Gentia - Agents Explained     Introduction: The Need For Intelligent Agents 

What are software Agents? 

Gentia, a commercial implementation of an Agent Architecture 

Uses to which Agents may be put 


A Planning Sciences White Paper   Planning Sciences International LimitedAbstract  
Introduction: The Need For Intelligent Agents 
Identifying the characteristics 
Information systems: changing the aims and goals 
Sifting the Information 
What are software Agents? 
Market Potential 
Gentia, a commercial implementation of an Agent Architecture 
Uses to which Agents may be put. 
Agents in the Future 

Abstract 
The scene is set for agent technology by analysing the issues that face businesses today as they try to leverage their investment in IT that has been made over the last 30 years. 
As the volume of structured information continues to grow, at a certain threshold the point and click type of interface breaks down and we need to look at how vast amounts of data can be viewed intelligently. Also, that unstructured or 'soft' information has been generally ignored. How can we better access and 'mine' large quantities of structured and textual data? 
The concept of software agents is introduced by looking at how managers currently delegate tasks to Personal Assistants and how performers employ business agents to handle tasks that are necessary to perform their job, but which can be accomplished by less skilled individuals. The potential market for software Agents in information analysis and retrieval is considered. 
How do software agents fit into the current IT architecture? How do they operate in a client/ server environment? We see that Agents extend the concept of client server to encompass a 'federated' Client/Server environment, with all machines on the network being able to participate in agent processing. 
How do agents manifest themselves in Gentia? The concepts of the Gentia Agent and Agency Architecture and user interface is introduced with worked examples. Agents and their roles are summarised  and the conclusions drawn that Agent Architectures are the  next logical step forward in commercial business computing. 

