base is similar to a database. In conventional programming, a database contains records with values of data items, which a program uses to compute its output, for example, a payroll. In a similar fashion, an expert system tool uses a knowledge base which contains explicit rules and facts. Just as a conventional program can easily process a different database and compute the payroll for another organization, so too an expert system tool can process a different knowledge base and conduct a different consultation. The execution of a knowledge base is called a consultation, because historically an expert system mimicked a consultation with an expert.

The second component of an expert system tool is the inference engine. The inference engine is a computer program which performs two major functions. First, it controls the order in which the rules in the knowledge base are executed. Second, it examines existing rules and facts and infers new facts when possible. These capabilities give expert system tools great flexibility compared to conventional programming languages for solving certain tasks.

Lotus Combined with an Expert

System

There are numerous advantages to combining the capabilities of a computerized spreadsheet like Lotus with expert system technology to facilitate portfolio investment analysis. Lotus provides a means of storing large amounts of data and of performing quantitative computations. Expert system technology supplies the means of storing intuitive rules and performing qualitative computations. Some expert system tools provide the facility to interface with Lotus worksheets. With such an interface, many benefits can be realized.

The first benefit is training. For the end user who previously performed financial (quantitative) analyses using Lotus, little additional training is required. The user can continue to employ Lotus to store all necessary data (both quantitative and qualitative). Commands to execute the knowledge base can be put in a batch file. All the user needs to do is enter the name of the batch file; no training in the use of the expert system tool is required.

A second benefit is more flexible reporting. Some expert system tools offer an interface with data base management files like dBASE. With such an interface, the results of a consultation with the knowledge base can be written to a dBASE file and subsequently reported by a dBASE program.

A third benefit is that data from more than one worksheet can be accessed during a consultation. This capability becomes more significant as the size of a system increases. For example, data concerning general industry characteristics can be kept in one worksheet and the specific data on each investment candidate can be kept in another.

A Case Study

The case considers a hypothetical investment firm which consists of a group of portfolio managers. Each manager controls a portfolio in a particular industry or industry segment. The investment firm's management desires that each portfolio manager follows the overall strategy of the firm. The case illustrates how this can be accomplished with expert system technology.

For the sake of brevity, this case study does not contain all factors that would be included in an unabridged investment analysis. However, it does contain sufficient quantitative and qualitative data from both the user and Lotus worksheet to illustrate the power of integrating expert system technology with Lotus worksheets. If desired, it is a straight forward process to extend the analysis to include additional factors.

For this case study, the overall evaluation of the investment candidates is limited to industry factors common to all candidates, managerial factors specific to each candidate, and financial factors specific to each candidate. At the next level of detail, the factors relating to the candidates' industry are limited to the product life cycle, competition, and labor. The managerial factors are limited to attention to customer needs, attention to the creative potential of each employee, policy on rewarding innovation, and administrative staff and organizational structure. The financial factors will be limited to the internal growth rate, earnings per share, return on total assets, and return on equity. A Lotus worksheet is used to store the qualitative managerial factors and the quantitative financial factors. See Table 1 on page 35 for the evaluation of four illustrative investment candidates from the semi-conductor industry. [Tabular Data Omitted]

The fictitious investment firm has a procedure to determine the suitability of each investment candidate at one of four possible levels: excellent, good, acceptable, and unacceptable. This procedure consists of four sets of rules. At the highest level, there are rules which relate the industrial, managerial, and financial factors to the suitability. One such rule is:

If the candidate's industry factors are

excellent, the managerial factors are excellent,

and the financial factors are excellent,

then the suitability of the candidate is

excellent.

Table 2 on page 38 contains a partial listing of the actual Portfolio Manager knowledge base, which was implemented with the expert system tool, VP-Expert. The above rule, which determines the excellent candidates, is encoded in Table 2 as rule A1. All the other suitability rules are included in Decision Table A of Table 2.

In order to determine the industrial factors, there is another set of lower level rules which relate the product life cycle, competition within the industry, and the labor climate to the industry factors. A typical rule, a part of rule B2 in Table 2, is:

If the primary product of the candidates is in

the growth stage, the competition in the

industry is many small producers of equal

size, and the labor climate is good, then the

industry factors are good.

The managerial factors are evaluated by another set of rules, which relate a candidate's attention to customer needs, attention to an employee's creative potential, policy on rewarding innovation and administrative staff, and organizational structure to the managerial factors. Although the rules for the industrial and managerial factors are entirely qualitative, both are included in this abbreviated case study because the industrial factors represent a variable which is the same for all candidates, whereas the managerial factors vary for each candidate. In an unbridged analysis, there are usually one or more variables of each type. An example of a rule for the managerial factors, a segment of rule C2 of Table 2, is:

If a candidate's attention to customer needs

is exceptional, its attention to employee

creative potential is exceptional, its policy

on rewarding innovation is sometimes, and

its administrative structure is lean and

simple, then the candidate's managerial

factors are good.

