STRA TXGY DEVELOPING EXPERT SYSTEMS FOR SMALL BUSINESS: AN APPLICATION FOR SELECTING A LEGAL FORM OF ORGANIZATION Ronald S. Rubin James M. Ragusa University of Central Florida ABSTRACT The success and iaiue ro citents of small business counseling programs such as those offered by Stnail Business Dei elopment Centers (SBDCs) and Smuii Business Insritutes (SBIs) are directly related to the availability and qua(in of expert advice. This article prthides an overview and erpiores the possibilities for PC-based erpert sysrem development with pumicular at ternion ro u stnuii business application. A proof of concept SBDC advisory system is described. It provides expert-based advice for selecting a legal form of business organizarioti. Issues of ciietit usability and liability are raised. Conclusions show that tire use of expert systems for sinai) business consulting offers an erpanded potential for relieving already overworked consult- ing staff tnembers, for offering clients quaiiry advice, and for providing a merhod of trainitig less e.rperienced consultants. INTRODUCTION The rapid acceptance of a branch of artificial intelligence (Al) known as expert systems has recently emerged from the realm of basic research into real-world business applications (Feigenbaum. McCorduck, & Nii, 1988). Artificial intelligence focuses on producing computer- based results that parallel the intelligent human decision process (Harmon & King, 1985; Rich & Knight, 1991; Waterman, 1986). The Al branch called expert systems is capable of emulating human decision capability with an interactive computer program (Frenzel, 1987; Liebowitz, 1988). Expert systems pro- grams integrate existing data and capture the essence of the experiences and expert rules-of-thumb (heuristics) in a field of specialized knowledge (Mockler & Dologite, 1 992), performing symbolic and numeric reasoning and solving a set of interrelated, semi-structured problems (Rangaswamy, Burke, Wind, & Eliashberg, 1987). Expert system technology offers a relatively new way to organize data and retrieve infor- mation for decision making by allowing a computer program to handle complex reasoning (Guterl, 1986; prerau, 1990). Such systems will provide the opportunity for sharing the recog- nized expert's knowledge and decision-making skills in a given field with novices and other individuals who possess less expertise (Liebowitz, 1988). Numerous expert systems have already 51 been developed for usc in business situations. To name a few: Mentzer and Gandhi (1992) have reported on marketing systems; Heath-Brandt, Carter, and Yozie (1991), MCI's pricing for potential customers; Springer, Buta, and Wolf (199I), automatic letter composition for customer service; McCann, Tadlaqui, and Gallagher (1990), creating retailer advertisements; and Ebersold (1991), meeting a compctitivc challenge in frequent flyer competition. With the prolil'aration of microcomputers and programs for designing expert systems applications, this technology is within the reach of client counseling programs and small businesses. Three works addressed the small business environment and the use of microcomputer- based expert systems. First, a case study by Sullivan and Shively (1989) identified several small business uses of expert systems, namely, serving as a staff assistant and owner surrogate, reducing owner role conflicts (manager of operations versus consulting expen), improving self-management, and providing product/service marketing support. In a second study Martin, Jones. McWilliams, and Nabors (1991) described the development of an PC-based expert system which provided Small Business Development Center (SBDC) counselors with client entrepreneurial trait assessments and financial statements to help explain financial and market reality data. In a third study Muller-Boling and Krichoff (1991) discussed shortcomings of expert systems over human decision making but concluded that these systems have value in small business start-up consultation activities. In order to actualize such potential. however, it is necessary to understand what expert systems are and how they can bc developed and applied to the small business environment. Toward this goal thc purpose of this article is to provide an overview and explore the possibilities for expert system development with particular attention to selecting a small business legal form of organization as an example. Throughout this anicle reference is made to the use of this technology for SBDC application because of the proof-of-concept system developed. However, the authors believe that application of this technology to Small Business Institutes (SBIs) is equally appropriate. AN OVERVIEW OF AN EXPERT SYSTEM STRUCTURE An overview of the development of expen systems should provide a better understanding of its potential benefits in the small business environment. In the past expert systems were constructed with the use of one of several programming languages. for example, LISP, PRO- LOG, and even PASCAL and BASIC. However, because of recent commercially available'oftware "shells," the user does not have to know these languages to construct an expert system. These programs have the built-in capability to function as a translator between the internal programming language and the English-like language of the rules and questions asked the computer user. As a result, these shells are much easier to learn than programming languages and an application can be more easily and efficiently constructed. The basic framework of an expert system is presented in Figure I. An expen system can be broken down logically into three separate parts (Harmon & King, 1985; Mishkoff, Schafer, & Ramus, 1988: Waterman, 1986). The