Opinion
The enterprise AI paradox: The arms race slipping out of budgetary control
Marrying a skilled AI work culture with smart cost control is the key to transforming AI from an expensive trend into a stable, secure growth engine.
The pressure to implement AI has created a budget drain and silent security threats. The solution is not to halt the technology, but to combine managed architecture, cost optimization, and a tailored, organization-wide training process.
No CTO or CEO in a modern enterprise can afford to be left behind in the AI arms race. The pressure from boards, markets, and competitors is clear: "Implement AI, and do it fast." Yet, behind the shiny POC presentations and promises of operational revolution lies a far more complex reality. This is the "Enterprise AI Paradox" where the critical need for innovation collides head-on with ROI metrics, cloud budgets, and data security.
The paradox begins to unfold the moment usage scales beyond limited pilot groups. According to a recent study by McKinsey and MIT, a troubling picture is emerging, showing that only about 5% of enterprise AI pilot programs ultimately translate into a real, measurable impact on the bottom line.
This reality stems, in part, from employees seeking shortcuts and creating a widespread phenomenon of "Shadow AI" the use of unmonitored tools that expose sensitive corporate data and breach strict security regulations. Meanwhile, CFOs and CIOs find themselves chasing mysterious budget overruns. The cost of querying Large Language Models (LLMs), cloud data processing, and network traffic is growing at an exponential rate, with no direct correlation to the business value generated. In the current state, calculating the ROI of AI projects in many enterprises is nearly impossible.
Into this crucible of escalating costs enters an equally deep paradox: the employee adoption and implementation gap. Many organizations rush to purchase expensive licenses and distribute them freely to "demonstrate innovation," only to find that a vast portion of them sit completely idle. Alternatively, untrained employees run incredibly heavy and expensive models for basic tasks that could easily be solved with much simpler, cheaper processing engines. Without a genuine alignment between the employee's role requirements and the tool provided, the organization generates a quiet but massive drain on resources, without measuring or quantifying the organizational value.
To resolve this paradox, enterprises must define a new operating model that integrates technological control with human capital development:
Cost Optimization and Complete Visibility: Implementing solutions that provide a precise, real-time picture of usage and costs across the organization - who is using which tool, for what purpose, and at what cost - to understand the true unit cost of AI within the enterprise.
Targeted Reskilling: One of the most practical and immediate steps enterprises must take regarding their existing human capital is investing in a comprehensive, customized reskilling program for all employees. When employees receive role-specific training, they learn how to correctly leverage the tools selected for them, upgrading and updating their professional skills. This enables the organization to implement precise licensing optimization - granting advanced licenses and models only to those who truly need them and generate measurable business value from them.
This exact strategy is currently being implemented at NESS, which employs over 6,500 people. As part of its technological leadership vision, the company is driving an organization-wide reskilling initiative. A significant portion of this training focuses on building Forward Deployment Engineer (FDE) capabilities. These key personnel are capable of not only developing technology but also embedding directly within client organizations, understanding their unique business needs, and building tailored AI solutions that are rapid, secure, and highly efficient on the ground.
In conclusion, marrying a skilled AI work culture with smart cost control is the key to transforming AI from an expensive trend into a stable, secure growth engine.
Shachar Efal is the CEO at Ness Technologies.
