Integration developers often face the challenge of connecting to unfamiliar systems and applications. In theory, this shouldn’t be difficult. Every system has API (Application Program Interface) documentation to help you build application connectors.

In reality, API design and documentation varies wildly. You’re probably familiar with some of the pain points caused by lackluster API documentation, namely wasted time and frustration. Learning APIs is already time-consuming. When they’re poorly documented or frequently altered, the result is more busy work as you try to re-engineer the API to interact with an application properly. 

I recently realized that this problem is an ideal use case for ChatGPT: Artificial intelligence (AI) can help developers spend less time deciphering complex API documentation, enabling them to quickly build application connectors. Here’s how I did it.

A Perfect Example of a Poorly Documented System

‍

I recently had a chance to compare and contrast how ChatGPT and Google can be used to solve an issue caused by poorly documented and complicated APIs. 

While building Boomi application connectors to push data into SAP Concur, I quickly realized that documentation for Concur’s API wasn’t readily available and the diagram they provided was confusing at best (the API uses five endpoints where one would do, see above). Upon testing, an unexpected behavior occurred: worker information was incomplete.

What follows is a comparison of how Google and ChatGPT can help developers solve integration problems like this one. As you’ll see, treating ChatGPT like a virtual Concur expert made it possible to solve this problem in record time with minimal headaches.

Deconstructing APIs: Two Ways to Problem-Solve

Now, let’s look at how I can use both Google and ChatGPT to solve my problem and build the Boomi connectors I need. 

Finding Documentation: Google vs. ChatGPT

With Google, I search for the official API documentation, scroll past ads trying to sell access to documentation, and finally find the correct API documentation on the second or third page.

With ChatGPT I type, “Give me the link to the official API documentation for SAP Concur” into the chat box, and the link is immediately generated. 

Trouble-Shooting Documentation: Google vs. ChatGPT

With the proper API documentation in hand, I then establish a connection and send a GET request to Concur to fetch all workers and avoid duplicates when creating new users. Upon testing the integration, I notice incomplete worker information. 

Using Google, I search for solutions and find discussions on Stack Overflow. After spending hours testing various endpoint URLs and adjustments, I determine that obtaining complete worker data requires three GET requests from different endpoints.

With ChatGPT, I simply use the bot as my own dedicated Concur expert, asking simple questions and explaining the situation like I would to a real person. After some back-and-forth, it tells me that I’m not receiving complete data for each worker, and I’ve saved hours of my time. 

Predictions About AI’s Impact on Existing Tooling

We’re already seeing the impact of ChatGPT, and AI in general, on existing tooling — tools across verticals and functions are rapidly releasing AI-assisted features. This is something I’m personally excited to experience more of in the platforms I use every day, like Boomi.

For example, I would love to see enhanced versions of existing features. Integrating ChatGPT with Boomi Suggest could streamline the mapping process and provide a more user-friendly experience. It could also remember earlier inputs, including any objectives and associated system information. 

Features like the above would streamline processes and save time, but will still require skill and practice to master.

Should All Developers Learn How to Use ChatGPT?

Absolutely, yes. 

Tech workers need to understand how to leverage emerging technologies like ChatGPT. While there are limitations and potential drawbacks of using AI-powered tools in development— they’re occasionally wrong, human fact-checking is still necessary, and the outputs sometimes need revising— there’s no denying their usefulness. 

This advice is especially important for Junior Developers, who are at a greater risk of having their jobs made redundant by AI. The good news is that if you understand development principles and how AI “thinks,” you can use that knowledge in your work. You can replicate the approach I described above for many different technical issues.

Conclusion

AI-powered tools like ChatGPT can markedly improve integration development work streams by streamlining the process of working with poorly documented or complicated APIs. 

By embracing these technologies and learning to work with ChatGPT and other AI-powered tools, developers can revolutionize the way they approach integration development, save time, and increase their productivity.