Introduction: The Need For Intelligent Agents 
As the millennium draws to a close, modern businesses are faced with an era of unprecedented global change. This change affects every part of the business from the expectations of the customer to competition, from development of new products and services to deregulation of the market, from rapidly changing political geographies to emerging standard electronic communication formats. There is a need to be able to develop and deliver world class products, whilst being able to continuously change the business practices and processes that make up the business. In order to be able to achieve this successfully, technology must be fully exploited as an agent of business change. 
The pressures on business during the 1990s and beyond are not simply a matter of being able to control and cut costs. Rather it is being able to change and develop the business practices to increase revenue at the same time as cutting costs. In today's cut-throat world this is the minimum that is required in order to stay competitive and stay in business. 
Over the last 30 years of commercial computing there have been a number of 'waves' both of technology and the use to which that technology has been put. It is possible to identify three main waves from the batch systems of the 1960s through the emergence of online transaction processing and fourth generation query tools in the 1970s, the acceptance of relational database technology, increasing information demands and the rise of the Personal Computer to a place in the strategic IT planning of organisations during the 1980s and early 1990s. Although these changes have been compared to 'waves', it would be more apt to consider then as ripples in a pond, expanding outward from a thrown pebble. At any one time more than one wave ripple will be in motion. While new information systems may make use of client/server architecture and GUI front-ends, the mainstream business systems may still be batch oriented or based on pre-relational database technology. The pace of technological change has been matched over the last ten years only by the changes that have taken place in the business world. 
This paper aims to identify the next steps in information system development, its changing aims and purpose together with the underlying technology architecture that will make this possible. 
Identifying the characteristics 
The characteristics of these Fourth or New Wave Information systems can be considered first by identifying those existing or near term technology features that would be a 'tick in the box' for any evaluation criteria matrix. These would include support for very large multidimensional databases, transparent access to SQL based data whether for data loading or for drill down from the bottom level MDDB to the underlying SQL databases, performed automatically and transparently and support for Distributed Multidimensional Database making full use of all the resources on a network for data storage and retrieval. Similarly support for Massively Parallel Machines  (MPP) for big databases. A fast, easily maintainable development environment is essential, using object oriented techniques as the enabling technology, but not requiring rocket scientists for programmers and developers. Indeed, the whole point of object orientation is make things easier and more intuitive. These development tools should allow for the support of distributed applications across heterogeneous machines and networks. The build once, deploy many paradigm. 
As well as the user interface and operation of the system, New Wave Information Systems attempt to tackle information areas that have been previously avoided. These areas have been avoided simply because they were too difficult to achieve, for a variety of reasons. Two of the neglected areas of information systems that have been avoided up until now are the retrieval of relevant details from unstructured or 'soft' information, essentially free format text, and the analysis of very large volumes of data looking for speculative trends and relationships between data dimensions. This has been christened data mining and until now has been restricted to those organisations that can afford specialist data storage and retrieval technology. It is one of Planning Sciences' missions to bring the benefits of information retrieval from free text and the power of data mining within the reach of smaller organisations without the need for specialist hardware and software. 
Workgroup oriented software is beginning to be widespread thanks mainly to the proliferation of Local Area Networks and the change in business practices toward case book type approach to business procedures rather than the functionally defined areas of a few years ago. It is essential that any development and deployment environments for the next generation of information systems should support workgroups and workgroup administration implicitly. 
Information systems: changing the aims and goals 
When industry is surveyed, the one overriding attribute that affects the bottom line of an organisation is the productivity of the work force. In a manufacturing company, this is easily measured for the blue collar workers, the guys actually stamping out the metal or assembling the TVs and Videos. Productivity in this area has steadily risen over the last twenty years by an estimated fifteen to twenty-five percent. 
In that time there has been an estimated six trillion dollars spent on computer and office systems, so there would be an expectation that white collar or managerial productivity has also increased alongside the manufacturing productivity. Not so. In this same time period white collar (also known as knowledge or managerial worker) productivity has risen by a paltry one half of one percent. Why? 
The simple answer is that the plethora of information systems that are in place are an addition to the information load rather than an automated replacement. If the normal information assimilation processes can be automated, then we may see an improvement in productivity and provide more time for better decisions with superior information. 
In the past almost all information systems concentrated on providing a view of the state of the business by reviewing the current and historical data that has been accumulated via operational systems. Sales revenue for this month or quarter could be compared against the similar month or quarter for previous years. Some extrapolation was sometimes attempted to forecast predicted revenue for budgeting purposes. The next generation information systems will take this a stage further. Rather than simply reporting upon historical data, these systems will attempt to identify new business opportunities or changes in the way of doing existing business from the data and information collected. Rather than calling these applications information systems, they can be better described as business development applications. For example, a well known US supermarket embarked on a data analysis exercise and discovered a correlation between the sales of disposable nappies, time (five to seven P.M. on Fridays) and males aged 25 - 35. The explanation was that Dads with young children bought the weekend supply of diapers on the way home from work on a Friday night. The supermarket cashed in on this information by placing the beer display next to the disposable diapers. Beer sales increased by 50% within a month! It is this type of business development, aimed at increasing revenue but minimising increases in costs, that business development systems are all about. 
Sifting the Information 
Information that influences decisions comes from a large number of sources: TV; telephone conversations; radio; electronic mail; computer information systems; printed reports and memos; personal and shared spreadsheets or external information providers such as Reuters or the World Wide Web. Information sources need to be combined into a complete information system in order to fully leverage and exploit computer technology. Structured information systems play a relatively small part in the information providing world. Having said that, increasing volumes of data are available that need to be handled in an intuitive and innovative way. This can be provided by Multi Dimensional Databases capable of handling Very Large Database (VLDB) sizes (>50 GBytes). 
The method by which our perception of the world is formed is very sophisticated. It is a synthesis of many information sources. Those sources bombard us with information virtually all our waking life. All of the above are sources of information that are used during the decision making process. For some, the sifting of this information can be delegated. A senior executive has a personal assistant who manages her diary and schedules activities. Junior managers and others will be instructed to report on specific items of interest to that executive and will act on her behalf to sift the relevant information from the morass of irrelevance. 
In the entertainment world, a talented performer employs a business agent or manager to handle the booking of venues, management of contracts and arrangement of fees, allowing the performer to concentrate on what they are good at, performing. It would be very beneficial if this type of delegation of human tasks could be somehow automated and made available to the majority rather than the minority. 
As part of the next generation information systems, these activities of  information sifting and the delegating of tasks to semi autonomous software objects will become commonplace. 
Traditionally the route to information of interest is to define the set of data to be interrogated, then define the type of conditions that are of interest. This may be trends, other curves, exceptions etc, and general relationships between specified dimensions. 
Thus the information that is of interest is simply the information that is believed to be of interest: we only know what we already know! While the volume of data and information was relatively low and it was in an easily handled structured form, this approach was feasible, but with the very large volumes of disparate data and access to the vast amounts of unstructured textual information, this approach is no longer valid. A different approach is required to information retrieval, one offered by the emerging technology of intelligent software agents. 

What are software Agents?