Web agents cut token costs greatly using app behavior models
Token Efficient Task Execution via Application Behavior Modeling for Web Agents
Artificial Intelligence
Summary
Processing language tokens to control web applications can be costly. The authors developed OdoBot, a web agent that learns how an application behaves by watching successful task examples. Using this knowledge, OdoBot completes web tasks with far fewer tokens than competing agents while accomplishing more tasks correctly. This makes automating web-based tasks cheaper and more reliable on platforms like online learning systems.
What this means in practice
- •For software developers: Build web automation tools that handle complex tasks more efficiently with fewer API calls by leveraging behavioral models of web interfaces.
- •For online platform operators: Improve the scalability and cost-effectiveness of AI-driven task automation on web platforms like learning management systems through reduced token use.
Authors
Alexandru Ianta, Eleni Stroulia
Abstract
The strong performance of AI Agents across an impressive variety of tasks is driving an unprecedented investment in agentic infrastructures, however the cost of processing tokens is fast increasing. Web agents automate the execution of web-application tasks described in natural language, by analyzing the web-application's user interface (UI) and interacting with it. This work introduces OdoBot, a novel web-agent architecture that completes tasks at a fraction of the cost when compared to conventional web agents. This is achieved by leveraging a behavioral model of the underlying application constructed by analyzing successful task-execution demonstrations. Our experiments with 45 tasks on the Canvas Learning Management System (LMS) demonstrate that OdoBot uses 44% and 80% fewer tokens than two state-of-the-art competitor agents (Agent-E and WebVoyager), while also surpassing WebVoyager in terms of task success rate.