-->

Enterprise AI World 2024 Is Nov. 20-21 in Washington, DC. Register now for $100 off!

Moveworks’ Agentic Automation Eases Complex AI Agent Development

Moveworks, the leading generative AI copilot for employee support, is debuting Agentic Automation, an automation engine designed to revolutionize the way enterprises create AI agents.

Traditional agentic AI developer platforms—which are built on integration platforms—operate through stiff, rigid commands that are incapable of addressing the nuance of natural language, according to Moveworks. With natural language inputs that may be difficult to translate into precise API commands, agentic AI requires more agile tooling that captures the subtleties of its inputs without sacrificing functionality.

“When you're talking about a large language model [LLM] calling a function or calling an API, there's a lot of ambiguity between [the] language that a person specifies and then the specificity that an API expects,” explained Varun Singh, president and co-founder of Moveworks. “What we found was [that] the developers—they wanted to build these agents that could do meaningful work, but they were stuck with two things. One is just the amount of work they needed to put into encoding… [the conversion of the] ambiguities of language into specificities of APIs. And second, even with that, they couldn't build more sophisticated agents.”

Prompted by this challenge, Moveworks’ Agentic Automation solution was developed to translate ambiguous natural language prompts into precise system commands, empowering developers to create complex agentic AI without needing an external integration platform. Powered by Moveworks’ agentic Reasoning Engine, which delivers hundreds of out-of-the-box agentic use cases, Agentic Automation facilitates the creation of AI agents capable of interpreting human language and executing a variety of automations for the end user.

Some of the key components of Agentic Automation include:

  • Manifest Generator: A new approach to interfacing with LLMs that analyzes plugins—or the tool that the LLM uses to interact with the underlying API—produces a manifest of plugins, and dynamically alters prompts to assist the AI agent in understanding the intricacies between plugins and how to effectively map them to user requests.
  • Slot Resolvers: A new method by which Agentic AI systems convert natural language into reliable, API-friendly IDs, enabling AI agents to correctly identify business objects and successfully execute tasks.
  • Policy Validators: An additional layer to Moveworks’ Reasoning Engine that enforces business rules to avoid misunderstanding or jailbreaking of LLMs’ instructions.
  • Action Orchestrator: An integration engine that informs the Reasoning Engine of actions that are planned, in-progress, or complete, enabling developers to focus on building rather than dialog generation, edge cases, and failures.
  • Moveworks Copilot: Agentic Automation can extend the power of the Moveworks Copilot by creating any agentic automation within any business application of their choice, integrating seamlessly with collaboration tools like Microsoft Teams and Slack to streamline the employee experience and minimize tool sprawl.

“In order to bring the reality of AI agents to life, there needs to be a tool that can expose automations in a manner that is friendly to large language models and allows them to interact with APIs,” said Singh. “On the other end, [it must be able to] interact with users and therefore bridge the gap between ambiguity of language and sort of the specific inputs that an API expects… [Agentic Automation is] a prerequisite towards bringing AI agents to life, and AI agents are the holy grail of bringing business productivity via AI, not just employee productivity.”

To learn more about Moveworks’ Agentic Automation, please visit https://www.moveworks.com/.

EAIWorld Cover
Free
for qualified subscribers
Subscribe Now Current Issue Past Issues