Skip to main content

Virtual Sales Assistant for the complete Customer Service Process in Digital Markets


Customer consulting and excellent service are key factors for the success in nearly all business domains. Especially in the arising age of digital marketplaces service becomes essential not only for SMEs. But service is difficult to implement with state of the art e-commerce concepts and systems, which are limited to the set up of catalogues with more or the less smart search engines.

The overall objective of the ADVICE project is the development and real-world testing of an intelligent virtual sales and service system beyond simple product listing or intelligent product search. ADVICE offers intelligent product advice and guides through the selection of products, instructs the application of products and provides step-by-step solutions for technical problems.
The system is designed for the consulting about craftsman tools, but the architecture is designed to be as flexible as possible to enable the adaptation of the system to other products or languages.

Existing "smart" systems limit consulting to intelligent products search by case-based reasoning or offer "reactive" dialogues based on reaction to keywords in the user dialogue. ADVICE develops a knowledge based multiagent system, which contains detailed knowledge on the products. Customers can communicate with the system using text input. The system advisor explores the needs of the customer and explains the products or at after sales service provides product application examples. The knowledge modelling technologies to be used in ADVICE are characterised by flexibility in the definition of the conceptual contents, the accessibility of the knowledge model components and the capability to support advanced user-system dialogues, with explanations and alternative conclusions.
The agent-based distributed architecture will provide the supporting intelligence for the virtual assistant. This enables the assistant to understand the needs of the potential customers and offer the most appropriate products according to these needs and the evolution of the user-system dialogue. It provides the possibility of adapting the selection of the products or services to be offered not only to the specific aspects of the product declared by the user but also to the characteristics of the user (experience, sales record). In this way, a higher level of customer satisfaction can be reached as well as an improvement in the sales results. This experience can be used to design models of customer behaviour that can be incorporated into the knowledge model, improving in this way the intelligent performance of the system.
To be prepared to the future the ADVICE system is multimodal (using text, graphics, speech), so that is accessible from a broad variety of devices, like the Internet, I-TV or mobile devices using the wireless application protocol WAP.

Work description:
The ADVICE system, is an agent-based system that simulates the competence of a human assistant in this activity domain. To achieve a flexible and coherent performance of the system (intelligent system behaviour) within the problem scenario and during customer interaction, a knowledge-based approach is selected in order to support an independent management of available knowledge and data. Several AI technologies (rules, frames, constraints) can be integrated to provide the optimal solution for each problem. This solution has been successfully used to develop systems in specialised domains where humans perform tasks based on their problem-solving experience (medicine, mechanical design, financial decisions, etc) as in FLUIDS, KITS (European funded projects) and will now be adapted to electronic commerce scenarios.

The ADVICE software architecture needs to include domain-dependent knowledge to offer the assistant services in which the ADVICE system is specialised. In order to support human-computer-interaction through natural language, ADVICE system has to include Linguistic Knowledge.
In the particular case of the ADVICE domain, a knowledge-based solution is also useful to model the different types of problem-solving strategies:

. During customer interaction: To manage dialogue-based interaction, considering well-known dialogues in such scenario or previous experiences (conversation models or dialogue scripts), client profile (frame of features), kind of explanations needed etc
. During assessment: To search for appropriate products according to the user needs, considering factors such as the client preferences, general characteristics of the products at different levels of abstraction, relations between products and preferences, constraints about certain product configurations, market strategies, etc.
. During post-sales services: To search the appropriate explanations according to the user questions or identified needs from conversation, etc.

The virtual sales assistant focuses on an intuitive, "fluid way" of interaction between the user and the system. The virtual sales assistant will make use of the virtual reality mark-up language (VRML), which allows a free movement of the assistant. ADVICE will develop a plug-in-free architecture for the execution of the assistant.
A presentation manager will handle the multimodal interaction with different devices.
The wireless ADVICE application will make use of the wireless application protocol (WAP) and implement a dynamic WAP service server.
To achieve platform independence Java is used for most of the system components on the client and server site.

ADVICE is targeting the challenges of the electronic commerce for the first time in parallel from four points of view:

. New concepts in artificial intelligence and knowledge based systems using distributed multi-agent systems
. Tight and economic straight-forward integration in existing e-commerce environments like shop-systems, product databases and enterprise resource planning systems.
. New light-weight and plug-in free interface technologies using a 3D environment
. Business concepts for the use of the system not only as an entertaining add-on, but as a mean of marketing and service

The ADVICE project will develop prototype sites at Festo Spa and Tooltechnic Systems. The customers of Tooltechnic Systems will get both sales and after-sales service for high-quality tools. The after sales service will not be limited to problem solving. The system will also explain the application of the tools for special purposes. The site at Festool will focus on after-sales service. ADVICE is expected to deliver new research results on intelligent multi-agent based systems for the electronic commerce. The user interface will contain new methods for the multimodal interaction (text input/output, point and click) between the customer and the virtual sales assistant. The interaction models are validated by usability testings. ADVICE develops business concepts for the integration and the maintenance of a consulting system in a productive environment.

Funding Scheme

CSC - Cost-sharing contracts


Hansastrasse 27C
80686 Muenchen

Participants (5)

Muenchensteinerstrasse 43
4052 Basel
V. E. Fermi 36/38
20090 Milano
United States
Bond Avenue 711
93108 Santa Barbara Ca
Avenida Ramiro De Maeztu 7
28040 Madrid
Keplerstrasse 7
70174 Stuttgart