first of these is the knowledge base which contains the facts of information that are widely shared, publicly available, and generally agreed upon, and heuristics (rules-of-thumb, rules of plausible reasoning, rules for guessing) that characterize an expert's decision-making process. This section is the "expert" knowledge the program needs to solve problems. This knowledge can be encoded in various ways such as hierarchical, network, relational data bases, semaniic networks, production rules, 52 Figure 1. Structure of an expert system. User User Interface (Input-Output System) May Entail: Menus or a Natural Language Question & Answer Graphics of Icons Inference Engine Inference Control Knowledge Base Facts Rules 53 and predictive calculus, or some combination of these (Rangaswamy et al., 19g7). The most common and popular method of encoding knowledge is "rule-based" representation in which decision rules of an IF (premise)-THEN (actions) design are used for consultation and recom- mendation development. By linking (or chaining) these decision rules, this knowledge-based system can simulate the thought and decision process of an expert. The second element of an expert system is referred to as the "inference engine." It is the coded software reasoning mechanism of the system. It acts as a control structure or rule interpreter. This software code controls the search techniques and regulates ways in which the rules in the knowledge base arc to bc applied to the problem. It.sequences through the rules of the knowledge base, asks appropriate questions of the user when additional infomiation is required, and then offers advice liased on the rules and the user's input. It is possible for thc system to start its reasoning in a "forward" or "backward" mode. Forward reasoning starts with facts (premises of a rule arc first checked to determine if they are true) and then follows an inference path through onc or more levels of rules to reach a conclusion about a goal (examines thc knowledge base to Imd rules that are applicable). Conversely, backward reasoning starts with thc goal (possible solution) und follows inference paths back to the details (looks through the knowledge base to find rules that justify that goal solution). Backward reasoning works best if there are fcw goals and many details-a situation common in everyday business problem- solving. The last part of thc system is the user interface. This man- machine bridge is used to collect information using menu prompts and other means needed by thc inference engine, to display results, and to facilitate updating and modifying of the knowledge base. The nuthors'tudy of one SBDC revealed several potential applications for such a system. DEVELOPMENT OF AN EXPERT SYSTEM APPLICATION Thc SBDC serves the emerging business community with seminars and private consultations on cvcry aspect of starting and maintaining a business. The SBDC mission is to assist individuals who are either considering thc start-up of a small business or who have reached a stumbling block in thc operation of their business. In addition to the director and various clerical staffcrs, the office usually has consultants in accounting and finance, management. marketing, nnd energy managcmcnt. These consultants are degreed in their specialty and have several years experience working in their individual fields. Some own or have owned small businesses. When a business person contacts the SBDC, a brief phone interview is conducted by the rcccptionist. Any client in business for six months or less is encouraged to attend onc of the seminars conducted by thc SBDC. These are typically one-day seminars that outline each of the functional areas of a business and the specifics in each area that should be addressed by thc entrepreneur. The speakers are drawn from the local business community and cover topics from setting up an accounting system to identifying a potential market. Thc SBDC has determined that this is the m&xst eflicient method of disseminating information to the largest number of new and potential business owners. The basic seminar is I'ree or low cost but is somewhat time. limited. If the client has been in business for more thatt six months. an attempt is made to schedule individual appointments with thc appropriate consultant. The purpose of these meetings is to help the client resolve individual business problems by identifying resources in the community and determining an appropriate course of action. Assistance is offered on thc functional level in solving specific problems as well as on the strategic level in deterinining an overall focus 54 of the business. This service is free to the client; however, clients often fail to call I'or assistance until there is a situation in need of immediate attention. The SBDC does a laudable job of pt'oviding an important service to a large clientele. However, because of short staffing, consultants are often booked well in advance. If a firm is in need ol'mmediate assistance, the waiting time inay be too long. Furthermore, although some prioritizing now occurs when clients are directed to either the seminars or published materiuls, thc system in place provides only these two additional resources before clients arc put into the queue to meet a consultant. Seminars help I'ill the gap. but they are by their nature general in scope and cannot be tailored to a firm's specilic needs. Published material is another source of information, but clients often find it difficult to make a decision based on their study of thc materials. What is required is another resource that will intelligently provide customized advice without involving more consultant time. IS AN EXPERT SYSTEM THE ANSWER? Thc