Build-time vs. runtime AI
Tines uses AI in two distinct ways, and understanding the difference helps you choose the right approach for your workflows.
Build-time AI
Build-time AI helps you create workflows. It runs when you're building or configuring an action, not when your story executes.
The best example is the Event Transform action in automatic mode. When you use automatic mode, you write plain English instructions describing what you want to do with your data. The AI takes those instructions and generates Python code to perform the transformation. You can review the code, test it with sample data, and regenerate it as many times as you need until it fits your desired input.
Here's the key part: once you save that code, it's locked in place. When your story runs, only the code executes. No AI is involved at runtime in this instance.
🪄Tip
Runtime AI
Runtime AI executes when your story runs. Each time the action triggers, AI processes data and makes decisions in real time based on the current input.
The AI Agent action is the primary example of runtime AI in Tines. When an AI Agent action runs, it analyzes the data it receives, applies the instructions you've provided, and generates a response. The output can vary based on the input, making runtime AI ideal for tasks that require dynamic analysis or decision-making.
🪄Tip
Understand AI model selection
When you use AI features in Tines, different features use different AI models depending on the complexity of the task. Your tenant is configured with two types of models: a fast model and a smart model.
Fast model vs. smart model
The fast model is optimized for speed and efficiency. It's designed for tasks that need quick responses and straightforward processing, like generating formulas or transforming data. Fast models prioritize speed while still maintaining accuracy for these focused tasks.
The smart model is optimized for complex reasoning and multi-step tasks. It's designed for features that require deeper analysis, decision-making, or handling more sophisticated interactions, like AI agents or conversational interfaces.
How models are configured
Your tenant owner configures which specific AI models are used as the fast and smart models in your tenant's AI settings. They can choose from various AI providers (like Anthropic, OpenAI, or AWS Bedrock) and select the appropriate models based on your organization's needs.
As a builder, you don't need to configure these models yourself. The AI features in Tines automatically use the appropriate model (fast or smart) based on what you're trying to accomplish. However, it's helpful to understand which features use which models so you know what to expect in terms of performance and capabilities.
Tines AI features
Tines offers several AI-powered features designed to make building and running workflows easier and more powerful. Each feature serves a specific purpose, and you can use them individually or combine them in your workflows.
🪄Tip
Human-in-the-loop approach
At Tines, we believe AI should augment human capabilities, not replace human judgment. This philosophy is built into every AI feature we offer.
You're always in control: AI in Tines doesn't run automatically or make decisions without your explicit direction. You decide when AI features are used, what data the AI processes, what instructions it follows, how outputs are used, and whether outputs require human review before action is taken.
Built for oversight and transparency: Tines makes it easy to review and validate AI outputs. You can test AI features with sample data before deploying them, review generated code in automatic mode, monitor AI Agent action outputs in your workflow logs, set up approval steps for AI-generated decisions, and track AI usage and performance over time.
Designed for critical workflows: AI features provide powerful capabilities while maintaining the reliability and control that critical workflows require. For example, you might use an AI Agent action to analyze and categorize security alerts, but you'd still have a human security analyst review high-priority alerts before taking action. The AI handles the initial triage and provides context, but the human makes the final decision.
Balancing automation and judgment: The goal isn't to automate everything; it's to automate the right things. AI in Tines helps you automate repetitive data processing tasks, speed up analysis and decision-making, and reduce manual effort on routine operations. But it keeps humans involved in the decisions that matter. You define the boundaries, set the rules, and determine when human judgment is required.