Intelligent agents are first and foremost tools which can be applied in numerous and different ways. However, Intelligent agents, in the true sense of attributed functions such as autonomy and pro-activity, do not yet exist. There are agent-like applications like Web crawlers and search engines which sometimes include collaborative filtering; in spite of these advances a software entity that combines all of these functions into an intelligent agent has yet to be developed. Still, it is only a matter of time when intelligent agents will play a decisive role in the electronic marketplace and therefore in competition. This paper explores the boundaries of what might happen in markets when intelligent agents are introduced and used by market participants. It discusses existing commercial agent-like applications and treats models on how different functions of agents could affect different market stages. Two types of markets - travel and bookselling - are examined, focusing on consumers' interests and the functionality of destination sites.

Contents

Introduction
Intelligent Agent Competition Model
The Dynamic Approach of the Intelligent Agent Competition Model
Case Study: Internet Travel Market
Case Study: Internet Bookselling
Impact of Agents on Competition

Introduction

Intelligent agents are software entities that can execute functionalities in an autonomous, pro-active, social and adaptive fashion. These functionalities include searching, comparing, learning, negotiating and collaborating. Subsequently, based on these functionalities, seven types of agents have been identified: collaborative agents, interface agents, mobile agents, information/Internet agents, reactive agents, hybrid agents and smart agents. Provisionally in electronic commerce, interface agents and information or Internet agents will be heavily used.

Theory and development on the use of intelligent agents has focused on two areas - making agents 'smarter' (that is, enlarging their functionalities) and creating a framework (an electronic marketplace) in which intelligent agents can operate. Hence in the near future, these developments may evolve into the establishment of an 'agent reality', aside from the 'information reality' (equating to the World Wide Web) and the physical reality.

In an electronic marketplace, the market process can be divided into stages. In their Consumer Buying Behavior (CBB) Model, Gutman, Moukas and Maes identify six stages: need identification, product brokering, merchant brokering, negotiation, 'purchase and delivery' and 'service and evaluation' [ 1]. Preconditions of electronic marketplaces are the establishment of ontologies (in short: unequivocal definitions of entities) and protocols or languages, such as KQML. Security and a 'transport mechanism' are also necessary. Basically, these electronic marketplaces represent frameworks to enable trade on the Internet (auctions are one practical example). These frameworks provide intelligent agents the ground to execute their functionalities with certain programmed strategies.

Competition can be seen as a function of market dynamics. Elements of these dynamics are firm-entry or -exit, price, costs, innovations, and profitability. Each element is related to the other. For example, more firm-entry will cause a pressure on prices. However, the effect of profits on competition is controversial. Although lower profitability is expected to trigger cost-saving innovations, in some capital-intensive sectors high profits may be necessary to finance certain activities such as research and development.

Intelligent Agent Competition Model

To establish the theoretical effect of intelligent agents on markets and competition, we assume that the functionalities of agents are connected to the market process, that is stages in the process. For each stage in the process we then can assess the effect on competition. This forms the basis for the Intelligent Agent Competition Model (Figure 1).

Figure 1: The Intelligent Agents Competition Model
Source: [2]

In a somewhat broader perspective, we can also examine the possible impact of intelligent agents on the distribution chain. To complete this picture, four 'autonomous exogenous factors' will be taken into consideration: taxes, jurisdiction, standardization and secure data transfer.

Agents in the Distribution Chain

It is obvious that since intelligent agents operate with a great amount of autonomy, they can play a major role in the distribution of products. Depending on their functionality, agents can act as information brokers (search and compare) or even as mature market players (negotiating and making specific transactions).

Intelligent agents, compared to traditional parties, can act without organizational implications on behalf of many users. For instance, if an intelligent agent is sent out to purchase the least expensive product, this information - once retrieved - is virtually available to all.

How can intelligent agents affect the distribution chain? Imagine a traditional purchasing agent, or group of traditional purchasing agents contacting directly manufacturers and distributors. This time-consuming process can be more efficient with intelligent agents working on behalf of a purchasing agent or group of purchasing agents, since agents act time independently and generally in an objective, non-influential fashion. Manufacturers, producers and distributors can also develop agents to provide a personal virtual relationship to customers, integrating profiles of customers into a group of marketing activities.

Intelligent agents do not destroy the role of the distributor. On the contrary, distribution to resellers and consumers will become more important as demand increases for products sold on the Web. Agents will act as catalysts, stimulating demand and heightening the importance of timely and efficient distribution.

