This paper presents a multi-agent prototype for an electronic voting system based on optical character recognition technology. This prototype is analyzed and compared to other voting systems. Trade-offs in voter training, ease of use, security and coercion across various systems are considered for the purpose of recognizing achievable improvements. Security improvements and new capabilities are proposed based on the use of N-version programming techniques. This architecture can work with and improve security and reliability on most any algorithmic or architectural security approach. We describe how multiple agent-based technology can be applied to design and implement better voting systems. This case study demonstrates how the voter’s authentication system and vote data transmission could further simplify and improve the electoral process by adding these new capabilities to the electronic voting system using N-version programming.