Intelligent Agents & E-Commerce

M.V. Sinmao
24 November 1999

Introduction
What is an intelligent agent?
How does it work?
How is it applicable to e-commerce?
Where can I find some examples?
Getting There

Appendix

Table 1. Three Dimensions of Intelligent Agents
Table 2. Intelligent Agent Network Infrastructure

References

Introduction

The internet has experienced a rapid shift from information and entertainment to electronic commerce. The amount of information available on the web, as well as the number of e-businesses and web shoppers, has been growing exponentially and the influx is difficult to process. A major challenge for marketspace participants has become sifting through an unwieldy amount of information to find products, services, and even each other – often relegating e-commerce a hit or miss experience.

Intelligent agents are a major evolution toward solving this difficult problem. Intelligent agents empower both buyers and sellers to accomplish e-commerce transactions by enabling efficient, precise, and comprehensive searches on the vast web community and information repository. Because of user simplicity and thoroughness, intelligent agents enhance user experience and satisfaction. By operating in the background in lieu of user intervention, intelligent agents also circumvent problems related to slow internet access and free up prohibitively expensive “surf” and data mining time.

The technology has evolved and the internet has matured to a point where sophisticated new generation intelligent agents proliferate. This page examines intelligent agents and their use by buyers and sellers in the e-commerce marketspace.

What is an intelligent agent?

Intelligent agents are software applications that have a predefined knowledge base and/or learning system about their user's goals and wishes and, through adaptive reasoning, use this information to execute their user's request. The concept has been around for many decades, ranging in functionality from a simple macro with prescriptive directions to new generation software that truly exhibits learning and artificial intelligence.

The continuum of intelligent agents may be characterized along three dimensions: agency, intelligence, and mobility. (See Table 1. Three Dimensions of Intelligent Agents.) Agents typically act independently of the user, learn and adapt from actions or the environment and are highly mobile and autonomous in their actions. For these reasons, intelligent agents are more interactive and can accomplish multiple tasks at a variety of locations. This characteristic differentiates intelligent agents, making them much more powerful and versatile applications than traditional search engines, spiders, and web crawlers.

How does it work?

Intelligent agents operate according to a Remote Programming Paradigm, or computer-to-computer programming. Remote programming enables the client (user) to store on the server (host) not only the procedures but also any accompanying instructions and peripheral data. Each time the events specified in the instructions occur, the server calls the procedures and executes on-site the instructions, without any further intervention from the client computer. As directed, the agent returns the results to the client either at the end of the session with a single host or after conversing with all hosts in the target community.

The intelligent agent network infrastructure has five components working together: an execution facility, a communication facility, a transport facility, a packaging facility, and an integrated security facility. (See Table 2. Intelligent Agent Network Infrastructure.)

How is it applicable to e-commerce?

E-commerce has grown at an astounding rate over the last few years and so has the number of participants. Although web business models differ from the traditional brick and mortar models in many ways, the fundamental needs of consumers and businesses remain much the same: consumers want to comparison shop products and services side by side for the best price; businesses want to grow sales by driving the right shoppers to their sites. These needs give rise to a variety of intelligent agents working for buyers and sellers of products and services over the web.

Buyer agents

According to a recent report from Deloitte Research, online shoppers will routinely use intelligent agents to e-shop by 2002. At that time, more than 70 percent of large companies are expected to be selling goods online. Rather than getting in a site you never wanted or playing a hit or miss guessing game regarding which keyword search result might match your needs, a variety of intelligent agents can be sent out by buyers to help locate stores, brands, product or service categories, products and services, and desired prices. These capabilities will drive a shift from a web-centric to user-centric e-commerce model, whereby buyers will be empowered to comparison shop even before going in to one particular seller's site.

Moreover, the buyer agents perform the work without user intervention, thereby simplifying life while improving the end results obtainable by the user. With more efficient and powerful search capabilities, agents also are expected to help further drive down e-commerce costs and make web-shopping more transparent. These effects will make e-commerce increasingly attractive to reluctant consumers, since convenience, costs, and on-the-fly availability of peripheral information are the major hooks of the digital marketplace.

Seller agents

In turn, buyer agents affect how vendors on the web need to operate their online businesses. In a more competitive and transparent marketplace, vendors of products and services on the web will need to hone and logically organizate their information so that buyer agents will be attracted. Sellers can use intelligent agents to track demand and market share changes, engage in competitive knowledge mining, and even learn through collaboration from buyer agents. Learning agents deployable by sellers are less obtrusive and typically receive greater acceptance than filtering agents that require users to make ratings and answer lots of questions.

These capabilities mean new a era in e-commerce, not easily replicated in the brick and mortars environment: mass personalization. Agents enable businesses to efficiently gain intelligence on the market to create personalized relationships with every single one of its customers. Thus, e-businesses have the advantage of being about to use intelligent agents to present the perfect sales pitch. To do so, businesses will need to deploy a team of seller agents in a multi-agent system. Multi-agent systems comprised of buyer and seller intelligent agents are essential to open digital marketspaces, given the many dynamic roles driving transactions. Buyers may need decision agents to comparison shop, while sellers may need a broker, provider, and merchant agent to sell a product. More sophisticated seller agents also may negotiate with and learn from buyer agents.

