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 execute actions.
This is because many actions require AI agents to call APIs with the right values. AI agents use language to converse with users, but they need to call APIs which use IDs. Moveworks’ Agentic Automation Engine, and specifically Slot Resolvers, are designed to change this dynamic, enabling AI agents to integrate with every business system reliably.
This article will share how Slot Resolvers bridge the communication gap between humans and systems, how they are able to put AI agents to work with words, and explore the technical components that make these remarkable abilities possible.
Systems don’t use natural language
When you’re building an AI agent for all systems, you need to be able to actually integrate with those systems. Unfortunately humans speak in natural language:
"Please add Jim to Project Orion." |
And your business systems speak in structured APIs:
Your users want to manipulate these business objects, but something needs to translate natural language (“Project Orion”) to system language (“67428930”). In the era of SaaS, this meant developers needed to configure “display value” and “value” for every dropdown and checkbox. However, in the era of AI agents, we call this challenge “entity resolution.”
Unfortunately, today’s solutions for building AI agents are challenged to perform entity resolution efficiently. They take the similar approach: In-API entity resolution. We took a look at code samples for building AI agents from the existing solutions. For those of you who don’t speak Typescript, let’s explore the implications.
1. Most submitted values need to be manually “cleaned up.” The LLM is going to submit everything from “Orion” to “Project Orion” to “the Project Orion” to “PROJECT_ORION.” Developers are going to spend hours writing text processing utilities instead of building integrations.
2. Your AI agents can make decisions that your users don’t expect. With in-API entity resolution, developers write “matching” code to predict which business object the user is referring to. Unless you get a perfect match, traditional automation tools can throw an error. Users are often left to brute force different inputs until they find one that matches – executing the API over and over again.
Introducing Moveworks Slot Resolvers
We believe that in-API entity resolution is not the way to build AI agents. Moveworks Slot Resolvers offer a different approach to the problem. Moveworks Slot Resolvers allow developers to configure data types on their inputs (slots). The AI agent isn’t tasked with submitting “consultant name” – it’s tasked with submitting a “User” object.
Our Agentic Automation Engine is packed with cutting-edge conversational technology like a stateful execution tracker and entity resolution strategies that empower the Moveworks Copilot to have conversations that “resolve” those slots with transparency and accuracy.
It automatically finds and matches slots to business objects where highly confident. In this case, it found a perfect match for Jim Telustria and kept the user informed about its selection.
It picks an API call to execute to retrieve possible business objects (projects in this case) and presents them to the user with relevant information and citations so they can pick the right one.
It allows the user to disambiguate and confirm the project they want.
It preserves system IDs between turns, without letting the LLM accidentally corrupt / mutate the values.
To top it off, developers aren’t limited to just using the IDs, they can access the entire business object in their actions. So if you need to add an additional step that collects an approval from the project owner, you can just reference it as “project.owner” – developers write even less code, and end-users get lower latency AI agents.
Slot Resolvers make AI agents robust
Moveworks Slot Resolvers offer a truly agentic solution to the entity resolution problem. Since we bind plugins to data types, the Moveworks Copilot is able to more effectively dynamically assemble plans across plugins, and pass resources reliably from one plugin to the next.
With this architecture, developers need to write MUCH less code. Think just one tenth of what you normally write.
That means MUCH less text formatting, parsing, matching, disambiguation, state tracking – and it feels like almost nothing compared to traditional approaches. The Agentic Automation Engine handles it for you.
Building with iPaaS and middleware can leave you with broken end user experiences that aren’t robust to support the ways that users speak in natural language. But whenever you build with the Agentic Automation Engine, you get more elegant, intuitive, and usable experiences that users love.
Enhancing developer productivity and efficiency
This robust approach towards entity resolution makes it possible to build automations that are able to act on every business object. Developers often face a relentless challenge: every unique user input requires some form of text cleanup. In traditional setups, they can spend countless hours tackling edge cases like inconsistent project names or unwanted terms that need filtering. Without automation, each variable can mean extra code for “cleaning up” inputs, such as removing specific words or handling null values, to ensure compatibility with backend systems.
Slot Resolvers help to streamline this process by automating the entire cleanup pipeline, allowing developers to avoid the text-processing grind. Instead of writing custom functions for each malformed input, Slot Resolvers can automatically map user entries to the correct system objects, freeing developers to focus on impactful innovations rather than “sanitizing” data. This reduction in low-level fixes is able to not only speed up development cycles but can also lead to cleaner, more maintainable code.
Humans speak in words, but systems speak in IDs: Bridge them with Moveworks Slot Resolvers
As businesses grow and integrate new tools, Slot Resolvers become essential for AI agents, enabling them to convert user language directly into actions across diverse systems without sacrificing reliability. Slot Resolvers give organizations a unified data model across their entire business.
Developer teams can build integrations that span multiple platforms seamlessly, reducing the need for repetitive coding, text processing, and data mapping. Moveworks’ Slot Resolvers aren’t just a convenience—they’re a necessary step-function change in how the world builds AI agents that can understand and execute tasks across any system, 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
