# How to find user behavior patterns with AI, and follow up with the people in them

> A behavior pattern is a group of people who did something in common: signed up and never came back, kept returning to pricing, got stuck on the same step. Here is how an AI agent finds them from plain words, why the rules matter more than the model, and what to do with the group once you have it.

Source: https://behavly.io/blog/find-user-behavior-patterns-with-ai · Behavly blog · Agentic Analytics · November 17, 2026 · By Ardit Karavidaj, Founder & CEO, FoundryX

A **user behavior pattern** is a group of people who did something in common. They signed up and never came back. They read the pricing page three times. They rage-clicked on the same step of setup. Patterns are where analytics turns into action, because a pattern is a group you can do something for.

Finding them used to mean building segments with filters and operators. With an AI agent you describe the group in plain words. The hard part is making sure the group you get is the group you meant. Here is how Behavly does both.

#### Two ways Behavly finds patterns

##### Patterns Behavly finds on its own

Behavly reads your tracked behavior and offers the patterns it found. Some you will see: **Signed in, never used it**, **Gone quiet**, **Power users**, **Keeps coming back**, **Hit friction** and **New faces**. Each one spells out its rule under **The rule**, because a name is a label, not a definition, and the numbers in the rule come from your own data.

##### Patterns you describe

Under **Looking for another pattern?**, or by asking **Behavly Agent**, you describe the group you want in your own words:

- “People who saw pricing but never bought.”
- “Signed up but haven’t come back in two weeks.”
- “Clicked the import button but never finished an import.”

Behavly turns the description into a rule and counts who matches it right now. The answer has the pattern’s name, its rules in plain words, the exact number of people and a sample of up to five of them. Nothing is saved until you say so.

#### Why the rules matter more than the model

A pattern decides who gets an email, so a wrong reading is not a small mistake. Behavly is built so the AI cannot quietly widen your audience:

- **It maps your words onto a closed set of behavior rules.** It does not write a database query, and it cannot invent a rule Behavly does not support.
- **It only uses what your site actually recorded.** When a description names something Behavly has not seen on your site, it says so and suggests things it has recorded that you could ask for instead.
- **It explains an empty result.** If a pattern matches nobody, it says why and offers a nearby alternative when there is one.
- **You see the group before it is saved.** The rules in plain words and the exact size come first. That preview is where you catch a description that was read differently than you meant.

#### A saved pattern keeps finding people

A saved pattern is stored as a **rule**, not as a list of names. Every time it is used, it runs again against current data. A campaign for “people who stalled in onboarding” takes in the people who stall this week, not only the ones who had stalled when you made it.

**10 min** how often an Always-on campaign re-checks who is in its pattern. Each person enters once, the moment they join.

#### What to do with a pattern

A pattern is a starting point for a **Pattern Campaign**. Ask Behavly Agent to build one, or choose the pattern on the campaign’s first step, **Who enters**. Then:

- **Once**: The group is read when the campaign starts, and only those people enter.
- **Always on**: Each new person who joins the pattern enters the moment they join. With **Also the people in it now** on, the people already in it enter when it starts.

Each email on the path can be written for one person from what that person did, so the people in **Hit friction** hear about the step they were stuck on, not a generic tip. For someone who kept trying to import a file, it could read like this:

To someone in a “started an import, never finished” pattern

> Subject: Your CSV import
>
> Hi Sam,
> 
> I saw you tried the CSV import a few times yesterday and it didn’t go through. Sorry about that.
> 
> The most common snag is a header row the importer doesn’t recognize. If you reply with the file’s first two lines, I’ll tell you exactly what to change.
> 
> Ardit

> Note: Someone who stops matching a pattern is not taken off the path. Use the path’s **Goal** or **Replies** for anything that should end it.

#### Try it

1. Add the Behavly snippet to your site and your signed-in pages.
2. Open **Behavly Agent** and describe a group: “people who signed up last week and haven’t come back”.
3. Read the rules and the count. If it is not the group you meant, say what to change.
4. Save it, and ask for a campaign for the people in it.

Patterns are one part of [agentic analytics](https://behavly.io/blog/what-is-agentic-analytics). For other questions to ask, see [Agentic analytics examples](https://behavly.io/blog/agentic-analytics-examples), and for the whole path from a pattern to an email, [From insight to action](https://behavly.io/blog/agentic-analytics-lifecycle-email).

Start on the free plan: https://behavly.io/auth/login?mode=signup
