Intelligent software agents were hyped as the most important Internet technology only a few years ago. Both respectable research labs and ambitious startups were announcing "smart" agents that could act for their "owners", performing such tasks as information collection, commodity price setting, negotiating, agreeing contracts, and auctioning. On the practical side, however, all we could see was "work in progress", without any real practical value. Needles to say, the whole idea lost some of its attractiveness and returned to the R&D; departments. However, e-commerce remains one of the areas in which the agent technology could unleash its full power. As Clive Thompson put it in one of the articles published in the ROB magazine, using intelligent agents to buy and sell goods can create truly rational market behavior - just the way economics is supposed to work.
Probably the best known use of agents in e-commerce for the average Web user is comparison shopping. There are literally hundreds of sites offering such services, and many of them are using some sort of agent technology. The limitations of HTML - intermingled structure, content and presentation - are becoming increasingly evident to the agent developers. At the same time, the limited descriptive capability of HTML often necessitates manual analysis. Wrappers facilitate access to Web-based information sources by providing a uniform querying and data extraction capabilities. Wrapper learning and maintenance resources provide valuable information for automating the information retrieval and extraction process. Several automated and semi-automated tools were described in previous articles on data extractors and SQL-style information retrieval. An article from Microsoft Systems Journal describes the development of a very simple comparison shopping agent for a Computer Finder, a fictitious company that offers users a Web page for specifying desired computer features, such as RAM, processor, and price. When the form is submitted, Computer Finder steps through its list of vendors, sending a query to each one based on the user-specified parameters. It then assembles the results and formats them for display to the user.
Another common application area for the "conversational" agent technology is Customer Relationship Management (CRM). Artificial Life, Inc. offers conversational bot-based products for customer service, consultative selling, Web site navigation, automated e-mail response and financial portfolio management. NativeMinds, Inc. and Novuweb, Inc. are also making the strong efforts in this field. The Chatterbot Collection by Simon Laven provides extensive background information and comprehensive instructions on how to interact with dozens of chatterbots. You may also want to check this article that provides a high-level overview of the various agent technologies used in Manna's FrontMind product. FrontMind is a high-end, mission-critical server for Internet Relationship Management that permits commercial sites to conduct online learning relationships with their customers through intelligent business rules.
Many researchers are focusing their efforts on the issues involved when multiple agents interact in multiagent systems. Following is a list of interesting, E-Commerce related projects and resources oriented towards intelligent agent marketplaces.
eMediator is a free-to-use next generation electronic commerce server. It provides several services that facilitate efficient commerce, both in terms of process and outcome. It uses state-of-the-art technologies developed by the Multiagent Systems Research Group at the Computer Science Department of Washington University in St. Louis.
MAGNET (Multi AGent Negotiation Testbed) from University of Minnesota provides support for complex agent interactions, such as in automated contracting, as well as other types of negotiation protocols. Business-to-business situations in which a company outsources production to outside contractors and needs to automate as much as possible the management of the supply-chain are one of the focal points of this project.
The Zeus Agent Building Toolkit provides an integrated environment for the rapid development of multi-agent applications in Java. Its Agent Component library provides a set of high quality, pre-written and pre-tested agent components. The rationale is to liberate developers from the complexity of agent technology, allowing them to concentrate on solving their application's problems instead. It also provides an integrated suite of editors that guide developers through the stages of the agent development methodology. During this process developers describe the agents within their application, how they interact, and the tasks they perform. The Visualization Tools collect information on agent activity, interpret it and display various aspects in real-time. This is the solution to the inherently difficult problem of analyzing and debugging a multi-agent system where all the data, control and active processes are distributed. These features are enough to make Zeus one of the best agent building toolkits available today. What makes it so appropriate for the E-Commerce environments is a set of in-depth manuals and case studies for building trading applications and simulated supply chains.
Geobot's InfoSwitch is an object-based, open-framework architecture that takes advantage of the latest technologies such as XML, intelligent agents, and Java to provide a versatile foundation for building an event-driven e-business network. Through the use of intelligent-agent technology, businesses can quickly incorporate new services and functionality without the cost of redesigning their networks.
AuctionBot a multipurpose Internet auction server developed at the University of Michigan. You can use this free service to to create automated Internet auctions according to your specifications.
IBM's Business Rules for E-Commerce project is investigating rule-based business processes for E-Commerce: both business-to-business (B2B), e.g., to integrate supply chains; and business-to-consumer (B2C). The main goal is to develop technology that is highly reusable and easy to integrate with a broad spectrum of networked applications. It is embodied as an extensible structured Java library, called CommonRules (formerly called DIPLOMAT, follow-on to RAISE and Agent Building Environment implemented in C++), with an preview release freely available at the AlphaWorks site.
The Fishmarket Project investigates communicational aspects of multiagent systems. It involves network-based marketplace prototypes based on the trading conventions present in the traditional fish markets.
Students interested in learning more about agent technology applied to the E-Commerce environments can find several interesting course programs on the Web. CC435: Agent Technology for E-Commerce at University of Essex, for example, provides practical instruction in the key design principles and implementation techniques for software agents. Another interesting source of information is a book Constructing Intelligent Agents With Java : A Programmer's Guide to Smarter Applications that describes a development of a complex multiagent marketplace application.
Developers of agent solutions for business-to-business integration will need to learn more on another emerging standard, called Universal Description, Discovery and Integration (UDDI). Introduced by Microsoft, IBM, Ariba, and several other major players in this field, UDDI promises to become an Internet standard for creating online business registries. Web Services Description Language (WSDL), the core technology behind this initiative will allow businesses to describe their offerings in a standard, machine readable way.
