AI agents face ROI test as enterprises shift focus to operating costs: McKinsey

In addition, information structure, including prompt design, context length, formatting, language and data organisation, plays a critical role in determining AI costs.As generative artificial intelligence (Gen AI) moves from pilot projects to enterprise-wide deployment, organisations are increasingly focusing on the economics of AI agents rather than the technology itself, with financial sustainability and return on investment (ROI) emerging as the next major challenge, according to a McKinsey report.The report, as seen by ANI, said the initial phase of Gen AI adoption over the past two years was centred on access, experimentation and deployment.

However, as agentic AI systems enter production, enterprises are now evaluating whether the business value they generate justifies the rising operating costs."For the first two years of gen AI adoption, most enterprises focused on access, experimentation, and deployment.

As agents move into production, a different set of questions is emerging...

The decision to scale an agent is increasingly becoming a complex and fast-changing economics decision, not a technical one," the report said.According to McKinsey, chief financial officers (CFOs) and chief information officers (CIOs) are increasingly demanding measurable business outcomes from AI investments, shifting the conversation from reducing technology costs to demonstrating tangible returns.Factors affecting operating expenditure of agentic AIThe consultancy noted that despite the rapid adoption of agentic AI, many organisations still lack robust mechanisms to measure the business impact of AI-driven decisions.

It added that nearly 60 per cent of the operating cost of agentic AI systems is spent on verifying and refining responses rather than generating them.The report identified six major factors influencing the operating expenditure of agentic AI systems.

Among them, long-lived context emerged as one of the largest cost drivers, with agentic AI tasks consuming nearly 1,000 times as many tokens as conventional code reasoning or chat-based AI applications."Per-token pricing has stopped being a useful measure for what enterprises actually pay for gen AI," the report noted, adding that context management now accounts for a significant share of AI operating costs.Challenges aheadMcKinsey also highlighted that refining AI-generated outputs is more expensive than producing the initial response, while autonomous AI systems introduce cost variability because expenses depend on the reasoning path, tools used and the number of retries required for each task.Another key challenge identified was the unnecessary use of advanced reasoning capabilities for routine tasks.

While sophisticated reasoning models deliver value for complex use cases, deploying them for simple workflows adds avoidable computing costs, the report said.The study further found that efficient agent orchestration—how AI agents coordinate with models, tools and each other—can significantly reduce operating expenses without compromising business outcomes.In addition, information structure, including prompt design, context length, formatting, language and data organisation, plays a critical role in determining AI costs."Non-English text, for example, gets fragmented into more tokens per meaning, so the same conversation costs more in some languages than others," the report said.McKinsey concluded that as enterprises scale AI deployments, the success of agentic AI will depend not only on technological capabilities but also on the ability to optimise operating costs while delivering measurable business value.