The idea of removing complexity is one of the things that drives us here at Integrail. Given all the complexities that AI entails under the hood combined with all the complexities inherent in any company, being able to connect systems is a critical function. The more seamless and efficient the connection process is, the more potential ROI there is to be gained.
This is especially true with regard to building autonomous AI workers.
The most advanced of the different classes of AI collaborators, AI workers can handle complex workflows and reasoning to deliver specific outcomes described by a human worker. These capabilities separate them from AI chatbots, which provide basic chat interactions to accelerate simple tasks; and AI agents, which typically consist of prepackaged prompts and simple workflows to complete repeatable tasks.
The Universal Connector is part of our AI Studio platform for building and employing AI workers. In sum, it is a tool for seamlessly connecting to any API in seconds. It is essential when you’re building an AI worker and need to connect to disparate systems to source the data that will drive the worker to complete tasks within relevant workflows.
More specifically, if you have some documentation -- it could be a human language description, an example of usage using programming language, or pretty much anything from which we are able to extract information about how to call the CPI – you can feed it into the Universal Connector. Then the connection will be created for you to connect to the desired API.
This is in contrast to what people used to do before LLMs, which was to code by hand. Now our customers can just build a custom worker that generates those connectors on the fly, allowing them to connect to any system they want in seconds.
Integrations are key to many parts of the modern world. And without integrating different systems, it’s basically impossible to get any useful work done.
As referenced above, the way we at Integrail think about agents – which coincides with the direction the industry is moving in – is that of autonomous AI workers that execute complex tasks. In essence, companies will have a biological workforce in the form of human employees and a digital workforce in the form of AI workers.
In order for AI workers to execute useful tasks, they need access to the systems a company uses, just like a human worker does, to get information and make changes as directed. This could be a CRM, an email system, and an ERP system, just to name a few.
For example, using the Universal Connector, an AI worker can change the status of an opportunity in the sales pipeline or create a draft email to follow up on conversation. Basically, anything a system can do, your AI workers can do as well.
The most common and straightforward use case we see is what people call “talk to your data.” Although it's a relatively basic use case, it allows people to get familiar with what AI workers can do for them. The idea is that an employee can create an AI worker that listens to what the human is asking and understands what needs to be done. From there it will generate several search queries to all relevant corporate systems, such as an internal Wikipedia, SharePoint, the email system – whatever system is needed. All of them can be connected through the Universal Connector.
From there, the AI worker will return relevant documents. A large language model (LLM) of the user’s choosing will analyze those documents and combine them into a final report that provides the knowledge a company has on a specific issue.
Another use case that is becoming more common as companies get smarter about how agentic AI can benefit their bottom lines is building highly customized AI workers.
In these cases, it often makes sense for customers to engage with our Professional Services team to not only identify high-value use cases and business processes, but to build tailor-made AI workers to execute them. Every company has many complex rule-based job descriptions that need to be performed on a daily basis. They are what we consider to be the low-hanging fruit because the AI workers remove the boring and mundane parts, leaving the human employees to focus on the more creative and strategic parts.
One of the questions we hear frequently with regard to Universal Connector is how it compares to MCP. It’s a fair question since, at a high level, both are about connecting different systems.
As Anton Antic, our Chief Product Officer, told me in the video interview we did on this subject, this is somewhat of a philosophical question. Universal Connector and MCP take different approaches to how the actual connecting is done and who can instigate the connecting.
MCP is a technical tool focused on developer use. Using MCP means you need to know Python and be able code your way to creating workers.
Universal Connector does not require users to have any coding experience. It is part of our no-code AI Studio platform. Unlike MCP, non-technical businesspeople are able to create AI workers that can connect to all relevant systems and perform useful tasks. That said, we don’t see the need for MCP.
MCP is a good protocol when you are a developer and need to work with APIs to enable integrations. But without a technical background, you are beholden to others to help you create AI workers.
With Universal Connector, all you need is the API documentation, in any form. It can be formal documentation, it can be an open API spec, or even just documentation written in plain human language. It can use virtually anything as a source.
So, we welcome MCP, we support MCP, but we find it to be redundant with what Universal Connector can do.
While Universal Connector is already part of our AI Studio platform, we’re excited to be working on the next iteration. We’ve learned a lot from seeing how customers use it and have gotten a lot of great feedback on how to make it even better.
Simplifying the UI is one of the big improvement areas. Another improvement will be providing out-of-the-box connectors to all the major corporate environments, such as Microsoft, Google and others. So, even though Universal Connector is not limited to the systems it can connect to, having connectors ready to go for the most widely used systems will make the time-to-value even faster for our customers.
The Universal Connector is not only the foundation of our platform, but part of the evolutionary development of the idea of making it easy for anyone to connect to anything, regardless of role or technical abilities.
If Universal Connector sounds like something that would benefit you and your company, I highly recommend watching Anton demo it in the video starting at the 12:20 mark. He walks through the steps and shows how easy it is to use, even if you’re not a developer.
I encourage everyone to use their imagination to come up with AI worker “job descriptions” to fit any workflows and tasks. Try building them with AI Studio for yourself (start for free here) or even just share them with us. We are always on the lookout for new ideas to build out and share with our community.