Data foundry company Boundfor said on Tuesday it has added an AI token spending management service, Pareto, to its AI data operations platform, DroPai.

Pareto blocks AI API requests before execution if they exceed administrator-set budget limits. It also provides real-time team usage and expected spending, helping users manage budgets before costs occur.

Rather than post-use monitoring, or "observability", that checks costs after usage, it is based on pre-emptive control that prevents budget overruns. It applies company-defined budgets and policies during API calls to block unnecessary costs in advance and supports operating AI agents and automation systems within set budget ranges.

It analyses monthly usage data to automatically identify key patterns that lead to unnecessary spending. It detects accumulated retries that build costs as failed calls repeat, duplicate requests that effectively process the same request repeatedly, and over-specified model use where high-performance models are used for tasks that could be handled with lighter models. It provides reports on areas where savings are possible.

Hwang In-ho (황인호), Boundfor's chief executive, said, "Pareto was born in the process of solving problems that Boundfor experienced firsthand at physical AI sites." He said, "In environments where robots and devices call AI 24 hours a day, there were limits to relying only on a method of analysing costs afterward." He added, "We will continue to advance our field-verified operational experience into services and support companies' overall AI operations, from AI data management to cost control."