proposed expert system would act as a supplement to the current methods of providing information to small businesses. By using the expert system, some of the simpler consulting tasks could be designated to a non-consultant staffer. If willing and able, the client could work with the expen system without the assistance of any SBDC staffer. This would provide a preliminary answer more quickly or supplant attending all or part of the business stan-up seminar. If thc client has already reviewed published materials, the expert system could provide guidance through the areas of specific interest for that firm. While not all clients can be expected to bc totally assisted by the expen systein. it should shift the caseload of the consultants so that clients with the most pressing and complicated needs will be assisted first. With the problem identified and the application of expert systems technology considered feasible. the next step is to select an appropriate SBDC application. After discussions with the Director of the University of Central Florida's SBDC, Aloyse Polfer. it was determined that a proof-of-concept system which helps clients select a legal form of organization would be a worthwhile target application. Polfer served as the expert supported by a guide he co-authored (Polfer. Dicks. & Holland. 1990). The next step in the building process was to test the application for appropriateness and potential for success. One of the critical factors in dcterinining the potential of an expert system is appropriate problem selection. Cenain applications lend themselves to the expen systein solution morc easily than others. Two models were used m the analysis to determine the applicability of the expert system solution to the proposed problem solution: the Texas Instru- ments (Tl) Project Advisor Checklist ("Knowledge Engineering." l 989) and criteria rules from Silverman (19g7). The proposed application passed every section of the Tl instrument except the political environment test, which was not considered to be important in this situation. The proposed project also passed the Silverman criteria and showed reservations only on the success- oriented section. A concern was raised that the system would find a management champion only if it could be proven that the use of the systeins would reduce a consultant's workload. ln light of the requirements of the project, it wus dgcided to limit the scope ol this prototype to advising a client of ways to choose the correct legal structure for a business organization. This is uppropnate because that is one ol'hc areas of greatest concern for potential business owners. Additionally. choosmg a legal form ol'usiness is now addressed only as part ol'hc start-up seminar. The expert system would provide u source of information in uddition to the 55 seminar. This application choice is meant to serve simply as a possible example of how expen systems could be employed, and that this application as well as any other should be subjected to substantial testing and development before it is employed by a small business advisory group, BUILDING THE EXPERT SYSTEM The prototype system is a knowledge-based computer program which uses expert systems techniques to assist a client in determining the best legal form of business organization. The program requires an IBM-PC or compatible and was developed using Texas Instruments'ersonal Consultant EASY shell. Other alternative products in the same price range ($500 or less) are VP-Expert by WordTech Systems Inc. and EXSYS El by EXSYS, Inc. Information on these and other shells is available in Harmon, Maus, and Morressey (1988) and Wielgos and Ragusa (1991). The prototype works by backward chaining through the goal rules. In order to obtain values for the parameters in the premise of these rules. the system must test and fire (activate) several rules. The parameters in the premises of rules are assigned values from the knowledge base or by prompting the user for appropriate information. From the user's perspective, the system appears to operate in a more straightforward manner. To determine an appropriate legal form for a business, one should address many factors in satisfying a client's objectives. When a consultation begins, the client is asked a series of questions about the characteristics of his or her business. These questions generally address six key issues: control, continuity, regulation, administration, liability, and financing, although the user may not be aware of this grouping. When the necessary questions have been answered, either a form of business is recommended for sole proprietorship, corporations (both S and C), partnerships (both limited and general), or professional association, or instructions are given which will assist in determining the appropriate legal form. The prototype system consists of 66 rules and 42 parameters. Two rule groups make up the system —issue rules and goal rules. Issue rules draw conclusions about the characteristics of a client's business as related to the important considerations mentioned previously. For example, an issue rule would be used to determine if a small business has sufficient accounting, legal, and administrative capability to suppon corporate regulatory and reponing requirements. Goal rules use the conclusions of the issue rules to select an appropriate form of business and)or make other recommendations. If, for instance, the issue rules conclude that the client lacks the required administrative capacity but has other compelling reasons to become a corporation, a goal rule would suggest the appropriate type of corporation and provide instructions for improving administration. Validation (building thc product correctly) and