Connecting OpenAI to action builder
Before you can include external actions in your action flows, you must connect the action builder to the external system.
- All external actions performed by an action flow are attributed to the user who connected the external system. Therefore, it is a best practice to use a dedicated service account rather than personal credentials when connecting to each external system.
- All integrations request access to necessary scopes. However, it's important that you review and validate the scopes before authorizing the connection to the external system.
- When managing credentials for API key-based tools, such as OpenAI, it's best to store keys in a secure vault or credential manager.
- In Admin Center, click
Apps and integrations in the sidebar, then select Actions > Action flows. - Create or edit an action flow.
- Open the step sidebar.
- Under External actions, click OpenAI.
- Click Connect.
- Use an API key to authenticate the account. Note:
- The connection isn't confirmed until you test or use an action flow with OpenAI steps.
- All external actions performed by an action flow are attributed to the user who connected the external system. Therefore, it is a best practice to use a dedicated service account rather than personal credentials when connecting to each external system.
After you've connected to the system, you'll see an indicator that it's connected and details about the instance you're connected to, as well as the actions available for OpenAI.
Using OpenAI actions in action flows
The OpenAI action steps make it possible for you to provide things such as AI-generated summaries of long tickets to agents, identify negative emotional tones in tickets so they can be routed to specialized teams, extract keywords that can be used as ticket tags, and use tailored AI models to analyze customer messages.
GPT 5.0 is supported for all of the OpenAI steps. This enhances automation capabilities by providing better contextual understanding and more natural responses.
Analyzing sentiment
Use the Analyze sentiment action to detect the emotional tone of text.
This action has the following inputs and outputs:
| Variables | |
|---|---|
| Inputs |
model, text
|
| Output | sentiment |
Summarizing text
Use the Summarize text action to conense large quanities of text into a brief summary.
This action has the following inputs and outputs:
| Variables | |
|---|---|
| Inputs |
model, text
|
| Output | summary |
Extracting keywords
Use the Extract keywords action to identify main topics or entities. These can then be used as tags for tickets.
This action has the following inputs and outputs:
| Variables | |
|---|---|
| Inputs |
model, text
|
| Output | keywords |
Sending a prompt
Analyze this ticket and categorize it into one of the following: Billing, Technical Issue, Feature Request, Bug Report, General Inquiry.
Ticket:[description]where
[description] is replaced with a value by clicking Add
variable.
| Variables | |
|---|---|
| Inputs |
model, text
|
| Output | content |