In the current socioeconomic context, IT sector companies can be interested in the diversification of their IT service offerings. The discovery of new markets and innovation potentials for existing software products is a challenge for IT SMEs, since products shall be adapted to these new targets. Hence, costs and risks of the production of adapted product variants can be really high. More particularly, a classical mean to explore the pertinence of a product variant consists in the confrontation of possible end-users to the user interfaces (UI). Yet, the current UI design practices are not industrialized enough to be able to develop and test efficiently product variants UI on users. Thus, the overall diversification of the service offerings could not be afforded by SMEs. Being able to explore a product variant innovation potential and its adaptation to a variety of contexts raises two important challenges: 1) the identification of variability and the combination of potential products 2) production time reduction of a given product.As a result, we propose an integrated approach that combines both adaptation and innovation providing semi-automatically a set of product prototypes. Our approach is based on model-driven Software Product Lines (SPL) which advocates the use of variability modelling in order to cope with the combination of situations (context and innovation). Thus proposing a set of products (said a product family). We rely on Model-Driven Engineering (MDE) to considerably reduce the production time of our UI variants without requiring developing each variant. MDE is a generative approach that relies on abstract models, which conforms to metamodels. In a previous project called GENIUS, we have experienced the efficiency of MDE for generating UI automatically from high level abstraction models (e.g., task models, kind of organized interaction requirements models).We combine together MDE for UI generation and the SPL approach to provide the set of products adapted to a given combination of situations. We consider important to produce prototypes that are aligned with the real end-user needs and truly usable. We propose to follow a user-centered approach that involves usability assessments and tests of product variants.Our methodology is about reconsidering the classical MDE SPL approach introducing UI specific properties and generation algorithms. In addition, we want to get the minimum of products covering the maximum of given contexts by identifying similarities between products. This phase will be followed by automated testing of our UI regarding functional and ergonomic properties. The formalization of human aspects related to ergonomics is an important scientific challenge that goes beyond this project but we expect being able to formalize some of them (for example the interaction length). Regarding the properties that we cannot assess by this way, we propose a standard ergonomic assessment.At each major step of our method, we are able to go back to the initial step, for calibration purposes, so as to improve models, transformations and to test algorithms. For instance, we can reconsider the similarity algorithm if we get too much or not enough products after this phase.Finally, our product family will be evaluated in a real work environment by professionals. The objective is to validate our method and our MDE-SPL tool suite by evaluating acceptance in real contexts.