Barry_A_Stevens@cup.portal.com (11/27/87)
EXPERT SYSTEM SHELL SURVEY Copyright 1987 Applied AI Systems, Inc. We recently sent a questionnaire to 1700 users of PC-based expert system development shells. One hundred seventy nine firms responded. The survey was intended as a snapshot of the expert system market during the preparation stages of a business plan. It was understood in advance by all parties concerned that: IT WAS NEVER MEANT TO BE AN ACADEMICALLY CORRECT SURVEY. IT WAS ONLY TO PROVIDE SOME GENERAL INFORMATION. If you can accept the above limits, what follows may be of interest to you. If an imperfect survey is an abomination, you can skip the remainder of this file. The purpose of the survey was to educate and inform, on a gross level, about the expert system shell marketplace. Information sought included: profiles of the shell users and their organizations; general strengths and weaknesses of expert system shells; the decision process followed by users when buying a shell; the reasons for getting into expert systems; expert system software in use; job titles of people using expert system tools; and applications that have been implemented using shells. One hundred seventy-nine questionnaires were completed and returned. The survey contained some questions whose answers are confidential. We thought that some of the results, summarized for bervity and sanitized to maintain confidentiality, might be of interest. WHAT CHARACTERISTICS MAKE A GOOD - AND BAD - EXPERT SYSTEM DEVELOPMENT SHELL? Many of the questionnaire respondents indicated that they had made studies of multiple expert system development shells. Thirty two of those respondents offered the following general comments about factors that they viewed as strengths and weaknesses of those tools. A strength was defined as a reason they would buy a tool as a result of a product evaluation, while a weakness would cause them to reject a tool. Strengths General strengths are described below, with the number of respondents mentioning each factor shown. The tool should be useful in many microcomputer, minicomputer, and mainframe environments, under different operating systems. (6) The capability to access other programs and data should be provided. (5) The tool should be capable of frames and/or object representation as well as representation by rules. (4) Math functions and numeric and text variables should be usable in rules (2) Rules should be easy to structure; (2) The product should be easy to learn. (2) The product should be easy to use. (2) Both editor and user interface should be in English, or natural language; Graphics and a good user interface should be available; Procedural components should be available for sequencing and interaction, including the ability to clear previous answers and ask questions again; The tool should handle probabilities, including fuzzy logic; The tool should learn by examples; Sophisticated WHY and HOW capability should be available to explain reasoning; Good support and training should be available; Good documentation should be available. Weaknesses General weaknesses of expert system development shells that were identified by users are described below. Cost of a product should be appropriate for its capabilities and performance; (the survey indicated that users are price sensitive, and price is a significant factor in purchasing a product.) (12) Special hardware requirements are a problem. The tool should run on a standard PC or other commonly available environment. (5) Knowledge base size limitations are a problem. Several available products limit the number of rules that can be defined. (3) Slow execution speed is a problem. Execution speed should be such that a large number of rules can be executed in a reasonable time. (2) Lack of flexibility in knowledge representation and use is a problem. (2) THE PROCESS OF MAKING THE DECISION TO BUY We asked about the process of decision making that went into the purchase of a PC-based expert system shell at a price of $400. We wanted to know who made the decision, and how it was made. Who (by organizational unit) made the decision to buy? INTERNAL ORGANIZATION NUMBER % RESPONSE Research and Development 58 31.9% MIS/Data Processing 36 19.8% Independent individuals 31 17.0% Operating (line) organizations 26 14.3% Management/staff function 19 10.4% Advanced Planning 10 5.5% Other 2 1.1% TOTALS 182 100% Who (by reporting relationship) made the decision to buy? WHO MADE DECISION NUMBER % RESPONSE Me 151 88.8% My Boss 17 10.0% Different department 1 0.6% My subordinate 1 0.6% My boss's boss 0 0.0% TOTALS 170 100% How was the decision to buy made? DECISION PROCESS NUMBER % RESPONSE No formal decision process 99 40.9% Product review and comparison 64 26.4% Internal needs assessment 26 10.7% Cost justification 40 16.5% Formal review and planning process 6 2.5% Other 7 2.9% TOTALS 242 100% What were the reasons for getting into expert systems? REASON NUMBER % RESPONSE To capture knowledge 76 16.0% It's a training tool 71 14.9% Part of overall corporate strategy 60 12.6% To improve quality of work 47 9.9% To improve quality of product 41 8.6% To learn expert systems 41 8.6% It's a competitive weapon 34 7.2% It's in fashion 27 5.7% To achieve a cost savings 23 4.8% It's a marketing tool 22 4.6% To provide an MIS/DP capability 19 4.0% Other 14 2.9% TOTALS 475 100% What types of expert system software are installed? We found that 76 distinct products were in use from nearly as many vendors. It is interesting to note the categories into which these products fell. Lisp Based Tools: total units installed: 80 number of products installed: 15 Prolog Based Tools: total units installed: 120 number of products installed: 12 Development tools/shells: total units installed: 223 number of products installed: 49 JOB TITLES OF PEOPLE USING SHELLS We were interested in the job titles of people who had built expert systems using shells. You may be interested as well. Advanced Programmer/Analyst Advanced R&D Project Engineer Advanced Technology Group Advisory Engineer AI Branch Chief Analyst Assistant Professor Assistant Vice President Associate Professor Audit Manager CEO Chairman of the Board Chair, Department of Communications Chemist Computer Scientist Computer Specialist Consultant Cost Analyst Design specialist Director Clinical Laboratories Director, AI in Business Director, Clinical Research Director, Computation Center Education Specialist Engineer Executive Consultant Financial Services Officer Graduate Assistant Group Leader Head Intelligent Systems. Lab. Instructor Learning Center Manager Lecturer Manager, R&D Manager, Analytical Chemistry Manager, Information Resource Management Manager, Operations Manager, Product Evaluation Office Automation Manager, Proposals Manager, Remote Sensing Lab Managing Vice President Materiel Operations Manager Owner Physician President Principal Professor of AI & ES Professor of Chemistry Professor of Real Estate Program