Autonomous Exogenous Factors in Competition

Even though Internet commerce is increasing, there are some factors that will affect its development. We call these conditions "autonomous exogenous factors." We use the phrase 'autonomous exogenous' to describe those factors that can affect business on the Internet but are not an essential part of the process itself. Four of these factors are taxes, jurisdiction, standardization and secure data transfer.

Taxes

Value added taxes generally exist in a geographically limited area. Intelligent agents, searching for information based on a given set of rules and instructions, generally will ignore geography, purchasing products anywhere in the world. This potential role of intelligent agents is leading some lawmakers to study new ways to raise taxes. There have been some suggestions that new taxes will focus on production, rather than consumption of services or goods [ 3].

Jurisdiction

Intelligent agents will execute financial transactions on the behalf of others; will it always be clear who is legally responsible and accountable for their actions? As intelligent agents become mobile and autonomous, who will monitor their actions? Will it always be clear to which address an agent delivers information or the results of its actions? As encryption develops and as the functionality of agents advance, it may be difficult to monitor their actions. That poses some real legal challenges for legal systems.

Agents will not actually work in a physical reality but in an 'information reality' such as the World Wide Web. As the population of agents grows, a subset of any 'information reality' may be defined as an 'agent reality' [ 4].

Standardization

Standardization is crucial to the effective work of intelligent agents. In the framework of an electronic marketplace certain behavioral aspects of financial exchange will be standardized. For example, in traditional auctions bidding and selling follow a certain set of rules and protocols. Digital equivalents of these protocols and rules will be needed by agents as a language to communicate and compete effectively in an agent reality. Ontologies will also need to be developed so that all parties can easily understand each other in this digital environment.

Some research is currently being conducted to standardize the behavior of agents. If successful, it would be possible for agents to be autonomous in a more natural (less artificial and less predictable) online world. Protocols will be necessary for agents to not only communicate with us but with each other in agent reality.

Secure data transfer

Encryption, firewalls and other tools and devices are increasingly needed as the amount and complexity of data transfer on the Internet grows. Secure data transfer will be vital to the development of intelligent agents. Agents have to be secure from manipulation to ensure their effectiveness.

Quantum computing may reduce some of the risks of data transfer in their use of 'qubits'. A qubit can take on eight different values, by making use of its 'super-position' or the ability to represent two values at the same time. Quantum computing may allow information to be transferred with little or no risk. This technology may reduce the need for complicated protocols and encryptions [ 5].

The Dynamic Approach of the Intelligent Agent Competition Model

Strategy of consumers: priorities of demands

Consumers set certain priorities in their demand for products or services. These priorities result create a set of individual priorities in the purchasing decision. These priorities vary from one product to the other. Ultimately, these factors can be defined as

• price;
• service;
• delivery time; and,
• quality.

Intelligent agents can handle these priorities through learning by experience from a given consumer - or group of consumers - the importance of these factors.

Methods, such as the technique of 'Constraint Satisfaction Problems' (CSP), will allow agents to quantitatively and qualitatively analyze priorities. In retail markets, CSP techniques are used to encode so called 'hard constraints' such as "I am not willing to spend more than 2,000 Euros for this products" and 'soft constraints' such as "availability is more important to me than price" [ 6].

Personalogic provides one example of how an intelligent agent could consider priorities (see Figure 2). In searching for a car, an agent asks a customer to differentiate the importance of certain factors such as price, size, features, safety, technical aspects and manufacturer. For each of these aspects the customer is asked to point out their relative importance.

Figure 2: Priorities of Demand Being Recognized by Personalogic (Source: Personalogic)

Strategy of Suppliers: Functionality of Destination Sites

An Intelligent Agent is an intermediary between its user and sources of effective, utilitarian information. The effectiveness of any agent depends on not only the parameters set by its user, but also on the kind of information encountered at different virtual destinations, which are sources of information. Information strategies by suppliers will play a large part in the functionality of agents (see Figure 3).

Figure 3: The Intelligent Agent Competition Model, a dynamic approach (Source: [7])

We will now take a closer look at integrating intelligent agents in travel and book industries.

Case Study: Internet Travel Market

The travel industry is increasingly becoming internationalized with increased direct marketing and the emergence of niche markets. Much of the strength of the market resides directly with airlines, accommodators and tour operators and these parties are putting pressure on the margins of products. Tour operators and the retail sector face high costs in communication and personnel.

