As enterprise technology continues to evolve, companies are turning to AI agents to streamline business processes, boost productivity, and enhance user experience. Yet, the potential of these AI agents has often been limited by their inability to intuitively select the correct tools for each unique task—a problem that’s especially critical when the agent must interact with thousands of distinct plugins or automations. Moveworks’ Agentic Automation Engine, and specifically the Manifest Generator, is designed to change this dynamic, enabling AI agents to navigate even the most complex landscapes of enterprise applications.
Selecting the right automation is HARD
If you’re trying to build an AI agent that can connect to every business system, it needs to be able to understand a user’s request and decide what plugin to select. However, without access to the underlying information within plugins, AI agents are incapable of discerning the right plugin to pick.
Let’s take a simple example. Imagine I have two plugins:
Update a PR - Built by a DevOps engineer to close Github Pull Requests.
Update a PR - Built by Procurement IT to amend an Ariba Purchase Requisition.
With traditional automation tools, all an AI agent knows is there are two plugins with the same title and similar inputs (an ID to update and some values to set like status, assignee, etc.). AI agents get confused and don’t know which plugin to pick.
As a consequence, developers need to become prompt engineering experts to come up with the “perfect” LLM-facing instructions. This may seem tractable when considering just these two examples, but imagine when you have thousands of plugins in one AI agent – you’ll spend more time beating your AI agent into behaving correctly than expanding it to support more types of tasks.
AI agents were supposed to eliminate the pains tied to process discovery, but instead, other solutions have just converted the “find the right portal” problem into a “find the right agent” problem.
Other LLM-powered applications split up plugins and data into multiple autonomous systems as attempts to focus knowledge and actions on the right type of content falls short.
It’s extremely challenging to deliver AI agents through creative prompt engineering or isolated agents, however, Moveworks Manifest Generator handles this problem differently.
Introducing the Moveworks Manifest Generator
Let’s say your business wants an AI agent for a domain (e.g. HR), a process (e.g. Onboarding), a system (e.g. Workday), or a persona (e.g. Sales Reps). In Moveworks, you build plugins, which represent the different automations that an agent should be able to execute.
Our Manifest Generator analyzes and validates its understanding of plugins when they’re configured. It then surveys the 1000s of plugins you’ve built, and dynamically produces a “manifest” (or list) of plugins to present to the Moveworks Agentic Reasoning Engine. This manifest of plugins serves as an agent for any domain, process, system, or persona.
Inside the Manifest Generator
When plugins are configured, the Manifest Generator analyzes them and interprets all the slots, policies, and actions. Much like a new employee will ask their onboarding buddy clarifying questions to learn during their orientation, the Manifest Generator asks the developer to clarify when the plugin should be used.
(1) We analyze and validate the plugin’s purpose
The developer doesn’t have to come up with creative utterances or creative descriptions, since the Manifest Generator picks the best ones by comparing the plugin and its contents to the other 1000s of plugins configured in your environment.
When it’s time to use the plugin in the Moveworks Copilot, our Manifest Generator continues to augment the Agentic Reasoning Engine’s intelligence.
Let’s take the following example: Say I’m hosting a customer at our company HQ and I need a spot for them.
The Manifest Generator first filters plugins to the best possible subset of plugins that could help the user.
(2) We filter to the best set of plugins
In this case, “Book a facility” or “Book a spot” could both be relevant plugins based on what the user described. Presenting them directly to the user would be problematic, as they wouldn’t know which plugin is the right one for them. To solve this problem, we rewrite the plugin descriptions to make them more semantically unique. Now, it’s clear that one will help Audrey reserve a room for the customer meeting, and the other would help him reserve a parking spot for his visit to New York City.
Finally, our Agentic Reasoning Engine steps in, to clarify Audrey’s intent and ensure the right plugin is selected, with human-like intelligence.
Under the hood, the Moveworks Manifest Generator has a ton of critical innovations that make this possible. From knowledge graphs to enterprise reasoning eval datasets to model distillation and fine-tuning techniques, you can be confident that the Moveworks Manifest Generator can help support nines of reliability when it comes to selecting the right plugin.
Enhancing developer productivity and efficiency
This level of specificity can be a game-changer not only for users but also for developers, who can now design AI agents without needing to code extensive prompt-engineered instructions for each task variation. Traditional approaches might have required developers to build detailed, custom prompt scripts or hundreds of individual plugins to cover every possible task the AI could encounter, resulting in both code sprawl and increased maintenance overhead. With the Manifest Generator, developers may save substantial time and effort, as the engine’s automation and contextual intelligence help to reduce the need for exhaustive customization.
By automating the plugin selection process, the Manifest Generator allows developers to focus on the big picture rather than writing code for minor interactions or error handling. For instance, instead of managing complex prompt logic to distinguish between “PR” plugins, developers can trust that the Manifest Generator will consistently guide the AI to the correct selection, even as new plugins and functionalities are added. This efficiency boost ultimately can translate to faster development cycles and reduced tech debt, as each new feature or plugin integrates more seamlessly into the existing system without requiring re-engineering.
Don’t build a copilot for every app. Build the AI agent for every system
As businesses continue to scale and add new software tools, the Manifest Generator can remain a crucial asset, helping to enable AI agents to operate across diverse systems and handle ever-growing libraries of plugins. With this kind of intelligent automation in place, companies are able to empower their teams to achieve greater productivity and a seamless experience across departments. In many ways, Moveworks’ Manifest Generator is more than just a step forward for automation—it also represents a leap toward a future where AI agents can better understand and meet the demands of diverse business environments, no matter how complex.
Want to dive deeper into the Agentic Automation Engine? Why not read our white paper to learn more.
Table of contents
