AWS Whitepaper | Page 14

AI INFRASTRUCTURE TEAMS REPORT SPENDING

30-50 % OF ENGINEERING TIME ON INFRASTRUCTURE

RATHER THAN VALUE-CREATING APPLICATIONS

barrier to deeper AI adoption, while 42 % of companies are abandoning generative AI projects entirely 5. With measurement systems that can connect agent outputs to business outcomes, boards cannot justify continued investment.
Legacy system integration blocks agents from operating across the enterprise. Seventy-eight per cent of enterprises report struggling to integrate AI with existing systems 6. Fragmented APIs, data silos and inconsistent interfaces prevent agents from accessing the information and tools they need to execute end-to-end workflows autonomously.
Ultimately, escalating costs, undefined business value or poor risk management are set to result in the cancellation of at least 40 % of agentic AI projects by the close of 2027 7.