Where can I find some?

Intelligent agents for e-commerce are not some future technology. The technology has evolved and the internet has matured to a point where sophisticated new generation intelligent agents proliferate. Some current applications of intelligent agents for e-commerce are listed below. Although the list is organized according to buyer agents and seller agents, the use of agents benefits all e-commerce participants.

Buyer Agents

Fido the Shopping Doggie. This is an early illustrative-only model of a shopping agent that searches a database of merchant pages and produces a list of product descriptions and prices, enabling one-stop shopping on the web.

BargainFinder. Agent returns the lowest music CD price found on indexed sites.

Firefly. An experimental project using an agent to study its community space and match e-commerce vendors with the people who would be most interested in their products and/or services.

YourCommand. @YourCommand has been working on consumer privacy since 1993 and developed IdentitySafe, an agent that will enable consumers to conveniently buy products and services on the Internet without revealing their
identity by incorporating tools like private messaging, private payments, and private shipments.

Seller Agents

ProspectMiner. This tool uses intelligent agents to generate sales leads by enabling marketing and sales professionals to interactively mine vast resources of web to find the most qualified sales prospects that fit a given customer or market profile.

ActiveBusiness. This eMarket Management software platform built on next generation collaborative intelligent agents architecture allows businesses to drive revenue and profit through new Internet-enabled market models (e.g., building customer loyalty, preventing brand commoditization, and avoiding channel conflict) and transform the way companies go to market and create value for their customers. ActiveBusiness comprises four components: ActiveMarket™ builds customer loyalty; ActivePartner™ helps your channel partners sell more; ActiveSupport™ helps your sales people sell more; and ActivePulse™ gives you superior business intelligence.

WebData. ExperTelligence earlier this year launched a database portal aiming to be the ultimate source for research, reference and shopping on the Web. WebData incorporates value-added features both for buyers and sellers. For example, WebData features current information on individual books from Amazon.com that automatically updates according to sub-categories any user browses. For the buyer, WebData features ExperSearch which uses an "Enter Once, and See Many Results on One Form" capability.

Getting There

For buyers and sellers to achieve full benefits of intelligent agents for e-commerce, a number of issues will need to be observed. Briefly:

Use. To see broad application, use of intelligent agents need to be free or inexpensively priced.

Design. Intelligent agents need to incorporate open design to allow continuous development. Agents also need to be compatible across platforms (Wintel, Mac, Unix, OS/2, PDA's etc.).

Repository quality. One major obstacle to intelligent agents turning the web into an e-commerce powerhouse is the quality of content across the web. Standards may be needed to be implemented to determine a site’s validity and trustworthiness.

Improved functionality. While current technology supports intelligence-gathering, advances in artificial intelligence can be incorporated so that intelligent agents can actually play a key role in decision-making process.

Agent-friendly environment. Finally, to facilitate a smooth multi-agent system sites will need to create an agent-friendly environment. Hosts can open their sites to agents. Hosts also can provide agent.txt scripts to converse with visiting agents. Developments in web security and authentication will help allay concerns about intelligent agents and support this necessary shift.

With these issues in mind, intelligent agents of the future will evolve beyond simply learning buyer/seller needs but also to analyze the results and present a user with comments on the results as well as suggestions on possible courses of actions based on the agent’s knowledge of the user’s ultimate goal. Such intelligent agent capabilities will enable buyers and sellers to turn the web into an e-commerce powerhouse.

Appendix

Table 1. Three Dimensions of Intelligent Agents

Agency: degree of independence. Agents are independent and proactive to varying degrees. Armed with the user’s directions, all agents must be able to function autonomously, without human intervention, to accomplish a mission on their user’s behalf. Agents must be able to communicate with repositories of information, as well as with other agents. This agent-to-agent communication entails one agent stating its intentions and objectives to other agents, who act as gatekeepers of information repositories needed to fulfill a user’s request. For example, an agent scouring the web for the lowest airfare will seek access to bids from a number of vendors. More sophisticated agents also possess the capacity for cooperation. Such agents are able to collaborate with other agents if the need arises. In a digital marketplace, an agent should possess a cooperative spirit to work in multi-agent systems together to perform mutually beneficial and complex tasks it can’t otherwise accomplish on its own. Vendors, for example, typically require agents to fulfill a variety of inter-related roles including broker, provider, and merchant.

