KPMG: companies spend $186 million on AI, but see no real returns — agents change the equation

KPMG published the first issue of its quarterly Global AI Pulse research — and the main conclusion is discouraging: the gap between what companies spend on AI and what they actually get in return is growing rapidly.

Figures That Concern

According to a survey of top executives of major organizations worldwide, the average planned AI budget for the next 12 months is $186 million. These are colossal investments — but the problem is not the volume of spending, but the fact that not everyone extracts concrete measurable business value from them. Many executives acknowledge: AI has been implemented, but the results are questionable. KPMG identifies several systemic causes of the gap:

• Companies implement AI in isolated pockets without a unified corporate strategy
• There are no clear metrics to assess AI's impact on operational and financial performance
• Most projects get stuck at the pilot stage and fail to reach industrial scale
• The technology stack remains isolated from key business processes
• Employees are not trained to work with AI systems at the operational level, which negates the potential effect

AI Agents: From Pilot to Margin

The key thesis of the research: AI agents — autonomous systems capable of performing complex multi-step tasks without constant human involvement — are becoming the primary tool for generating real margin. Unlike chatbots or co-pilots, agents don't just answer questions. They act: they independently gather data, make intermediate decisions, interact with other systems, and complete tasks without constant operator prompts. It is precisely this architecture that makes it possible to automate not individual actions, but entire operational workflows — and it is here that concrete financial returns are generated.

"Organizations that learn to scale agent systems will gain a competitive advantage that will be extremely difficult to overcome,"

KPMG analysts conclude.

Corporate Playbook

KPMG proposes a specific plan for the transition from investments to measurable results. The first step is to identify processes with high volumes of repetitive tasks and clear execution rules. This is where agents deliver maximum effect with minimum risk: financial reporting, customer onboarding, supply chains, compliance checks, initial request processing.

The second element is quality data infrastructure. An agent works only as well as the data it interacts with. Without a reliable data layer, the transition to an agent model will inevitably fail — and this will be an error not of technology, but of architecture.

Finally, KPMG insists on measurable KPIs from day one. Each agent use case must have a specific target metric: reduction in operational cycle time, savings in man-hours, reduction in error frequency, increase in gross margin. Without this, it is impossible to either justify investments to the board of directors or understand in which direction to scale.

What This Means

The era of AI experiments is ending; the era of AI operations is beginning. Companies that are already building agent architecture with clear metrics are gaining real competitive advantage. Those who continue to "explore possibilities" without concrete financial goals are simply burning budgets.