We consider a domain where each agent is an expert in a particular task type and can ask other, expert agents, to perform tasks for which it is not an expert. Agents are self-interested and respond favorably to requests for help only if the requesting agent is estimated to provide reciprocal benefits. It has been shown that self-interested agents can develop mutually beneficial cooperative relations with other like minded agents of complementary expertise in such domains. Previous work in the area presented a mechanism for forming coalition based on previous interaction history and expected future interactions. One constraint of that work was the assumption of fixed agent expertise. In a dynamic environment with continuously varying task distributions, however, agents will have to change their area of expertise to increase profitability and maintain competitiveness. The agent maintains a successful coalition till it is profitable. In this paper, we present an adaptive mechanism for choosing task expertise that estimates the likelihood of forming beneficial coalition with agents of complementary expertise allowing the agent to improve its utility. We augment this decision mechanism with an adaptive exploration strategy to improve robustness. Based on the experimental results, the new adaptive mechanism is shown to be more effective and responsive to the changes in the environment than other non-adaptive strategies.