In today’s fast-paced environment, manual processes can be a bottleneck. The AI agent action in Tines is designed to break those barriers by integrating intelligent decision-making directly into your automation workflows.
Read this article to help you get started with AI agents in Tines.
While the AI agent offers two modes, chat and task, this guide focuses on task mode, which enables the agent to independently make decisions and execute actions without human intervention. This is the ideal solution for fully autonomous workflows.
Getting started with task mode
Configuring an AI agent action for task mode is straightforward. Follow these steps to get your agent up and running:
Add the action: Navigate to your Tines story. Select the AI agent action from the story toolbar and drag it onto your canvas.
Select task mode: With the action selected, open the properties panel on the right. Under the mode settings, choose task mode.
Configure instructions: Once in task mode, you can begin defining your agent’s behavior through system instructions.
Crafting effective system instructions
Your system instructions are the "brain" of your agent. They define the tone, domain, and specific persona the agent will adopt. A well-defined prompt leads to more precise and reliable outcomes.
Best practices for your prompts
Define a persona: Give your agent a clear identity. For example, if you are creating a cybersecurity assistant, define the persona as a cybersecurity analyst responsible for reviewing alerts, and provide context like: "You work on a security team tasked with investigating potential threats."
Stay organized: As your system instructions grow in complexity, use clear, distinct sections. This improves maintenance and helps the AI understand its role and boundaries more effectively.
This article goes through more best practices for the AI agent.
Integrating dynamic data
The power of the AI agent lies in its ability to process real-time information. You can inject dynamic data into your prompts by using data pills.
How to add data: Within the prompt input field, simply click the + button.
Referencing values: This allows you to select values from upstream actions or resources. Once selected, you can develop your JSON path to ensure the agent receives the exact data it needs to perform its task.
Example: You might prompt the agent to "summarize the data," followed by a data pill referencing upstream data. This ensures the AI always has the most current information to work with.