Communication is one of the main problems in the Dutch travel industry. Paradoxically, this industry has sophisticated networks at its disposal but these systems are not often well linked. Many travel agents retain a great deal of information about their clients on individual preferences which is not shared with other agencies. Because this information translates into business, it is usually accessible only to individual travel companies, and not the industry as a whole. In summary, communication is more of an organisational issue where economic interests play a vital role rather than an unsolvable technical problem. In this environment, intelligent agents have a great potential.

Consumers and the Internet Travel Market

The success of Internet travel sites readily indicates that consumers have adopted the Internet as a medium for the purchase of travel products. Content providers, such as airlines, accommodators and car rental agencies, can be directly approached far more easily than in the real world. Some travel agencies are having difficulties in coping with Internet-savvy and well-informed clients. What then is the added value of traditional travel agents when travel information is easily available on the Internet?

Functionality of Destination Sites

What kinds of travel Web sites are available to an intelligent agent? A quarter of Dutch travel companies offer online booking facilities. Nearly one-third (31%) offer extensive online booking information, in some cases without price details. Besides Web-based information, many companies provide opportunities to request more information on demand. For nearly 40% of the Dutch travel industry, no online booking information is available. In some cases, 8% of those companies surveyed are not on the Internet; 18% do not provide booking information. For 13% of those companies with a Web presence, booking information is provided on demand but not instantly provided on a given Web site.

With nearly 300 Dutch travel Web sites, is it possible for a consumer to find comprehensive travel information? The answer must be a firm 'no'. Travel organizations often do not give full access to all information; often only special offers are fully described and specified.

However, being an 'information-intense' industry, the travel industry offers huge opportunities for intelligent agents. Improved information searching, product brokerage and cost savings will be the major added values made by intelligent agents.

Table 1: Functionality of Dutch Travel Web Sites
measured October 1998
Source: [ 8], based on De ReisLijst, September 1998.

Facilities	N	Percent (%)
URL Not Found/Not Applicable	22	8
No online booking information	51	18
Booking information on demand	38	13
Booking information available	55	20
Booking information both readily available and on demand	30	11
Online booking available	72	25
Other	15	5
Total	283	100

There is a differentiation among the market players when it comes to the functionality of their sites. Tour operators and suppliers of accommodations, transportation or recreation offer frequent opportunities to book online.

Table 2: Online Facilities of Different Dutch Travel Sites
measured October 1998
Source: [ 9]

Online Facilities	accommodation/ transportation/ recreation	tour operators/ travel agencies/ direct sales	tourist agencies/ tourist bureaus
URL Not Found/Not Applicable	5	0	0
No online booking information	13	30	29
Booking information on demand	24	16	7
Booking information available	21	16	29
Booking information both readily available and on demand	11	10	7
Online booking available	21	24	14
Other	5	4	14
Total	100	100	100

Case Study: Internet Bookselling

Online bookselling has become a huge success. Buying books online is widely recognized as more convenient, less expensive and less time consuming. In response, the Dutch Internet book industry is in the process of re-examining itself.

Portals are playing a major role in book searching; booksellers are competing for few exclusive positions on portals. Increasingly, independent shopping sites are providing price comparisons for a broad selection of booksellers; there is some evidence that one-to-one marketing by personalized Web sites is paying off.

The Internet book market provided initially lower costs and other financial benefits to those companies that moved early to the Internet. Now, newcomers face a more difficult process in establishing a prominent place in the Internet economy. For some, prominence can be achieved with partnerships to key Internet services such as service providers or search engines.

Agents will have an impact on competition in this market because of their efficiencies in comparing prices and other factors. For many the most efficient agent will be the one that finds the lowest price for a give book. In this market, highly sophisticated agents will not be necessary.

The functionality of any given Web site will be important not only to human visitors but also for software agents. The following characteristics will differentiate functional sites from those that are less utilitarian:

• online information available to help consumers decide on a given purchase;
• information available on demand by e-mail;
• online distribution and shipping information (time, costs, etc.); Amazon gives about 8 pages shipping information; and,
• online recommendations and reviews for books to fit moods, preferences and other factors.

Effects of Agents on Competition in Bookselling

Table 3 compares the functions of intelligent agents with the actual functions of some bookselling Web sites.