Intelligence: degree of learning. Agents must be able to reason things out based on its current knowledge of its user and environments and on past experiences. Reasoning may be: rule-based, where agents use a set of user predefined conditions to evaluate conditions in its external environment; knowledge-based, where agents are given sets of data about prior situations and resulting actions upon which the agents can define its future moves; or artificial evolution-based, where agents give rise to newer generations of agents capable of carrying out even higher reasoning capabilities. By examining its environment, agents must be able to adapt their actions to achieve its goals and/or mission. An agent accomplishes this by processing information about its environment and the history, successes and failures, of previous actions taken under similar scenarios, and then adapting its next actions based on all that learned information. Thus, an agent that returns the lowest airfare is useful but an agent that learns and returns a better overall travel package (e.g., airfare, airline, travel dates, and seat reservations) each subsequent time it is used will be far superior to the e-commerce user.

Mobility: degree of travel. Agents may be static, residing and operating within one machine. Agents engaged in e-commerce must also be able to travel from computer to computer, gathering information until search parameters are exhausted. Agents that repeatedly cycle back and forth between client and repository are more secure than those that accumulate data from computer to computer but do not return to the client computer until completion.

Table 2. Intelligent Agent Network Infrastructure

The execution facility is the agent’s run-time environment, or the hardware and software needed to execute agent programs in the agent’s environment. Unlike hardware, where the choice has little bearing on an agent, the software programming language of an agent is the determining factor in the range of functionality possible in an agent application. There are three language categories for an agent: general purpose programming and scripting languages for simple applications (examples: SQL, Perl); general purpose mobile code languages for flexibility in transmitting procedures across the network for execution at the host end (example: Java); and mobile code languages for intelligent agent applications (example: Telescript). The first isn’t a true agent language. The second, taking Java as an example, wasn’t originally developed as an agent language but is a mobile code language that has more mobility and safety features than the C++ program it resembles. The third category, taking Telescript as an example, was designed specifically to exploit the possibilities for mobile intelligent agents and might be the standard for agent-based communication.

The communication facility addresses the process in which information or a message, with its inherent content and context, is exchanged in an agent transaction. This facility must be able to communicate simultaneously with different agents and must support synchronous/asynchronous communication.

The transport facility addresses what and how an agent is communicating. The facility allows for the movement of an agent from one execution environment to another and for the distribution of uninitialized agents. Movement usually occurs in an established data transmission protocol such as HTTP, SMTP, or TCP/IP.

The packaging facility deals with the encapsulation of an agent’s state, authentication, and goal information with an agent’s capabilities and method/plan information.

The integrated security facility deals with the identification and verification of an agent to determine that agent’s owner and place of origin. Security concerns include such aspects as resource protection, non-refusal of services rendered, information privacy, and agent termination rules and procedures. Each intelligent agent application usually incorporates its own security protocols.

References

Business and Industry. E-Commerce Efficiencies Could Drive Prices Substantially Lower. 23 June 1999.

Business and Industry. Inside an Online Search agent. 1 September 1999.

Business Wire. AlphaServ.com's AlphaCONNECT StockVue 99 Wins Top Rating From ZDNet; Book on Intelligent Agents Receives Praise From Online Buyers and Reviewers. 9 June 1999.

Business Wire. Exterprise Inc. Introduces New Software Category, eMarket Management (eMM) That Transforms the Way Buinesses Go to Market. 13 October 1999.

Business Wire. Intarka Extends SalesLogix's Sales Automation System by Mining the Web for Sales Leads. 18 October 1999.

Business Wire. IntelliSeek, Inc. Receives $6 Million in Venture Capital Funding. 1 July 1999.
Canadian Business and Current Affairs. Agents at your service: Intelligent agent software will give consumers the uper hand in Net transactions. 23 September 1999.

Do et al. Intelligent Agents & The Internet. bold.coba.unr.edu/Tara/paper.html

Fingar, P. A CEO's Guide to eCommerce Using Object-Oriented Intelligent Agent Technology. June 1998. home1.gte.net/pfingar/eba.htm

Hauser, R. NetWatch Top Ten - Intelligent Agents/Information Agents. www.pulver.com/netwatch/topten/tt9.htm

Miller, D. Three Types of Intelligent Agents, Their Uses, and Their Future. November 1996. www.netwcapeworld.com/netwcapeworld/nw-11-1996/nw-11-agents.html

PR Newswire. AlphaServe.com Signs Agreement with Leading Danish Web Portal; Two-year Contract Underscores Value of Intelligent Agent Technologies in Internet Marketplace. 17 November 1999.

PR Newswire. ExperTelligence Enhances WebData.com with Search Agent for Amazon.com Books; WebData.com

Now Offers Access to Most Current Information on Amazon.com Books by Category and Context.

PR Newswire. Open Sesame Intelligent Agents to Impact Internet Privacy, Intranet Development Costs, and Web Site Personalization. 16 July 1997.

Sinmao, D.V. Intelligent Agents + Internet = DSS. 1996.

Intelligent Agents Examples:

ActiveBusiness

BargainFinder

Fido the Shopping Doggie

Intelligent Agent Resources on the Web:

www.research.ibm.com/iagents/home.html

www.grailsearch.com/Intelligent%20Agents.htm

ksi.cpsc.ucalgary.ca/courses/547-95/bentley/547talk.html

www.comp.lancs.ac.uk/computing/users/gary/design.reuse/int.agen