Generate and Transform Text with the GPT Action

The GPT powered by OpenAI action lets a workflow draft, summarize, translate, or classify text on its own. The result becomes a value you can reuse in a later step, such as an email, a text message, or a custom field.

What it can do

Pick an Action Type:

• Analyze Text Sentiment
• Summarize Text
• Translate
• Custom: write your own prompt, for example "Write a friendly follow-up to this message: {{contact.message}}".

Add it to a workflow

1. Click + where the AI step should go and search for GPT powered by OpenAI.
2. Choose an Action Type.
3. Select a Model. Larger models handle complex reasoning and long inputs; smaller ones are faster and cheaper for short, repetitive jobs like classification.
4. For Custom, write the prompt and insert merge fields, such as the contact's first name or a custom value, to personalize it. An optional system prompt can set tone, such as "friendly and concise".
5. Save, then test with a contact to review the output.

Use the output

The generated text can be:

• Stored in a contact custom field.
• Inserted into a Send Email or Send SMS step.
• Posted to your team through Slack or a webhook.

Writing better prompts

• State the role and the exact format wanted ("Write two sentences. No emojis.").
• Include a short example of a good answer.
• Wrap long pasted content in clear markers so it is easy for the model to separate from instructions.
• Be specific. "Summarize in 3 bullet points" beats "Summarize".

Cost and limits

AI steps are metered by the amount of text processed, both in your prompt and in the reply, and usage is shown in your AI billing settings. Switching models later only requires editing the action.

Good to know: Review AI-written messages in a test run before enabling them for real contacts, especially anything sent automatically.