Improving the speed and energy-efficiency of AI agents
June 25, 2026
A new system, known as Murakkab, optimizes the design and deployment of multistep workflows that power AI applications.
Agentic workflows are artificial intelligence-powered software systems that chain together multiple models and external tools to tackle complicated tasks, like analyzing a video and answering questions about it.
But the way these highly fragmented systems are designed and deployed often causes inefficiencies that can lead to wasted computation, energy, and cost.
To improve efficiency, researchers from MIT and Microsoft developed an intelligent system that streamlines the process of designing agentic workflows and automatically optimizes how those workflows are implemented.
With this new method, a developer can describe what they want the agentic workflow to do in plain language, without needing to specify all the details of their application in advance.
The system automatically figures out the best models and tools to use, as well as the ideal hardware configuration and computational resource allocation when the workflow is executed by a cloud provider.
It adjusts those configurations on the fly based on each user’s priorities, such as minimizing costs or maximizing speed.
When tested on several agentic workloads, this new system reduced the number of computational units needed for deployment, significantly cutting energy requirements and costs compared to traditional approaches without hampering performance.
“Agentic workflows are getting very complicated and quickly becoming the backbone of what cloud providers are doing. Energy usage is a huge concern, so we need to be very careful about how efficient these workflows are,” says Gohar Chaudhry, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on this system.
He is joined on the paper by Adam Belay, an associate professor of EECS and a member of the MIT Computer Science and Artificial Intelligence Laboratory; senior author Ricardo Bianchini, technical fellow and corporate vice president at Microsoft Azure; and others at Microsoft Azure. The paper will be presented at the USENIX Symposium on Operating Systems Design and Implementation.
How Murakkab works
With this new system, called Murakkab (an Urdu word that means a composition of things), the researchers sought to optimize the entire agentic workflow process.
First, Murakkab enables developers to create an agentic workflow by describing their intent for the application in high-level terms, rather than detailing how the many components of that workflow should be combined.
Murakkab takes the developer’s straightforward specifications and automatically identifies the best existing models and tools to put together into the workflow. It also determines which components need to run sequentially and which can be run in parallel to boost performance.
When the cloud provider deploys that application for a customer, Murakkab optimizes the workflow by configuring its components to meet the user’s constraints, such as prioritizing accuracy while meeting a latency requirement.
It adaptively identifies ideal hardware allocations and deployment schedules to maximize efficiency in real time, then generates a workflow that is ready for the cloud provider to execute.
Results
When tested on diverse agentic workflows for video Q&A and code generation, Murakkab met user requirements while using only about 35 percent of the computation required by other methods. It consumed only about 27 percent as much energy for less than 25 percent of the cost.
The dynamic nature of Murakkab also enables users to balance tradeoffs. In one instance, the system lowered energy consumption of an agentic workflow by more than an order of magnitude with only about a 2 percent drop in accuracy for the customer.
This research was supported, in part, by the Semiconductor Research Corporation and the U.S. Defense Advanced Research Projects Agency.
Paper: “Murakkab: Resource-Efficient Agentic Workflow Orchestration in Cloud Platforms”