verification (building ihe right product) testing of the prototype system was accomplished in the manner described by Grover (I982). Using this method, progressively morc difficult client scenarios identified by the expen were tested until thc system consistently behaved (u) within the established domain, and (b) in a manner approved by the original expert. Figure~ 2 and 3 illustrate several sample scenarios representative of the expert system in operation. As is shown, the client responds to u series of questions. When answered, the system displays a client conclusion containing: (a) a recommended legal form of organization, (b) a "why" rationale. and (c) a summary of client confidence factors. Thc stronger the rationale 66 and conlidence factors. the more robust is the system recommendation. Also, if assistance is needed in answering an application question. the system provides access to a help module through the Fl key. For example, if a user was not familiar with the definition of "passive income." the term would be clarified. DISCUSSION Several issues remain unresolved at the completion of this prototype or need to be resolved before there is complete implementation of the system. Because the SBDC deals with such a variety of clients, the problem of creating an approp- riate user-interface is very important. The design approach taken for this prototype assumes that the user has minimal personal computer skills and has at least a high school education. Those assumptions may not always be true and may result in a system not always as user-friendly as intended. A further problem is that PCs may not readily be available to clients since the SBDC has a limited number of units. If the system is to be implemented as part of the client assistance plan, purchase of additional systems may be required. Research time spent with the staff consultants was very limited due to their extremely busy schedules. In order to refine the system further and introduce more of the heuristics of client consultation, more time with the SBDC staff would be required. A liability issue emerges in the creation of this system since any recommendations made by the system and followed by the client may establish a liability claim against the SBDC and the system developers. If a client answers questions in a less than truthful or less than accurate manner and as a result the system makes a recommendation, is the SBDC liable? What if a client answers completely but because of unforeseen circumstances the system makes a recom- mendation that causes damage to the client? These questions of liability must be raised and resolved by the SBDC before the system is presented to the public. One possible solution is to have the clients sign release forms, stating that they must not base a decision for a legal 'orm of business strictly on the recommendation of the expert system. The client is informed that each of the issues addressed by the expert system should be discussed with an experienced lawyer, who will advise them of requirements, options, and responsibilities. CONCLUSION In addition to supplying information for use in developing the prototype system, the interviews with the SBDC director and the staff identified several benefits possible through full implementation of a small business consulting expert system. Such a system will allow clients to get help without involving the already overworked staff. Thus, consultant productivity will be improved because clients can obtain much assistance from the system. The expert system will also fil I the need for immediate assistance at times when consultants are booked to capacity. Another benefit lies in its potential to provide flexible and comprehensive quality answers (o their questions. Although SBDC seminars and published materials are excellent, they may not be convenient, or they may not allow a client to locate answers quickly to particular questions. Furthermore, through comprehensive questioning of the client, the expert system is likely to arrive at a better solution than the client could reach by independent study and analysis. 57 Figure 2. Sample scenario dt I. Client Dialogue: Do the principals have enough capital for the business? YES Could the firm�'s activities cause personal injury or property loss? YES Is the potential liability too big to be carried by individual owners? YES Does the liability exposure or the need for liquidity and or continuity justify the additional administrative costs and tax consequences of becoming incorporated? YES What type of owner(s) does the company have? INDIVIDUALS Is the firm's passive income greater than 20 percent of total income'& NO Is the business already incorporated with only one class of stock? NO Is the company a U.S. corporation with only one class of stock? NO Conclusions: BILL —since your business is already incorporated, you should be incorporated as a C Corpo- ration. Why: —Liability, liquidity, or continuity are factors —There are 35 or less owners or stockholders —The owners are composed of individuals —Passive income is less than 20 percent —The firm is already incorporated and has more than one class of stock Client Confidence Factors: —Increased Liability Exposure 100% —Need for Liquidity 80% —Need for Continuity 0% —Need for Administrative Skills —100% 5&t Figure 3. Sample scenario i'. Client Dialogue: Do the principals have enough capital for the business? NO Can capital be borrowed for banks or venture capitalists? NO Could the firm's activities cause personal injury or property loss? YES Is the business a profession e.g., Dr., CPA, or lawyer, etc.? NO Is the number of owners less than or equal to 35? YES What type of owners does the company have? ESTATES