Table 3: Functionality of Intelligent Agents and Bookselling Web Sites
Source: [ 10]

Function of Agents	Search	Compare	Learn	Negotiate	Collaborate
Results	available	available	not available	not available	not available
Functions of Web site	search and browse millions of books	online information on prices, delivery time, shipping costs	find books based on preferences, profiles	not possible	not possible

Potential Effects on Competition

The world wide consumption of commodities like books will create enormous competition among suppliers. Who will survive in this market? The cost of transactions may be much lower digitally compared to traditional distribution channels but marketing online services can be quite expensive. These marketing dollars will pay for

• online and off-line promotion;
• increased volume;
• creating exclusive positions with search engines, portals and service providers; and,
• developing cooperative arrangements.

These costs may make a marketplace for only a few bookstores. However, intelligent agents may completely alter the economics of bookselling in the near future.

Among Dutch booksellers and book consumers, using agents for price comparisons does not make sense because book prices are regulated. Agents may only be successful in searching for books or bookdealers based on factors other than price, such as service and stock on hand.

Impact of Agents on Competition

The main potential effects of intelligent agents on markets and competition are:

• Higher market volumes;
• Cost reduction;
• increased importance on customer profiles, rather than products.
  Basically, the impact of intelligent agents is cross-sectoral; as a result, marketing will change its focus from product marketing to 'lifestyle-marketing';
• Communications.
  The "conversations" between intelligent agents will require extensive standards development and legal parameters;
• Mass individualization of the marketplace;
• Increased market efficiencies;
• Increased opportunities for innovation;
• Increased pressure on prices.

Agents will lead to less "surfing" and random information acquisition and more focused analysis of digital sources of information. In turn, agents will alter the ways in which products and services are advertised or otherwise promoted electronically [ 11].

Functionality of Destination Sites

One might assume that the functionality of destination sites is linked with the use of the Internet within a given company. Hence a parallel assumption would argue that it is just a matter of time before fully functional intelligent agents will be able to secure any information on the Web. The functionality of agents is directly dependent on the kind and quality of information made available publicly in digital form.

As we noted in the Dutch travel industry, there are great differences in the kinds of information available electronically. Large and well-known tour operators may only put a brief description of their travel options on their Web site, while relatively small companies - often with specialized services - provide far greater information and resources on their sites. Clearly, strategies are emerging on the Web over information access by certain commercial sectors.

The ways in which companies present their products on the Internet will critically affect the functionality of agents. There is an emerging 'interdependence' between intelligent agents and the kinds of information available on specific destination sites. With the increased use of agents in the future, it is most probable that information will be more readily available electronically.

About the Author

Kees Jonkheer is a senior researcher at EIM, one of the largest independent economic research institutes in the Netherlands.
E-mail: kjo@eim.nl

Notes

1. R.H. Guttman, A.G. Moukas and P. Maes, 1998. "Agent-mediated Electronic Commerce: A Survey," Knowledge Engineering Review (June), p. 2.

2. K.R. Jonkheer and A.M. Jansen, 1998, Intelligent Agents, Markets and Competition. Zoetermeer: EIM, at http://www.eim.nl/uk/uk/intelligent%5Fagents%5F%5Fmarkets%5Fan.html

3. K. R. Jonkheer and W. Dolfsma, 1999. "Belasting in de 21e eeuw - gevolgen van Internet," (February).

4. This notion is developed in Co-Habited Mixed Reality Information Spaces (COMRIS) at http://arti.vub.ac.be/~comris/

5. http://www.starlab.org/

6. R.H. Guttman and P. Maes, 1998. "Agent-mediated Integrative Negotiation for Retail Electronic Commerce," paper presented at the Workshop on Agent Mediated Electronic Trading, (Minneapolis/St Paul, May 10), a workshop of the 2nd International Conference on Autonomous Agents (Agents'98); paper available at http://www.iiia.csic.es/amet98/paper5.pdf

7. K.R. Jonkheer and A.M. Jansen, 1998, Intelligent Agents, Markets and Competition. Zoetermeer: EIM, at http://www.eim.nl/uk/uk/intelligent%5Fagents%5F%5Fmarkets%5Fan.html

8. K.R. Jonkheer and A.M. Jansen, 1998, Ibid.

9. K.R. Jonkheer and A.M. Jansen, 1998, Ibid.

10. K.R. Jonkheer and A.M. Jansen, 1998, Ibid.

11. O. Do, E. March, J. Rich and T. Wolff, 1996. "Intelligent Agents & the Internet: Effects On Electronic Commerce and Marketing," http://bold.coba.unr.edu/odie/paper.html

Copyright © 1999, First Monday