Corporations and government agencies that use Avatars claim there are substantial benefits for using them in their respective organizations; including 24/7 service availability, quick answers without a phone queue, and improved consistency in the responses provided. “There are also potential cost savings by having an Avatar answering questions compared to using personnel” (Lind and Salomonson, 2006). However, these benefits may not be great enough as the lack of possible human communication may lead to alienation between individuals and organisations. Furthermore, a robot may “miss out” on business opportunities that a human would act on. A robot will not hear and understand nuances in speech, with the risk that a potentially problematic situation may not be adequately resolved, leading to dissatisfaction with products and services delivered. Many companies measure the satisfaction with Avatars by analysing question and answer logs to see if the Avatar appears to give satisfactory answers. Few of these companies have actually asked their customers (e.g. IKEA and SAS until recently) what they really feel about the quality of the answers they receive. User Centered Design, Participatory Design and other methods are the preferred ways of developing such systems, but these do not include all stakeholders. This thesis addresses this exclusion of all stakeholders by applying a co-design research approach for developing avatars for e-Services. Case studies from Mark Municipality, Sweden and Scandinavian Airlines Systems (SAS) are presented in this thesis showing how improvements of service quality aspects with Avatars can be managed by applying a four-step Co-Design research approach. From the first step of Co-Design, through interviews, log analysis and a channel survey, findings show that the failed dialogues with Avatars Eva (SAS) and Elin (Mark) are mainly concerned with five factors: interactivity; dialogue capability; consistency; knowledge; and synonyms. In the second step of carrying out customer workshops, a number of ideal scenarios are suggested for the Avatars to perform better. In the third step, SAS decision makers decided to implement the first three scenarios: Eva’s synonyms, knowledge and consistency. Mark decision makers decided to shut down their Avatar Elin, as they did not believe they had the necessary resources. In the fourth step, another channel survey was carried out for SAS as well as a new log analysis in order to know the impact of the redevelopment of the above three scenarios. An important result of the study was that the company adopted the continuous use of Co-Design as an approach to continuous improvement of the service quality performed by the Avatar Eva. This, for example, led to an increase of 14 percentage points on the users overall satisfaction level. The results also open a new set of questions framing the relation and transformation between Co-Design as a research approach for knowledge creation and Co-Design as a method for innovation and service quality improvements. This thesis also presents an Extended Co-Design Model, which illustrates how Co-Design inspires SAS staff. In addition, the staff of the supplier of the Avatar use it for other functions within and without SAS.