Lesson 5 of 7
Adding an AI step
6 min read
Your flow can move data from app to app all day. But what happens when a step doesn't just need to move the data — it needs to actually understand it?
A smart step in the middle
Until now, every step followed a fixed rule. An AI step is different: you drop a language model — the technology behind ChatGPT — into the flow, give it a plain-English instruction, and it reshapes the data in ways rigid rules never could. Messy text in, clean result out.
An AI step is a module like any other — but instead of a fixed rule, it reads the data and reshapes it using a prompt you write.
What it's good at
AI steps shine at the fuzzy jobs a simple rule can't handle: summarise this long email, pull the order number out of this message, decide whether this review is happy or angry, or draft a reply in your tone. The judgement calls — that's where an AI step earns its place.
Between two apps, an AI step is your translator, summariser, or judge — turning messy human text into tidy data the next app can use.
The AI can be wrong or make things up. For anything that matters, add a human check or a rule to catch bad output before the flow acts on it.
The shape of it
- —An AI step is a module that runs a language model on your data.
- —You steer it with a prompt; it summarises, categorises, extracts, or rewrites.
- —It handles fuzzy jobs fixed rules can't — but it can be wrong, so check it.
What is an AI step best at, compared with an ordinary rule-based step?
Continue in the app
Take the whole Automation 101 course — tracked
Get your personalized path, progress and streaks in the app — this lesson and every next one, in order.