If you have made a purchase at Amazon, you would have noticed that as soon as you complete your purchase procedures, Amazon presents you with a list of books that customers who purchased the same item as you did also bought along with this item. During your next visit, you will find a list of books that are similar to the ones you purchased previously that you might be interested in buying. Another popular site eBay displays a list of similar products that other customers who did the same search looked at. The above two sites are examples of two ways of implementing software agents.
A software agent is a computer program that is capable of intelligently and autonomously executing a given task based on the instructions provided by the user. It basically operates on behalf of its owner. The goal of a software agent is to maximize the owner’s interests which it does by maximizing a particular utility function. Most software agents can learn from experience and adapt themselves based on the feedback obtained from the environment thus making it an autonomous process with no human intervention. A multi-agent system involves multiple software agents.
Software agents are ideally suited for a wide variety of applications, in particular, process and workflow automation, electronic commerce, distributed problem solving and Internet applications (Acronymics, Inc. 2004). Some application examples include agent-based supply chain management, personal or user agents, e-business systems, e-market, e-banking, and e-investments, etc. Merchandise management in the retail industry is one area of application of the software agent. Merchandise management involves selection of desirable merchandise, disposal of slow-selling ones and ordering and distributing them appropriately.
The competition is fierce with the eCommerce making it even more difficult due to diverse customer preferences low merchant loyalty. The only way to tackle this problem is by understanding the changes in customer demands quickly and responding appropriately through merchandise management. However, due to the sheer number of merchandise and brands, managing merchandise effectively becomes an impossible task. This is where software agents are helpful by performing the merchandise managers’ jobs autonomously, continuously and efficiently.
These software agents help in evaluating and selecting merchandise and predicting seasons and building purchase schedules autonomously. An effective merchandise management helps reduce the inventory level while increasing sales and profits. (Jae Heon, P. , Sang Chan P. 2003) Rob Meijer describes the following commercial applications in his article: 1. Watcher agents – These autonomously look for the specified information and provide personalized versions based on the end-user’s preferences, an example is personalized versions of newspapers.
To generate a personalized paper, a user must first provide with relevant personal information, preferences, and special interests. The agent then provides personalized news based on these details. With each visit, the agent improves and provides personalized newspaper without much human intervention. Fishwrap (MIT) and Newshound are examples of Watcher agents. (Meijer, R. 2005) 2. Learning agents – Based on the behavior of their user, they can learn to tailor their performance. Examples of the same are Firefly, Similarities Engine, WegHunter, Open Sesame, and InterAp.
Similarities Engine was a Web-based music recommending system that worked similar to Amazon’s recommedations of new books; the technology has now been purchased by Microsoft. (Meijer, R. 2005) 3. Shopping agents – These agents are capable of comparing and finding the best price for an item. BargainFinder is an example of Shopping agents that searches the inventory of vendors on the internet to find the lowest prices on the desired product helping you save both time and money. (Meijer, R. 2005) 4.
Information retrieval agents – These agents search for information intelligently and are capable of summarizing the information as well. Netsumm is an example that can provide an abstract of the most important sentences of a web page. (Meijer, R. 2005) The Intelligent Software Agents Lab (The Robotics Institute – Carnegie Mellon University) has created an information retrieval agent for Aircraft Maintenance. Considering the searching of relevant information along with time constraints, the mechanics are provided with wearable computers based on RETSINA infrastructure.
When a discrepancy is encountered, all he needs to do is fill out a form on his computer and the agents would seek relevant information. “The advantages of wearable computers with agents include automatic location and retrieval of information relevant to repairs, utilization of historical repair data, increased efficiency of access to information from manuals, and reduction in average time for repair. The overall result is timely, quality maintenance. ” (The Intelligent Software Agents Lab, 2001-2009)
Another example on their site is WebMate which is a personal digital assistant that “provides URL recommendations based on a continuously updated user profile; offers ever more relevant web documents based on the ‘Trigger Pairs Model’ approach to keyword refinement; responds to user feedback by selecting features from documents the user finds relevant and incorporating these features into the context of new queries; compiles a daily personal newspaper with links to documents of interest to the user. ” (The Intelligent Software Agents Lab, 2001-2009)
5. Helper agents – These are generally used for network management and standard maintenance functions. LANAlert is an example of Helper agent. (Meijer, R. 2005) Agents can also help automate the bargaining process i. e. based on the user preferences, products and services along with related conditions (like warranty and delivery time) can be made flexible to meet the client’s expectations. (Gerding, E. H. 2004) Google AdWords is an example of such an agent. On typing search keywords, it provides recommendations to users as sponsored links.
