Uploaded January 2026 | Updated September 2026, 2 weeks ago
AI Building Trust in AI Agents – Szilárd Széll (Eficode) and Katja Saarela (Eficode)
Agentic software development offers significant efficiency improvements, but replacing the entire development process with agents presents challenges, especially concerning quality assurance. We identify two major hurdles: the communication problem and the oracle problem. Implementing an interconnected system of AI agents requires careful design to create a robust and efficient process. The communication problem, inherent in all software development, persists even with AI agents, making clear prompts and sufficient context crucial. Additionally, the “oracle problem” highlights the difficulty of determining correct AI output without a clear ground truth, a challenge amplified by multiple connected agents. Our presentation will explore how these communication and oracle challenges impact software development and how they are addressed, ultimately guiding the design of a new AI agentic development process for high-quality outcomes.
AI Building Trust in AI Agents – Szilárd Széll (Eficode) and Katja Saarela (Eficode)
Agentic software development offers significant efficiency improvements, but replacing the entire development process with agents presents challenges, especially concerning quality assurance. We identify two major hurdles: the communication problem and the oracle problem. Implementing an interconnected system of AI agents requires careful design to create a robust and efficient process. The communication problem, inherent in all software development, persists even with AI agents, making clear prompts and sufficient context crucial. Additionally, the “oracle problem” highlights the difficulty of determining correct AI output without a clear ground truth, a challenge amplified by multiple connected agents. Our presentation will explore how these communication and oracle challenges impact software development and how they are addressed, ultimately guiding the design of a new AI agentic development process for high-quality outcomes.