AND TRUSTS Is the firm's passive income greater than 20 percent of total income? NO Have you spoken to an investment banker about your finding requirements? NO Does the firm have sufficient legal, accounting, and administrative resources available? NO Conclusions: BILL —your business should be set up as an S Corporation. However, you should consult an investment banker before incorporating. You will also need to acquire additional administrative, accounting, and legal services. Why: —External equity capital is required —There are 35 or less owners or stockholders —Primary owners are estates and trusts —Passive income is less than 20 percent of income Client Confidence Factors: —Increased Liability Exposure 70% —Need for Liquidity 0% —Need for Continuity 0% —Need for Administrative Skills 90% 59 Finally, thc small business consulting expert system can benefit the SBDC by offering assistance and "what-it" training experiences to less expcricnccd consultunks, allowing part-time volunteer counselors to pcrfonn more cffcctively and efficiently. Thc prototype cxpcn system has not solved all the probltnns in selecting an appropriate legal forin ol'usiness, but it exemplifies thc way in which a computer's flawless memory and systematic thnught process can enhance a client's comprehension of the multiple factors influenc- ing thc selection of an appropriate business legal structure. Morc work remains to be done. While this article has focused on an SBDC application, thcrc is no reason to believe that other small business advisory organizations such as SBIs could not bcnelit Crom such systems. Microc&miputers and commercial expert system shell software Cor developing and economically deploying advisory programs are now being used by numerous cnd-user organizations. Hope- fully, other upplications like the one described will help to ptovidc solutions to thc needs of small business advisory groups and the important segment of our economic system which they serve. REFERENCES Ebersold, B. (1991). Meeting a competitive challenge in the frequent flyer competition. In R. Smith & A. C. Scott (Eds.), Imioi arire Applications rif Anificial Intelligence 3 (pp. 11-19). Cambridge, MA: The MIT Press. Feigenbaum, E., McCorduck, P., & Nii, H, P. (1988). The rise of the expert company. New York: Times Books. Frenzel, L. E. (1987). Crash course in arrificial intelligence anrl erperr systems. Indianapolis, IN: Howard Sama. Grover, M. D. (1982). A pragmatic knowledge acquisition methodology. Redondo Beach, CA: TRW Defense Systems Group. Guterl, F. V. (1986, October). Computers think for business. Technrrlrrgy, 30-37. Harmon, P., & King, D. (1985). Expert systems: Arrificial intelligence in I usiness. New York: John Wiley & Sons. Harmon, P., Maus, R., & Morressey, W. (1988). Experi Systems iools and applications. New York: John Wiley & Sons. Heath-Brendt, P., Carter, S. H., & Yozie, M. A. (1991).MCI's pricing system. In J. Liebowitz (Ed.), Operational erperi system apph'carions in rhe Um'ied States (pp. 89-97). New York: Pergamon Press. Knowledge engineering; Application qualification. (1989, August). Texas Instruments seminar at the Eleventh International Joint Conference on Anificial Intelligence, Detroit, MI. Liebowitz, J. (1988). Inrroducrion to erpen systems. Santa Cruz, CA: Mitchell Publishing. Martin, W. S., Jones, W. T., McWikiiams, E., gt Nabors, M. V. (1991).Developing artificial intelligence applications: A small business development center case study. Journal of Small Business Munagemenr, 29(4), 28-32. McCann, J., Tadlaqui, A., &. Gallagher, J. (1990). Knowledge systems in merchandising: Advertising design. Journal of Retailing, 66, 257-277. Mentzer, J. T. & Gandhi, N. (1992). Expert systems in marketing: Guidelines for development. Juurnul of tire Arudemy of Marketing Science, 20(4), 73-80. Mishkoff, H. C., Shafer, D., & Rasmus, D. W. (1988). Understanding anificial intelligence. Indianapolis, IN: Howard Sams. 60 Mockler, R. J., & Dologite, D. G. (1992). Knowledge-based systems: Anintroducrion to experr systems. New York: Macmillan. Muller-Boling, D., & Krichoff, S. (1991). Expert systems for decision support in business start-up. Journal of Small Business Management, 29(2), 83-90. Polfer, A. T., Dicks, D., & Holland, K. K. (1990). A practical guide to starting & managing ii biisiness in Cenrral Florida. Orlando, FL: Business Development Associates. Prerau, D. S. (1990).Developing and managing expen systems. Reading, MA; Addison-Wesley. Rangaswamy, A., Burke, R., Wind, J. & Eliashberg, J. (1987). Expert systems for marketing: A working paper (Report No. 87-107). Cambridge, MA: Marketing Science Institute. Rich, E. & Knight, K. (1991). Anificial intelligence. New York: McGraw-Hill. Silverman, B. G. (1987). Should a manager 'hire'n expert system. In B. G. Silverman (Ed.), Erpen systems for busiuess (pp. 5-23). Reading, MA: Addison-Wesley. Springer, S., Buta, P., & Wolf, T. C. (1991). Automatic letter composition for customer service. In R. Smith & A. C. Scott(Eds.), Innovative Applications ofAnificial Inielli genre 3 (pp. 67-83). Cambridge, MA: The MIT Press. Sullivan, J. J., & Shively, G. O. (1989). Expert system software in small business decision making. Journal of Small Business Management, 27(1), 17-26. Watabe, K., Holsapple, C. W., & Whinston A. B. (1991). Solving complex problems via software integration. Tlie Journal of Computer Information Systems, 3I, 2-8. Waterman, D. A. (1986). A guide to erpen systems. Reading, MA: Addison-Wesley. Wielgos. R., & Ragusa, J. M. (1991).A Review and comparison ofselected erpert system shells. (Research Report No. I). Orlando, FL: University of Central Florida, Intelligent Multi- media Applications Laboratory. 61