The last set of rules rate the quantitative financial factors. This assessment is based on the candidate's internal growth rate, earnings per share, return on assets, and return on equity. An example of a representative rule is:

If a candidate's internal growth rate is more

than ([is greater than]) twice the minimum required

internal growth rate, its earnings per share is

more than ([is greater than]) twice the minimum required

earnings per share, its return on assets are

more than ([is greater than]) three times the minimum

required return on assets, and its return on

equity is more than ([is greater than]) three times the minimum

required return on equity, then the

candidate's financial factors are excellent.

Thus expert system technology can be used to implement in a straightforward manner the procedure to evaluate investment candidates.

To demonstrate how expert system technology can be used to support innovations, assume that due to changes in the investment firm's strategy, it is desired to restrict good candidates to those that have only excellent industry factors. Rule A2 in Table 2 determines the good candidates. To make that rule more restrictive, the "industry factors = good" clause is removed from Rule A2. Once this change is implemented and the copies distributed to the portfolio managers, this new strategy will be followed by all the managers. Thus, the entire group of portfolio managers can apply their innovative skills to selecting investment candidates knowing that the analysis performed by the computer system is consistent with the overall strategy of the firm.

Some Applied Examples

Expert system technology can be used to implement innovations which would be too costly to develop and maintain with conventional technology. Consider, for example, the Authorization Advisor developed for the Travel Related Services division of American Express. This computer system uses expert system technology to condense the six inch thick Authorization Manual and the expertise of their best authorizers. The result is a less stressful job for the operator (previously, the operator had to study 16 screens of data in less than 90 seconds) and improved consistency - 24 hours a day, 7 days a week, around the world [1]. In addition, American Express can maintain its innovative policy of not imposing any credit limit on individual purchases. This is a strategic advantage over the other credit card firms which impose a monthly limit.

As another example, consider Coopers & Lybrand's corporate tax planning package, Exper-Tax. The software runs on PC-compatibles and is used by Coopers' auditors at over 1,000 U.S. corporate client sites [2,4]. Because Exper-Tax utilizes expert system technology, Coopers can update the system in a timely fashion whenever there is a change in the tax code and/or Cooper's tax strategy. This allows Coopers to leverage their tax expertise in an innovative fashion.

As the above two examples illustrate, once such systems are in place, it becomes possible to use them to implement innovations quickly. If a revised procedure is developed for credit authorization, American Express can incorporate it into the Authorization Advisor and have it used by all their operators. If Coopers & Lybrand develops a new strategy for corporate tax planning, it can be integrated with Exper-Tax and used by Coopers' auditors at all client sites.

Conclusion

Computer systems similar to the portfolio analysis system described above can be used by many organizations to stimulate innovation. Consider a planning department in any large organization. When developing new products, the business planners must be continually updated on changes in governmental regulations, products offered by the competition, and/or internal parameters, such as manufacturing capabilities and marketing strategies. Keeping conventional hard copy manuals up-to-date is inefficient. Maintaining conventional computer programs is costly. On the other hand, using expert system technology and Lotus worksheets together can provide a cost effective means to keep business planners informed. With such a system the planners can apply their innovative skills to new product development knowing that the analysis they perform with the system is consistent with the latest overall strategy of the organization.

Table : Table 2 Portfolio Manager Knowledge Base

Decision Table A: Suitability

This table determines the suitability of a particular candidate based on an abbreviated analysis of the industry factors, managerial factors, and financial factors. The values for these variables are found in Decision Tables B, C, and D respectively. The four possible values for suitability are excellent, good, acceptable, or unacceptable.

Rule A1

If   industry_factors     = excellent and
     managerial_factors   = excellent and
     financial_factors    = excellent
then suitability          = excellent;

Rule A2

If   industry_factors     = excellent or
     industry_factors     = good and
     managerial_factors   = excellent and
     financial_factors    = excellent or
     financial_factors    = good
then suitability          = good;

Rule A3

If   industry_factors     = excellent or
     industry_factors     = good or
     industry_factors     = fair and
     managerial_factors   = excellent or
     managerial_factors   = good and
     financial_factors    = excellent or
     financial_factors    = good
then suitability          = acceptable;

Rule A4

If   industry_factors     = excellent or
     industry_factors     = good and
     managerial_factors   = excellent or
     managerial_factors   = good and
     financial_factors    = excellent or
     financial_factors    = good or
     financial_factors    = fair
then suitability          = acceptable
else suitability          = unacceptable.

Decision Table B: Industry Factors

This decision table determines the industry factors which affect the candidates' industry. The industry factors are limited to: product life cycle, competition, and labor. The values of these factors are the input by the user.

