While there is still a great deal of incredulity in the market concerning the use of object-oriented technology for writing applications, it has been in use commercially since 1988 via AICorp’s KBMS Knowledge Base Management System. Waltham, Massachusetts-based AICorp is positioning its products against traditional, procedural languagessuch as Cobol and focusing on the knowledge-based generation of applications using forward-chaining, backward-chaining and hypothetical reasoning within an object-oriented system. The company was set up in 1975 by Larry Harris and launched its first product, the natural language system Intellect in 1981. This offers natural language access to data on the mainframe and distribution deals were signed with Cullinet, IBM and Information Builders Inc. Until 1986 AICorp was a research-driven, very academic-oriented company with one product.
Mainstream computing
Then a group of Cullinet Software managers joined AICorp, led by Bob Goldman who is now chairman of the company. Goldman began to promote the idea that AICorp should move away from its image of being a specialist artificial intelligence company and should move more decisively into mainstream computing and conventional data processing. Having come from the database and applications generator side the industry, Goldman urged the exploration of the use of expert system technology in these areas. In 1988 AICorp moved into the system development market with the launch of KBMS – user base for both KBMS and Intellect stand at round 650 apiece, and as the KBMS product offering includes a subset of Intellect, the two user-bases are counted separately. Intellect, however, is acknowledged by UK managing director Keith Deane to have a static user-base, while growth is expected from KBMS, which is now available under OS/2, MS-DOS and VAX/VMS and will be delivered to the Sun and RS/6000 Unix environments by year end. AICorp is an IBM Business Partner, a Digital Equipment Corp Complementary Software House and has a key Independent Software Vendor role with Oracle Corp – the last being a deal signed about the time Ingres launched its Knowledge Manager. Elements of KBMS are being used in the Initiative for Managing Knowledge Assets, funded by DEC, Ford Motor Co, Texas Instruments Inc and US West Advanced Technology, and developed by the Carnegie Group. The Initiative is developing knowledge-based decision support applications that can be embedded (CI No 1,588). In March 1989 AICorp came to Europe via a strange financing deal created so as not to dent the company’s bottom line before a public flotation. AICorp Europe is an independent company, 91% owned by Schroder Ventures with the remaining 9% owned by AICorp in Waltham. Schroder is bankrolling the start-up costs to a limit of $7m and for five years AICorp gets a 50% royalty on everything the European company sells. It is the parent company’s intention to buy Schroder out in three to five years but that is not a condition of the agreement. Instead, if AICorp does not take up the option to buy the European firm then royalties will drop from 50% to as low as 5%, the operation will be free to carry other products and Schroder can sell to another bidder. AICorp is into the second year of its five year option to buy the European company and went public last year.
By Katy Ring
Over the past year AICorp has grown by approximately 50% and expects to come in at year end with profits of approximately $5m on turnover of $22.5m. So, bearing in mind that several rival companies have disappeared over the past year including AI Ltd and Expertech Ltd, what is the secret of AICorp’s success? It sells KBMS as another software tool in the tool bag offered by large vendors and by consultancy groups and software houses such as Logica Plc, where a tool is required to embed a judgemental element in a wider suite of code. For example, whereas financial accounting packages and payroll applications are ideal for procedural languages, there are other areas such as underwriting, credit authorisation and scheduling that can be
volatile and where a knowledge-based system is more appropriate. Deane believes that IBM’s AD/Cycle endorsement of knowledge-based systems through the launch of The Integrated Reasoning Shell has helped independent vendors get a hearing from more mainstream technology buyers. He thinks that knowledge-based systems have unduly suffered from the conception that they are just a diagnostic technology, forming judgements from rules via applications that live in a world of their own, somehow separate from the rest of the data processing department. However, KBMS has database access, which means that unlike many knowledge-based applications, it can access corporate data held in a variety of relational databases. This means that applications can be built to embody judgement by reasoning across supported environments, accessing data and processing power on the mainframe. Furthermore, KBMS has portable code, which means that an application can be built on one machine type and delivered on another without changing a single line of executable code. The environment in which the application operates is hidden as everything is addressed at the object level – the code, written in Intellect, doesn’t include network calls, database calls and so on as KBMS generates the appropriate calls below that object level. So, for example, SQL calls in and out of the database can be embedded via KBMS so that an application transparently calls in KBMS rather than, say, DB2 and presents the information to the user via a standard graphical user interface.
Added kudos
So what types of areas are suitable for KBMS? Areas where the problems are changing frequently such as pricing, distribution and government regulations are appropriate, since with KBMS an end-user can maintain the relevant algorithms. Similarly areas where many interrelated and interdependent rules come into play, as in scheduling and routing applications, are difficult to optimise in a procedural language such as Cobol. Then there are the more recognised domains of knowledge-based systems: rule enforcement in, say, credit authorisation and auditing; and diagnostic systems that could be used as part of a data processing helpdesk. The wider commercial uses for artificial intelligence are becoming more commonplace and are creeping into software engineering and database products, so why buy from a specialist like AICorp when you can buy rules-based components from vendors with well-known application development offerings such as Software AG’s Natural Expert? Deane reckons that with Natural Expert, embodying rules in a programming language doesn’t buy you much. He thinks AICorp’s three-year lead in the market gives it a head start, making it hard for others to catch, but he welcomes their entry into the market, as it gives KBMS added kudos.
