Saturday 11 October 2025
The automation of tool creation and enrichment is a crucial step in unlocking the full potential of large language models (LLMs) for enterprise applications. A new framework, called ACE, aims to streamline this process by transforming enterprise APIs into LLM-compatible tools.
At its core, ACE is designed to address the challenges posed by poor documentation, complex input or output schema, and the sheer number of operations involved in tool selection and invocation. To achieve this, the framework employs two primary mechanisms: generating enriched tool specifications with parameter descriptions and examples, and incorporating a dynamic shortlisting mechanism that filters relevant tools at runtime.
The first component of ACE is responsible for enriching tool specifications by providing detailed information about each tool’s capabilities, parameters, and expected outputs. This includes adding contextual descriptions, parameter types, default values, and example inputs to aid developers in selecting the correct tool for their needs. By making this information readily available, ACE reduces the time and effort required to discover and understand the capabilities of individual tools.
The second component of ACE is focused on dynamic shortlisting, which enables the framework to filter relevant tools based on specific requirements or constraints. This mechanism allows users to specify a set of criteria, such as tool type, functionality, or compatibility with specific platforms, and ACE will then provide a list of suitable tools that meet those criteria.
To demonstrate the effectiveness of ACE, the authors conducted experiments using both proprietary and open-source APIs. The results showed significant improvements in tool selection accuracy and reduced payload formation complexity. Specifically, ACE was able to reduce the error rate associated with incorrect tool selection by up to 25%, while also decreasing the time required to discover and understand tool capabilities.
ACE’s impact extends beyond improved tool selection and invocation accuracy. By automating the creation and enrichment of tools, the framework enables developers to focus on higher-level tasks that require domain-specific expertise. This, in turn, can lead to increased productivity and efficiency, as well as reduced costs associated with tool development and maintenance.
While ACE is primarily designed for use with LLMs, its principles and mechanisms can be applied to other areas of automation and tooling as well. As the demand for intelligent automation continues to grow, frameworks like ACE will play a critical role in unlocking the full potential of AI-powered tools and applications.
Cite this article: “ACE: Automating Tool Creation and Enrichment for Large Language Models”, The Science Archive, 2025.
Large Language Models, Enterprise Applications, Automation, Tool Creation, Enrichment, Framework, Ace, Api Integration, Intelligent Automation, Ai-Powered Tools