Manager Programmer Project Coordinator Project Engineer Regional Manager Research Agronomist Research Assistant Research Forester Research Manager Research Scientist Seismic Processing Analyst Senior Analyst Senior Engineer Senior Research Chemist Senior Research Fellow Senior System Analyst Senior Tax Manager Senior VP & Senior Trust Officer Special Projects Director Staff Machinery Engineer Staff Research Engineer Staff Scientist Systems Officer Technical Advisor Technical Consultant Technical Journalist Technical Staff Member Technology Assessor Underwriting Director Unit Head Conventional Safety Standards Vice President Wildlife Ecologist While there are a few titles that indicate specialization in AI, most are seen to be outside that category. Expert systems are being built by technical professionals even with limited computer experience. APPLICATIONS FOR SHELLS We asked about the applications that were either built or in development using shells. In many cases, the reply indicated that the nature of the application was confidential and no application name was given. The names below are those that were listed in user responses, with little editing and no attempt to explain. Account business assessment Advertising copy development Advice on single family home purchase Advice on stock and commodities trading Advise nursing students on the care of patients Advising on choices for new technology Advisor on choosing soy bean varieties Advisor on design of new magnetic components Aid for financial futures traders Aid for isolating failing chips Aid in salmon stocking rates, species selection Alarm management system Analysis of simulation results in bank product planning Analysis of soil site characteristics Analysis of X-rays Application sizing based on similar applications Assist in compiling tax planning ideas Assist in diagnosis of computer console messages Assist in identification of rare antibodies Assist new users of DOS Assist service desk in troubleshooting application problems Assistance in search for part numbers Augment expertise of resource manager Broker syndication planner Call screening to interview users with application problems Career development Causal model of account marketing Chemical process diagnosing and troubleshooting Choosing a living or testamentary trust Choosing an executor for trusts Classification of data from satellites Classifications of software programs Closing and issuance assistance Commercial loan credit analysis Commercial loan documentation check list Computer modeling support Computer system configurator Configurator for selecting, sizing, and writing parts list Configuring programmable controller system Conservation equipment tillage selector Correct selection of cost codes Cost/benefit assistant Create standard loan documents based on characteristics Credit control system Crop management and irrigation simulation Customer assistance in selecting types of investments Customer service advisor for problem resolution Customer water quality analysis Data communications troubleshooting Decision support for correct testing by auditing Detailed analysis of hardware and software problems Detailed design for asphalt concrete pavement Determine correct mixture for propellant ingredients Determining best shipping documentation and routes Diagnose telecommunications difficulties Diagnosis of sports related injuries Diagnostic advisor for pulp bleaching DP production support system Epidemiology expert system Equipment fault diagnosis Equipment troubleshooting Estimate employee's potential retirement salary Estimating construction costs Evaluation of commodities purchases Evaluation of multi-family housing projects Evaluation of stock purchases Fault diagnosis for electronic hardware Federal contract management Fertilizer recommendations Fertilizer, climate, and soil interaction Finding phases present in super alloys Forecast snowfall accumulation Forecasting severe convecting weather Futures, stocks, and options trading Gas turbine troubleshooting Geographic information system analysis aid Grading of graft vs. host disease Hardware and software selection Hardware failure analysis Hardware sizing assistant Hazardous chemical ranking Implementation planning assistant Industrial training Interpretation of statistical quality control data Invention patentability expert Irrigation and pest control management Lime recommendation system Line diagnosis and fault detection Local area network selection Machine advisor for grinding, milling, turning Manufacturing resource planning aid Market segmentation and positioning Marketing advisor for process control systems Material selection by engineers Materials selection for specialized component parts Medical decision making Medical diagnosis MIS decision support system Mortgage credit analysis Network operations systems diagnosis Papaya management system Pavement performance diagnosis Pavement rehabilitation PC configuration PC Hardware and software configurator Perform hematological diagnosis Personal tax advisor Pest management and soil interaction analysis Portfolio construction Power plant boiler tube failure identification Problem diagnosis for local area networks Problem diagnosis for printers on a SNA network Product development support system Product performance troubleshooting for salesmen Product selection system Production scheduling Psychiatric interview Quick proposal estimator Radar mode design workstations Rating for substandard life insurance Real estate appraisal Real estate site selection Real time process control Real time troubleshooting for wastewater process control Recommend documentation to computer users Relay diagnosis Risk assessment of error or fraud in financial statements Salary planning Sales order analysis Salmon diagnosis and treatment Select pension types Select, recommend library reference materials Selection of non-materials in aerospace applications Selection of solvents for chemical compounds Service network assistant Software development risk analysis Software system diagnosis model Software vendor risk analysis Soil acidity analysis Soil characterization and utilization Solid waste disposal management assistant Space shuttle payload on-orbit analysis Strategic alternatives for a fragmented industry Strategic marketing and planning aid Structural damage assessment Student financial aid eligibility Submarine approach officer training System to identify feasible rehabilitation strategies System to prepare process estimates Tactical battle management Teaching mineral and rock identification Telephone system configurator Toxicity of laboratory chemicals Training in gas turbines Training new financial planners Troubleshooting airplane starting systems Underwriting assistance Underwriting guidance for line underwriters Weed identification When to perform a physical audit Applied AI Systems, Inc. and Barry Stevens may be reached at PO Box 2747, Del Mar, CA, 619-755-7231.