Prompt AI
Use Prompt AI to turn the text you provide into an answer that later steps can use. Ask for plain text when you need a summary or message, or define named fields when your workflow needs separate values or a list.
When to Use This Action
Use this action when:
- You want AI to analyze, summarize, classify, or transform text in your workflow
- You need to extract structured data from unstructured text, such as a priority, category, and summary from an email
- You want an answer based on the instructions and information you provide
Don't use this action for:
- Complex multi-step AI tasks: Use Invoke Agentic Action when you need AI to call tools, query APIs, or reason across multiple steps.
How It Works
Think of this action as a question with a reusable answer: your Prompt supplies the instructions and information, and Response Format chooses how you receive the answer.
Your prompt includes the information AI needs
Include both the task and the content to analyze. Text expressions such as {{email.body}} become the actual email text before AI receives the prompt.
The action does not look up records or call other tools. If you need information from a remote database, retrieve it in an earlier step and include it in the prompt, or use Invoke Agentic Action with the tools it needs.
Choose an answer your next step can use
Choose Simple for one text answer, such as a ticket summary. Choose Structured when you need named values, such as a priority and summary, or an array of search terms. In Structured mode, use Output Schema to name each field and choose its type.
Save the answer for later steps
Configure Store the results in Variable with a name such as company_search. Later steps can use that variable to read the answer or select an array for a loop.
Configuration
| Field | Description |
|---|---|
| Prompt | Describe what you want AI to do and include the information it needs. Use text expressions such as {{email.body}} to include available data. |
| Response Format | Choose Simple for a text answer or Structured for named fields that later steps can use separately. The default is Simple. |
| Output Schema | In Structured mode, name the fields you want back and choose their types. For a list of search terms, add query_strings as an array of strings. |
Returned Variables
The expressions below assume you name the results variable answer.
| Response Format | Example expression | What you receive |
|---|---|---|
| Simple | {{custom.answer}} | The text answer |
| Structured | {{custom.answer.summary}} | The value of a field named summary in your Output Schema |
Replace answer with your results variable name and summary with the field you defined.
Array previews help you configure later steps
For an array-of-strings field such as query_strings, the console shows a sample containing one empty string: [""]. This tells you the field is a list and lets you select it for a loop. The sample is not an AI answer; when the action runs, the field contains the strings AI returns.
For an array of enums, the sample contains the first choice you configured. The console also shows examples that match your other field types: numbers for numeric fields, true-or-false values for booleans, and text for string fields. These examples help you configure later steps; they do not predict the AI's answer.
Example Use Cases
Simple Text Analysis
Prompt: Summarize this email in one sentence: {{email.body}}
Response Format: Simple
Store the results in Variable: ticket_summary
Use {{custom.ticket_summary}} in a later step to read the summary.
Company Search Terms
Use AI to suggest search terms, then let later steps search your database.
Prompt: Return between 1 and 5 search terms for this company: {{input.company_name}}
Response Format: Structured
Output Schema:
- query_strings (array of strings): Terms to search for the company
Store the results in Variable: company_search
For a company name such as Acme Widget Company, AI might return:
{
"query_strings": ["Acme", "Widget"]
}
In a later loop step, select custom.company_search.query_strings as the array to iterate over. Each item is one search term.
The prompt asks AI for one to five terms; the array field type alone does not enforce that count. Review AI-generated terms before relying on them for an exact company match.
FAQs
Do I need an external AI account or API key?
No. You do not need to supply your own AI account or API key. Access to this limited-release action depends on your MSPintegrations subscription; contact your account manager if you need access.
Why does my array preview contain an empty string?
The console uses [""] as an example of a list of strings while you configure the workflow. Run the action with a company name to receive actual search terms.
Can this action search my company database?
No. It can suggest search terms from the company name you provide, but it cannot search the database itself. Use a database or integration action in a later step to search with those terms, or use Invoke Agentic Action with the appropriate tools.