Intelligent agents and electronic commerce

Citation

(1999), "Intelligent agents and electronic commerce", European Business Review, Vol. 99 No. 5. https://doi.org/10.1108/ebr.1999.05499eaa.001

Intelligent agents and electronic commerce

Intelligent agents and electronic commerce

The Internet provides the potential for unprecedented access by individuals and organisations to information with which they can make better-informed choices when purchasing goods and services. However, the lack of refinement and imprecision of existing search engines has been a major constraint. The development of intelligent agent software is providing better access to information and enhancing the ability of purchasers, suppliers, vendors and intermediaries to communicate with each other.

Most markets require some form of mechanism or intermediary to provide access to different suppliers' products or product information, the Internet is no exception. Organisations, such as eSmarts, have been established to provide consumers with information about the quality of service offered by vendors. eSmarts provides reviews of sites, it collects customer views of the service provided by Web sites and records other, predominantly qualitative, information about online stores.

This type of qualitative review of online stores is complemented by the use of intelligent agents which collect quantitative information, predominantly about price, for consumers. The delivery of exact product or service information to the desktop, permitting consumers to browse products from dozens of online stores and compare prices is now possible for some products and services. Intelligent agent software could have considerable implications for producers, intermediaries and consumers.

Agents are a rapidly emerging new paradigm in artificial intelligence and computing. Put simply, an agent is like a personal secretary that can help to do things on your computer and the Internet (Anon, 1997). Despite a great deal of hype and the development of some interesting applications in electronic commerce, agent technology is still in its infancy. But as intelligent agent technology develops it has the potential to play a crucial role in the way we use the Internet to search, filter and retrieve information to undertake electronic commerce.

Numerous, generally overlapping, definitions of agents and intelligent agents exist. A thorough review of the many definitions and the thousands of agents available online can be found at BotSpot. They prefer to use the colloquial term "bot" for an agent. This is derived from robot, which itself is derived from "robota" being Czech for "work".

Two different technical categories of intelligent agents have been identified (Anon, 1997) - reactive and deliberative agents.

Reactive agents are primarily data driven and respond to stimuli. Reactive agents can be adaptive and perform symbolic reasoning. Examples of a reactive agent include those which monitor and respond to alarm system and commercial agents, such as Autonomy's Knowledge Update with Live Alert which monitor changes or additions to Web pages or live newsfeeds and alerts users to information or stories of specific personal interest (Agentware, 1998).

Deliberative agents are goal driven. They can maintain belief models (of the user's interests), use symbolic reasoning and are able to perform complex types of reasoning and planning. Examples of deliberative agents include those which undertake mission planning and commercial applications, such as Acses and Bottomdollar shopping agents, which search numerous online shops for books and CDs and provide users with information about the availability and price of specific items they require, thus facilitating comparison shopping.

Intelligent agents can be helpful to consumers, vendors and intermediaries. They can help shoppers to find a way through the labyrinth of thousands of stores and millions of products available on the Internet. Acses and Bookfinder can obtain information about the availability and cheapest price for a specific book from more than 25 Internet book stores in less than 30 seconds. Infospace and Bottomdollar provide a similar service for CDs and books. KillerApp and ComputerShopper (formerly Netbuyer) are shopping agents for computers and software products. Expedia UK and Travelocity offer a similar service for the consumer in the travel market.

Although the "intelligence" of some agent software has been questioned, several businesses that developed agents have become highly profitable. Junglee's Shopping guide and Jango's Netbot enable shoppers to search for any product by typing details of the product, style, price range or any other relevant information. They differ from the preceding agents because they search for many different products (not just a single product category - computers, books or CDs), but they only search in the databases of merchants who are partners in the system. In many ways they are little more than earlier virtual shopping malls with a user friendly product information search system. However, both of these agents attracted commercial interest. Junglee is now part of the Amazon.com group and Jango was purchased by Excite for $35million in October 1997.

The capability probably exists now for agents to be created which will provide a one-stop-virtual intermediary able to search for any product from any online shop (not simply partner stores) on the Internet. Intelligent agent software could provide potential purchasers with a very stark, generally price dominated, basis for comparison of one company's offering with numerous other online suppliers. This is probably a frightening prospect for many online stores.

However, intelligent agent software can also be used by online stores to target content to their users based on their preferences and behaviour. The benefits of this "personalisation" are thought to be more site visits, higher rates of return visits and greater customer loyalty. Companies which use agents in this way might also be better able to differentiate themselves from competitors.

Agents can be used by vendors to develop profiles of users through the completion of questionnaires or through monitoring site visitors' habits and interests as they browse, shop or communicate. Over time these agents can build an ever-expanding database of information about customers' tastes, habits, needs and preferences. This information can be utilised to build better relationships with users and target adverts, special offers and discounts more effectively. These rich profiles of user transactions and interests enable interactive marketing to be targeted at very small niches or even types of individual.

Examples of these types of agents include relatively simple systems, such as Firefly that when first developed asked users to complete a taste-defining questionnaire. It then made recommendations about other items users may be interested in on the basis of the tastes of like minded previous users. Firefly has developed a Passport system that maintains the privacy of users' information and automatically provides personalised content at a number of Web sites, such as BarnesandNoble.com (an online bookshop) and the Yahoo! search engine.

Some agents do not require users to complete a questionnaire. Instead they monitor site visitors' habits and interests as they browse, shop or communicate. The best examples are Net Perception's Recommendation Engine and WiseWire. Net Perception technology was used by Amazon.com and CDnow and is still used by iVillage (a women's interest site). Some developers have acknowledged the threat that users may feel to their privacy. WiseWire, now affiliated with Lycos, has stated that it will not pass on contact information (names and addresses) and Firefly now markets the privacy component of their product strongly.

These agents could provide online stores with the flexibility to target different types of customer from a single site. Unlike the terrestrial world it may not be necessary to have multiple outlets to cater for different customer groups. One store could be able to appear in distinct ways to different users.

Paul FoleyDirector of the International ElectronicCommerce Research Centre, De MontfortUniversity, Leicester, and Visiting Professor,School of Accounting and InformationSystems, University of South Australia.E-mail: pdf@dmu.ac.uk

Web Sites

A selective review of Web sites that provide details about agents and other sites mentioned in this paper (see Table I). Dynamic links to these pages can be found at the De Montfort University International Electronic Commerce Research Centre home page: http://www.dmu.ac.uk/ln/ecommerce/ebr5.html