AI Assistant Data Flow

This article describes the data flow for AI Assistant.

Overview

AI Assistant is the embedded, centralized prompt library that helps employees create, edit, and improve business content directly within their existing workflows — while maintaining full brand, compliance, and administrative control. Unlike standalone AI tools, AI Assistant operates inside Templafy’s governed ecosystem, ensuring every suggestion aligns with company standards, tone of voice, and regulatory requirements.

How it Works

1. The flow starts in the user’s Outlook, Word, PowerPoint or Excel. The user can open the Templafy add-in and select the AI Assistant tab. In this tab, all the active prompts will be displayed as buttons with an icon and short description of the prompt name. The user can also select some text in these applications to get an in-context floating menu of prompts.
2. If the admin has pre-defined a company tone of voice within Templafy, the AI Assistant will include that in the chosen prompt action. Additionally, context is provided from all the uploaded files in the Admin Center and the knowledge connector bringing additional context from M365 storage like SharePoint, OneDrive or third-party applications or external AI platform like Glean. A complex prompt containing the selected text, the selected action, the tonality and the context provided in the Admin Center is then created.
3. This complex prompt is sent to the AI generative service, which in the current set up is Azure OpenAI Service, which will generate the output text and send it to the AI Assistant. The customer also has the flexibility to connect their preferred AI model to it. Finally, the output is displayed within the Templafy add-in, and the user can either copy it or ask the AI Assistant to generate a new output. Once the text is suitable for the user, they can click on Insert text and the text will be inserted in the email, document, or presentation instead of the text that had originally been selected by the user.