Smart Agents Do Deals On The Net

Intelligent software agents which trawl the Internet getting the best deals for both buyers and sellers may soon be a reality.

CSIRO Artificial Intelligence (AI) researchers Dr Maria Lee and Dr Ryszard Kowalczyk have been working on ways of applying artificial intelligence techniques to support e-commerce.

"Say you are looking for a particular make and model of car in a particular colour with a particular set of features," says Dr Lee. "All you will need to do is to tell your software agent what you want and it will locate a dealer, negotiate the best price and close the deal in a matter of minutes."

For buyers who like to stay in control, the agent will suggest the best options and leave the final choice to the buyer.

"This is the logical next step in e-commerce," says Dr Kowalczyk. "At the moment buyers just choose from what is on offer and pay the advertised price, which is often higher than the in-store price."

Intelligent agents, programmed with their owners' personal preferences and requirements, will not only obtain the best possible price but also ensure that other requirements, like colour, warranty, trade-in, delivery date and after-sales service, are met.

Dr Kowalczyk says that people often have difficulty weighing up choices when the number of issues increases.

"'Should I get the green one if it's a bit cheaper? What's more important to me, the sporty wheels or air-conditioning?"

"By allocating an importance rating to each feature and setting acceptable limits on variation, the buyer can be sure the result will be satisfactory."

Agents will keep negotiating and trying new strategies until they deliver the goods. They don't get tired or frustrated so they are always at their best in a negotiation.

Software agents will also have the potential to simplify the complexities of supply chains by ensuring that all the parties along the chain have their requirements met.

This revolutionary idea is built on some key Artificial Intelligence principles.

Coalition formation means that agents get together to increase their bargaining power when it is in their interests to do so. An agent will only join a coalition if the terms are to its advantage, so agents need to negotiate the formation of coalitions.

Another big plus is that the agents can learn from previous negotiation experience. If a particular strategy results in a favourable outcome, the agent will learn to use it in preference to a strategy which has not worked as well.

They can also learn to adapt their strategies according to the situation and even based on prior experience with a particular buyer or seller agent.

Developing a system which can work in practice requires some complex problems to be solved.

"One of the big issues is how agents communicate," says Dr Lee. "Agents need to speak the same language when they talk about concepts such as colour, length, currency and time."

"We are developing systems which free agents from concerns about units and conversions and which provide a uniform language for negotiation."

The use of intelligent agents is not restricted to retail commerce.

CSIRO Business Manager Kevin Cryan foresees uses in a wide range of sectors.

"We would expect that intelligent agent technology will soon become an important aspect of a wide variety of e-commerce applications in areas such as finance, health and the delivery of commercial and government services."