This research paper centers on enabling older adults to use digital-based healthcare applications effectively and independently. It delves into instructional methods tailored for aiding older individuals in dementia prevention activities through digital platforms. The core of this investigation is to assess the impact of employing virtual agents with distinct therapeutic and peer roles, coupled with demonstration-based instructions, on enhancing the ease of use and understanding for the elderly. By employing a 2 (demonstration: demonstration vs. no demonstration) x 2 (split persona: one vs. two) between-subjects design, the research examines the effects on elderly participants' task comprehension and their attitudes towards agents and the application. Key findings reveal that demonstrations significantly enhance both perceived and actual task performance, as well as ease of use of the applications, while leading to a more human-like perception of the agents. Conversely, the use of multiple agents, divided by roles primarily impacts perceived difficulty without significantly improving performance or attitudes towards the application. These results suggest that while demonstrations and agent design are crucial in addressing the digital challenges faced by the elderly, the role of agent persona needs careful consideration to avoid cognitive overload. The study underscores the need for intuitive, user-friendly digital health solutions for the elderly, emphasizing the importance of demographic factors like age and gender in future design and research. In conclusion, this research contributes to the optimization of digital healthcare applications, aligning with user-centric approaches for improving the quality of life and independence of older adults.