First of all, there are limited ads that can be displayed on a search result page. The ads are displayed only on certain keywords selected by the advertiser. Also, since limited ads are there and also the order of displaying the ads (from top to bottom) makes a difference, advertisers have to compete for these spaces. This is done by choosing the best keywords based on the information provided by Google regarding search counts and setting an upper limit on the amount that you would pay if you have a visitor. The advertiser with the highest bid wins in this case.
Gary Anthes provides the examples of the following companies who have been successful in utilizing software agents for complex processes: • Procter & Gample Co. (P&G) used software agents to convert their supply chain management into a supply network connecting to 5 billion consumers in 140 countries saving $300 million annually. (Anthes, G. 2003) • Southwest Airlines Co. optimized cargo routing using software agents. • For Air Liquide America LP, software agents helped reduce both production and distribution costs. (Anthes, G. 2003) • Software agents helped Merck & Co.
find more efficient ways to distribute anti-HIV drugs in Zimbabwe. (Anthes, G. 2003) • Simulation of buyer preferences using software agents helped Ford Motor Co. come up with car options that “optimized the trade-offs between production costs and customer demands”. (Anthes, G. 2003) • Software agents helped Edison Chouest Offshore LLC to optimize service and supply vessels deployment in the Gulf of Mexico. (Anthes, G. 2003) Supply Chain Management through software agents not only help save costs but also help in inventory management through reduced inventory, and improved customer service.
(Anthes, G. 2003) Gary Anthes also explains how software agents have helped P&G. Basically, each component of the supply system from trucks to drivers to stores was represented by a software agent. Using rules, they defined the behavior of each component that represented its actual behavior such as “Dispatch truck when full”. P&G then used the simulations to perform what-if scenarios and checked out inventory levels, transportation costs and in-store stock-outs. Many alternate rules were considered such as on ordering and shipping frequencies, demand forecasting etc.
(Anthes, G. 2003) Through the use of these agent-based models, P&G realized that while sending trucks with less than full loads increases transportation costs, frequency of in-store stock-outs often resulting in lost sales is reduced. Similarly P&G relooked into many of its rigid rules and relaxed them to improve the overall performance of the supply chain. They also had to be more flexible in manufacturing – instead of one product at a time, they had to be able to produce every product every day, reducing stock-outs.
Flexibility in distributing products within 24 hours instead of the customary 48 to 72 hours was another change that software agents helped P&G realize and change. (Anthes, G. 2003) References Jae Heon, P. , Sang Chan P. 2003. Agent-based merchandise management in business-to-business electronic commerce. In Decision Support Systems (Volume 35 , Issue 3 , June 2003). Elsevier Science Publishers B. V. Amsterdam. http://portal. acm. org/citation. cfm? id=794070 This article explains how an agent-based merchandise management system can help retail companies better manage their merchandise.
Gerding, E. H. 2004. Autonomous Agents in Bargaining Games: An Evolutionary Investigation of Fundamentals, Strategies, and Business Applications. PhD thesis, Technische Universiteit Eindhoven. http://eprints. ecs. soton. ac. uk/15640/ This article explains how autonomous software agents could be used in the bargaining process and some of its applications. WebMate. The Intelligent Software Agents Lab. The Robotics Institute. Carnegie Mellon University. http://www. cs. cmu. edu/~softagents/webmate. html
This article explains WebMate, a personal digital assistant, a real-life application of software agents. Aircraft Maintenance. The Intelligent Software Agents Lab. The Robotics Institute. Carnegie Mellon University. http://www. cs. cmu. edu/~softagents/aircraft. html This article explains how the aircraft maintenance issues have been resolved using software agents. Anthes, G. 2003. Agents of Change. In Computerworld. http://www. computerworld. com/action/article. do? command=viewArticleBasic&taxonomyName=Software+Development&articleId=77855&taxonomyId=63&pageNumber=1
This article provides examples of successful implementation of software agents by several companies and how it has helped them improve their processes as well as their thinking. Meijer, R. 2005. Intelligent Software Agents: Perspective for business. In The IPTS Report (Issue 5). http://ipts. jrc. ec. europa. eu/home/report/english/articles/vol05/art-it1. htm This article provides types of commercial applications of software agents along with implemented examples. Acronymics, Inc. 2004. Why, When, and Where to Use Software Agents. http://www. agentbuilder. com/Documentation/whyAgents. html This article provides an overview on software agents.