Box Inc. (NYSE:BOX) CEO Aaron Levie said the future of artificial intelligence will be driven by multi-model agent systems that combine advanced reasoning models with lower-cost AI models to improve efficiency and expand enterprise adoption.
Levie Highlights AI Agent Cost Breakthrough
On Monday, in a post on X, Levie highlighted the growing importance of multi-model AI agents, arguing that companies will increasingly rely on different AI models for different stages of complex tasks.
"Multi-model agentic systems clearly are the future," Levie wrote.
He added that Cursor’s research found that using a frontier model as a planner and orchestrator alongside a cheaper workhorse model could deliver a "15X cost improvement" by reducing total token usage.
Levie explained that the most powerful AI models should focus on tasks requiring deeper reasoning, such as breaking down problems, making design decisions and evaluating trade-offs.
"Few moments in a large task genuinely require frontier intelligence, such as the original decomposition, the design decisions, and certain trade-offs," Levie quoted from Cursor’s research.
He added, "Once a frontier planner has collapsed the ambiguity into a detailed, explicit instruction, less expensive models simply have to follow it."
Levie said this approach is becoming a key design pattern for advanced AI agents because large portions of work often do not require the highest level of reasoning.
"By routing to different models based on where you are in the task, you get much greater efficiency overall," he wrote.
Agentic AI Reshapes Tech Industry
Earlier, AWS CEO Matt Garman said the company was advancing AI-powered software through new products, including AI assistants and industry-specific applications, while expanding access to models from OpenAI, Anthropic and Meta.
He said agentic AI could transform productivity and how businesses build and use applications.
Meanwhile, Marvell Technology said agentic AI could fuel a new wave of infrastructure spending as AI systems increasingly interact across large-scale workflows, boosting demand for networking, optical connectivity and custom silicon.
Microsoft CEO Satya Nadella also pointed to a major shift toward "agentic computing," with the company’s AI business surpassing $37 billion in annual recurring revenue and focusing on AI-powered systems across productivity, coding and security.
Photo courtesy: Shutterstock
Login to comment