Rule B1

If   product_life_cycle   = growth and
     competition          = large_and_small and
     labor                = excellent
then industry_factors     = excellent;

Rule B2

If   product_life_cycle   = growth and
     competition          = large_and_small or
     competition          = small_only and
     labor                = excellent or
     labor                = good
then industry_factors     = good;

Rule B3

If   product_life_cycle   = introduction and
     competition          = large_and_small and
     labor                = excellent
then industry_factors     = good;

Rule B4

If   product_life_cycle   = maturity and
     labor                = excellent
then industry_factors     = fair
else industry_factors     = poor.

Decision Table C: Managerial Factors

This decision table evaluates the managerial factors for each candidate in terms of its attention to customer needs, its attention to an employee's creative potential, its policy on rewarding innovation, and its administrative staff and organizational structure. The values of these factors are read from the Lotus worksheet.

Rule C1

If   attn_cust_needs_wks[i]        = exceptional and
     attn_creat_potential_wks[i]   = exceptional and
     reward_innovation_wks[i]      = always and
     admin_structure_wks[i]        = lean_simple
then managerial_factors            = excellent;

Rule C2

If   attn_cust_needs_wks[i]        = exceptional and
     attn_creat_potential_wks[i]   = exceptional and
     reward_innovation_wks[i]      = always or
     reward_innovation_wks[i]      = sometimes and
     admin_structure_wks[i]        = lean_simple or
     admin_structure_wks[i]        = lean_complex
then managerial_factors            = good;

Rule C3

If   attn_cust_needs_wks[i]        = exceptional or
     attn_cust_needs_wks[i]        = moderate and
     attn_creat_potential_wks[i]   = exceptional or
     attn_creat_potential_wks[i]   = moderate and
     reward_innovation_wks[i]      = always and
     admin_structure_wks[i]        = lean_simple
then managerial_factors            = good;

Rule C4

If   attn_cust_needs_wks[i]        = exceptional or
     attn_cust_needs_wks[i]        = moderate and
     attn_creat_potential_wks[i]   = exceptional or
     attn_creat_potential_wks[i]   = moderate and
     reward_innovation_wks[i]      = always or
     reward_innovation_wks[i]      = sometimes and
     admin_structure_wks[i]        = lean_simple or
     admin_structure_wks[i]        = lean_complex
then managerial_factors            = fair
else managerial_factors            = poor.

Decision Table D. Financial Factors

This decision table focuses on the financial factors of the internal growth rate, earnings per share, return on assets, and return on equity. The values of these factors are read from the Lotus worksheet. The minimum acceptable values are obtained directly from the user. [Tabular Data Omitted]

PHOTO : Patrick J. Lyons is Associate Professor of Management at St. John's University in Jamaica, New York.

PHOTO : Anthony Fabiano is an MBA student majoring in Finance at St. John's University, New York.

Faculty Notes

Editor's Note: If you would like to contact faculty members about any of the activities reported below, please direct your inquiries to the Business Research Institute.

Jack Marshall (Economics and Finance) presented a two day seminar on swap finance for the New York Institute of Finance. He has a new book entitled Swaps and Related Risk Management Instruments, which was released in August 1990. He will present a paper entitled "A Multiperiod Portfolio Model" at the Financial Management Association Meetings in October 1990 and a tutorial on swaps at the Southern Finance Meetings in November 1990.

Jack Marshall (Economics and Finance) and Vipul K. Bansal (Economics and Finance) were awarded three grants for their research on financial engineering. The grants are from the Chicago Board of Trade, Bear Sterns & Company, and the Continental Bank of Illinois.

Lawrence Deckinger (Marketing) and Herbert Katzenstein (Marketing) were the co-authors with Louis Primavera (Psychology) and James Brink of Grey Advertising in New York City of a paper entitled "How Can Advertising Teachers Better Prepare Students for Entry Level Advertising Agency Jobs," which appeared in the December 1989/January 1990 issue of the Journal of Advertising Research. They were the co-authors with Rita Dunn (Education) and Pam Withers from the College of Mount St. Vincent of a paper entitled "Should College Students Be Taught How to Do Homework," which will appear in the Winter 1990 issue of Illinois School Research and Development.

Patrick J. Lyons (Management) had a paper entitled "PC-based AI Environments" published in the May 15, 1990 issue of Systems AI.

References

1. Bowerman, R. and D. Glover. Putting Expert Systems into Practice. New York: Van Nostrand Reinhold, 1988.

2. Clark, M.J. "What Every Wall Street CEO Should Know About AI." Wall Street Computer Review, June 1988, p. 27-40.

3. Duchessi, P., H. Shawky, and J.P. Seagle. "A Knowledge-Engineered System for Commercial Loan Decisions." Financial Management, Autumn 1988, pp. 57-64.

4. Leonard-Barton, D. and J.J. Sviokla. "Putting Expert Systems to Work." Harvard Business Review, March-April 1988, p. 92.

5. Rackoff, N., C. Wiseman, and W. Ullrich. "Information Systems for Competitive Advantage: Implementation of a Planning Process." MIS Quarterly, Vol. 9, No. 4, December 1985.