Engineers building modern artificial intelligence applications increasingly rely on complex, multi model workflows rather than single standalone systems. As AI evolves from isolated models into coordinated networks of specialized agents, developers face mounting challenges in manually designing and orchestrating these pipelines. To address this bottleneck, researchers at the Harvard John A. Paulson School of Engineering and Applied Sciences have introduced Orla, an automated framework that streamlines the construction and deployment of AI workflows while dynamically optimizing for computational cost, accuracy, and execution speed. The project is led by Gordon McKay Professor of Computer Science Minlan Yu and Thomas Watson Sr. Professor of Computer Science Michael Mitzenmacher, alongside postdoctoral fellow Rana Shahout and Boston University software developer Hayder Tirmazi. The initiative extends the group prior research in computer systems optimization, including scheduling, load balancing, and memory management for AI inference. Shahout, the lead author of the recent demonstration paper, explained that the team is applying established systems engineering principles to manage entire AI workflows as they transition into collaborative agent networks. In Orla architecture, engineers simply define the desired workflow in high level terms. The system then autonomously determines the most efficient execution strategy, selecting appropriate models, allocating resources, and coordinating task sequencing without manual intervention. This approach eliminates the traditional trial and error process of custom building pipeline structures. During testing, Orla demonstrated measurable improvements in both performance and economics. The framework successfully reduced computing expenses and response latency while maintaining high output quality, proving its capacity to handle the demands of commercial applications without compromising reliability. The researchers detailed their findings in a paper presented at the ACM Conference on AI and Agentic Systems in 2026, with full proceedings published through the conference archives. As enterprise and consumer AI deployments grow increasingly complex, automated orchestration tools like Orla are becoming essential for scalable development. By abstracting the intricacies of model coordination and resource allocation, the framework lowers the barrier to entry for developers while enabling organizations to deploy sophisticated AI systems more efficiently. The Harvard team work signals a shift toward infrastructure level automation in the AI stack, positioning workflow optimization as a foundational requirement for next generation intelligent applications.