Add AI to a workflow
Give an AI step a bounded job with a result you can check.
IN THIS LESSON
- Provide relevant input and a clear output requirement.
- Separate text generation from exact business rules.
Define one job
AI can summarize, extract, and transform text within a workflow. Select an available AI node, configure its model and credentials as required, and provide the input through a variable.
A useful instruction is specific: summarize the supplied message in two sentences, preserve dates, and say when information is missing. Keep instructions separate from the content being processed.
Check the output contract
If the next step needs a category or a JSON object, configure and test that output explicitly. Do not assume generated text always matches a schema. Include examples with missing or ambiguous information.
Keep deterministic work deterministic
Use ordinary rules for exact comparisons and known thresholds. Use AI for language tasks where variation is expected and review is possible.
Try it
Test a summarization step with a short message, a long message, and one that lacks key information. Inspect all three outputs before connecting a downstream action.
Go deeper
See AI nodes, branching, and agents and protocols.