Google makes breakthrough in efficiency for AI agents

"Test-time scaling for AI agents is increasingly shifting from longer thinking to controlling tool calls. In many practical applications, such as web search and document analysis, the number of external actions determines how deep an agent can dig. Each tool call increases the context window, increases token consumption, and incurs additional API costs. For companies, this can quickly add up."

"As a first step, the researchers introduce Budget Tracker, a simple module that continuously informs the agent about the remaining budget. This approach works entirely at the prompt level and does not require retraining. The agent receives explicit signals about resource usage and can adjust its strategy accordingly. In Google's implementation, the tracker also includes guidelines that indicate which behavior is appropriate for different budget levels."

Agentic AI faces scaling limits because external tool calls increase context window, token consumption, latency, and API costs, constraining depth of search and analysis. Many agents lack awareness of available compute and tool-call budgets, causing them to pursue leads too long and waste budget on dead ends without quality gains. Budget Tracker is a prompt-level module that continuously reports remaining budget and provides behavior guidelines for different budget levels, requiring no retraining. Agents receiving explicit budget signals can adjust strategies to control tool calls. Experiments with ReAct-like search agents show reductions of over 40% in search calls, nearly 20% in browse calls, and more than 30% total cost savings.

Read at Techzine Global