AI Will Make Code Cheap, But It Won’t Shrink Most Engineering Teams

As AI agents increasingly generate code, many technology leaders wonder if this will render human engineers redundant. However, Winter CTO Rajesh Rajagopalan argues this assumption is likely incorrect for most enterprises.

In 1865, English economist William Stanley Jevons observed that more efficient coal engines didn't reduce coal consumption but instead expanded its use across more industries. This 'Jevons paradox' appears to be playing out in software development, where AI has made code production significantly cheaper without reducing the need for engineering work.

The Forgotten Project Backlog

Every enterprise maintains two project backlogs: the official one and a much larger unofficial list of initiatives that were never pursued due to cost or time constraints. These include internal tools that were put on hold, modernization efforts perpetually delayed, and countless spreadsheet workflows maintained by analysts.

With AI reducing development effort by approximately 70% in modernization programs, these previously unfeasible projects are becoming viable. At one client, an agentic order-entry system built for one division proved so effective and inexpensive that a second division independently funded its own implementation shortly after.

More Software, More To Manage

Every piece of software created requires ongoing maintenance, security updates, and operational support. As more software gets built, the operational burden grows rather than shrinks. Much of this maintenance work will also be handled by AI agents, but humans must still design these systems and define exception handling protocols.

Shifting Engineering Focus

As code generation becomes cheaper, the valuable engineering work shifts toward specification, validation, and architecture. Engineers must understand business rules deeply enough to translate them into requirements, determine how new services integrate with existing ones, and take responsibility when systems fail outside normal business hours.

While total engineering demand may grow, specific roles could diminish. The key differentiator between valuable and obsolete skills lies in their durability—understanding business processes and judging quality remains essential, while framework-specific knowledge becomes less critical. When asked about team sizing, Rajagopalan suggests first considering what projects would be pursued if development were truly inexpensive.