Today, most of us have conversations with AI systems like ChatGPT. Usually, these AI systems use publicly available information to answer your questions, but they do not connect to your company’s systems to find answers.
With customized ChatGPT environments you can provide an environment that goes a step further, it allows for accessing company information in a secure way to answer your questions. This information can come from uploaded files but can also come from services that deliver information real-time.
This article describes how to set up a custom ChatGPT environment and demos how you can ask questions and get real-time data back from a company API. In this way you can create a smart assistant that increases your productivity.
Setup a custom ChatGPT environment
A custom ChatGPT is a tailored environment that can execute specific tasks that apply to your business domain. You will need a Plus, Teams or Enterprise ChatGPT subscription to create custom ChatGPTs.
Setting up a custom ChatGPT is very easy. Next to name and description you specify instructions, conversation starters and actions. An action can be a connection to an API or data that is uploaded.
Define a custom action to retrieve real-time data
In this example we will link to a company API that retrieves time tracking data so users can for example analyze time spent on projects over a period of time.
The API is specified using the OpenAPI specification. To retrieve data, ChatGPT can specify a period with start and end date and the API will return data. Note that a privacy policy for the data is required.
Authentication and authorization
ChatGPT business subscriptions work with the concept of company workspaces that have their own settings, members, and resources. In this way a private environment is created.
For authorization and authentication between ChatGPT and company services an API key or OAUT2.0 flow can be used to allow ChatGPT to connect on a user’s behalf in a secure way. This enforces not only authentication but also authorization of the user.
Let’s try a conversation and ask a business question!
The custom ChatGPT environment is connected to the business service. Now the user can ask questions. If the questions apply to the description of the business service, ChatGPT will use the service to retrieve answers.
When watching the video, note the following:
- ChatGPT is smart enough to translate the question into API parameters, it derives a start date and end date for quarter 1 and uses this to retrieve data. Of course, you must make sure that the dates are correct.
- In the conversation, a follow up question is asked to further refine the results.
What next?
This article demos a ChatGPT conversation where live company data is retrieved and summarized. The example is about time tracking data, but the idea is applicable to almost any business process where data is involved.
Hopefully this example makes you think about your business. With a conversational user interface, it is very easy to retrieve and analyze live data. And that is a great way to increase productivity for your employees.
Recent Comments