Researchers from MIT and Microsoft have developed a system called Murakkab that reduces the energy and computational cost of agentic AI workflows, addressing a growing constraint on cloud providers. In tests, the system used as little as 35% of the compute resources and 27% of the energy compared to other methods, while maintaining accuracy.
How Murakkab optimizes workflows
Agentic workflows combine multiple AI models and external tools to complete complex, multi-step tasks, such as analyzing video or generating code. These agentic AI workflows are becoming critical for cloud providers, but their fragmented design often wastes computation, energy, and money.
Murakkab allows developers to describe an AI application in high-level terms instead of manually selecting every model, tool, and hardware configuration. The system then identifies suitable components, decides which steps to run sequentially or in parallel, and picks hardware resources for cloud deployment. It also gives cloud providers more visibility into workflows, enabling IT and development teams to allocate compute resources more efficiently across tasks.
Efficiency gains in tests
In tests involving video-question-answering and code-generation workflows, Murakkab met user requirements while using about 35% of the computational resources required by other methods. It consumed roughly 27% as much energy and cost less than 25% as much as the comparison approaches.
In one case, the system reduced energy consumption by more than an order of magnitude with only a 2% drop in accuracy. The researchers plan to expand Murakkab to more complex workflows and larger computing clusters.
Why this matters for IT and development
Agentic AI systems are growing more complex and resource-intensive. Murakkab shifts the focus from optimizing individual models to optimizing entire AI workflows and their cloud deployment. This matters because energy use, compute costs, and data center capacity are becoming central constraints on AI growth. For IT and development professionals, tools that reduce infrastructure costs without sacrificing accuracy directly address budget pressures and sustainability targets in cloud-heavy environments.
Your membership also unlocks:
