Agentic Engineering and the Future of Software

Summary

A joint essay by ModernPath founder Pasi Vuorio and Aito founder Antti Rauhala argues that AI is changing two sides of software at once: how systems are built and what they do after deployment. On the development side, the authors say the difficult problems are no longer primarily code generation, but grounding agents in a real architecture, granting limited access, coordinating multiple agents, and governing the resulting changes. Their proposed Agentic Engineering OS keeps an architectural knowledge core, requirement ledger, evidence states, traceability, approval gates, and runtime feedback; it is intended to work above tools such as Claude Code, Cursor, Codex, and Copilot rather than require one model vendor. On the application side, they describe a shift from storing and presenting information to predicting and acting, with people supervising rather than manually operating every workflow. They identify interactive human collaboration, agent-ready data infrastructure, and coherence across facts, text, vectors, relationships, and predictions as unresolved problems. Their proposed pattern combines language-model reasoning with calibrated predictions: high-confidence cases can act, middling cases draft and ask, and low-confidence cases abstain, while outcomes feed back into the system. Aito’s examples include a sales assistant and an e-commerce reference application using a predictive database, which the article says is in public beta. The authors argue that both software construction and software operation follow the same loop: ground generation in system reality, gate it before it matters, and learn from results. They present this as an analysis rather than a guaranteed company-specific forecast, acknowledge that genuinely new possibilities cannot be derived from past patterns alone, and invite readers to a Future of Software working session.