How Agentic AI Accelerates Code but Complicates Software Development

Agentic AI is transforming software development by accelerating code generation and boosting team productivity. However, as Joe Bertolami, CTO and co-founder of Clifton AI, points out, it does not address the industry’s fundamental problems: ambiguous requirements, integration complexity, and real-world system support. Instead of reducing development time, AI reveals new challenges, such as the need for rigorous generated code validation and quality management.

Companies adopting agentic AI face risks of uncontrolled costs and operational disruptions. Examples from Uber and an unnamed company that spent $500 million on unoptimized AI processes highlight the importance of strict financial and technical oversight. Without centralized management, AI experiments can lead to duplicated work, rising expenses, and security vulnerabilities. Special attention must be given to access control for AI agents to prevent loss of control over critical systems.

Successful implementation of agentic AI requires a strategy that includes multi-model approaches, performance measurement based on business metrics, and redefining engineers' roles. Instead of traditional metrics like lines of code or task completion speed, companies should evaluate business impact: user retention, failure rates, and code longevity. Engineers must shift from writing code to managing systems and AI agents, demanding new skills and performance evaluation approaches.

Reducing headcount before adapting strategies to new realities can lead to long-term issues. The goal is not to shrink teams but to transform them for more strategic tasks. Agentic AI is a tool that enhances engineering capabilities, but its effectiveness depends on proper management and understanding of its limitations. Without this, companies risk facing increased technical debt, failures, and unexpected costs.