Tendem: A Hybrid AI+Human Platform
This addresses the need for reliable and efficient task execution in real-world applications, though it appears incremental as it builds on existing hybrid approaches.
The paper tackles the problem of improving task performance by introducing Tendem, a hybrid AI+human platform that combines AI for structured work and human experts for verification, resulting in higher-quality outputs and faster turnaround times compared to AI-only or human-only workflows, with operational costs comparable to human-only execution.
Tendem is a hybrid system where AI handles structured, repeatable work and Human Experts step in when the models fail or to verify results. Each result undergoes a comprehensive quality review before delivery to the Client. To assess Tendem's performance, we conducted a series of in-house evaluations on 94 real-world tasks, comparing it with AI-only agents and human-only workflows carried out by Upwork freelancers. The results show that Tendem consistently delivers higher-quality outputs with faster turnaround times. At the same time, its operational costs remain comparable to human-only execution. On third-party agentic benchmarks, Tendem's AI Agent (operating autonomously, without human involvement) performs near state-of-the-art on web browsing and tool-use tasks while demonstrating strong results in frontier domain knowledge and reasoning.