The Generate text action uses generative AI to create content automatically during a workflow. You can use it to generate messages, policy drafts, summaries, task instructions, or personalised responses, based on your prompt and the context of your workflow.
This helps reduce manual work, speed up response times, and deliver more relevant, consistent, and human-like content to your employees.
You can use the output of this action in later steps, such as:
- Sending a notification
- Creating a task
- Updating a model property
- Displaying content on a page
What is an AI prompt?
An AI prompt is the instruction you give the AI to tell it what to write. The clearer and more specific your prompt, the better the result.
For example:
Too vague: “Write a policy”
Better: “Write a short remote work policy for UK-based employees in the finance department. Use a friendly but formal tone.”
You’ll get better results by including:
- Audience – Who is this for? A new starter? A manager?
- Goal – Should it be a quick summary or a detailed explanation?
- Tone – Professional, encouraging, lighthearted?
Providing this context helps the AI generate content that fits your tone of voice, audience, and intent.
👉 For tips on writing good prompts, see:
Use retrieved content to improve AI results
You can now feed real-time content into your AI prompt by using the Retrieve Information workflow action before Generate text.
For example:
- Search your Knowledge or Documents model using keywords.
- Retrieve the most relevant content using Global search or model-based filtering.
- Pass the retrieved results into your AI prompt using tokens like
Retrieved Knowledge results.
This helps the AI generate smarter, more accurate responses based on your organisation’s actual content.
💡 Example:
“Write a welcome message for a new starter in the finance department. Use the following content as context: {{Retrieved Knowledge results}}”
You can also use the retrieved record count to change behaviour in the workflow, like skipping the text generation if no relevant articles were found.
Run as a user
To make responses more relevant, you can set the workflow to run as a specific user, often the triggering user or the content owner. You can then include that user’s role, department, or other properties in the prompt.
Example:
“Explain this for a new starter in {{@triggeringUser.department}} with no technical background.”
“Write this for a senior HR partner reviewing policy content for compliance.”
This helps tailor the AI response based on the user’s context.