The implementation of AI tools in the software development process is leading not to personnel replacement, but to an augmentation effect, creating a double financial burden on companies.
What Happened
Analysis shows that instead of the expected headcount reduction (substitution), companies are facing an augmentation effect. This results in a combination of fixed engineer salary costs with a sharp rise in variable token expenses. In agentic workflows, token costs can be 5–30 times higher than when using standard chatbots.
Context
There is a risk of the Jevons paradox occurring: the decreasing cost of AI models themselves is offset by an exponential growth in the volume of their consumption within complex automated cycles. Furthermore, the increase in individual engineer productivity is often accompanied by side effects such as a 30–41% increase in technical debt and a rise in the number of incidents.
Why It Matters for the Industry
For the AI industry and businesses, the cost modeling paradigm is shifting: AI development tools should be classified not as fixed-price licensed software, but as a variable cost center, similar to cloud services (Cloud/FinOps). This is driving demand for new disciplines, such as AI-FinOps, to control inference costs and manage model utilization efficiency.
Why It Matters for Users
Developers and team leads need to implement observability tools to monitor token consumption in real-time and conduct audits of agentic cycles. Work efficiency should be evaluated not by the number of lines of code written, but by quality metrics and complexity-aware velocity to avoid uncontrolled cost growth.
Sources
Author
Look at AI, Editorial Team