Julius onboarding

An AI data analyst: upload a file, ask in plain English, get charts and analysis.

Julius

Prompt templates as a data viz workflow library

Julius turns saved prompts into workflows with example outputs and visible steps, runs them in a side panel, and publishes them as SEO category pages.

B2CB2BonboardingactivationFeb 2025

01

What do you want to analyze today?

The empty thread offers a prompt box and, below it, ready workflows with run counts.

Julius: What do you want to analyze today?
  1. 1

    Start without a file (working)

    "Add a file or start a conversation now and add files later" removes the first blocker.

    Reduce friction

  2. 2

    Unexplained mode chips (not working)

    "Default", "Tools" and "Advanced Reasoning" sit under the box with no hint of what they change.

    Clarity

  3. 3

    Workflows with run counts (working)

    "44267 runs" on Quick Visualization tells new users where to start.

    Social proof

  4. 4

    Teams upsell in the sidebar (working)

    "Julius for Teams" is a small dismissible card, not a modal.

    Contextual timing

02

Quick Visualization

Opening a workflow shows what it does, an example result and the exact prompts it will run.

Julius: Quick Visualization
  1. 5

    Three ways in (working)

    "Clone workflow", "Preview" and a dark "Run workflow" suit both tinkerers and doers.

    Flexibility and efficiency of use

  2. 6

    Keyword tags as clutter (not working)

    "Graphing", "Bar Chart Maker", "Line Graph Maker" read like SEO keywords, not useful filters.

    Signal vs noise

  3. 7

    Files Needed (working)

    Says up front that you'll need to bring data.

    Expectation setting

  4. 8

    Example output (working)

    A real chart ("Distribution of Squirrel Fur Colors") shows what you'll get before you upload anything.

    Show, don't tell

  5. 9

    Prompts in the open (working)

    Workflow Steps show the exact prompt text, so the AI step isn't a black box.

    Transparency

03

Data Cleaner

A cleaning workflow that suggests changes before applying them.

Julius: Data Cleaner
  1. 10

    Suggest first, act later (working)

    "Receive suggestions ... without making any changes yet" keeps the user in charge of their data.

    User control and freedom

  2. 11

    Illustration, not output (not working)

    Unlike Quick Visualization, the preview is a generic folders graphic, so you can't see a result.

    Show, don't tell

  3. 12

    Explicit hold step (working)

    Step 2 says "don't do anything yet", a human checkpoint written into the prompt.

    Human in the loop

  4. 13

    Show more covers the text (not working)

    The "Show more" button sits on top of step 3's text, making it unreadable.

    Aesthetic-usability effect

04

Significance Testing

Stats for non-statisticians: describe the comparison and Julius picks the test.

Julius: Significance Testing
  1. 14

    No jargon needed (working)

    "Without needing extensive statistical knowledge" names the barrier and removes it.

    Reduce friction

  2. 15

    Real result preview (working)

    A boxplot with t-statistic and p-value shows the level of output to expect.

    Show, don't tell

  3. 16

    Plain-language input (working)

    "You don't need to know the name, you can just explain what you want to know".

    Recognition over recall

  4. 17

    "Yolo it" in a stats tool (not working)

    The visible prompt says "Yolo it and follow up if any issues", which undercuts trust in the result.

    Trust signals

05

Time Series Analysis

An advanced workflow with a decomposition chart preview.

Julius: Time Series Analysis
  1. 18

    Wall of jargon (not working)

    ARIMA, SARIMA, stationarity tests in a five-line paragraph, the opposite of the significance workflow's tone.

    Cognitive load

  2. 19

    Usage as proof (working)

    "9056 runs" shows this is not a niche template.

    Social proof

  3. 20

    Decomposition preview (working)

    Observed, trend, seasonal and residual panels preview the output.

    Show, don't tell

  4. 21

    Same overlap bug (not working)

    "Show more" again covers step 3 ("Specify the column ... time series data").

    Aesthetic-usability effect

06

Running the workflow: provide your data

The run opens as a thread with the workflow steps pinned on the right.

Julius: Running the workflow: provide your data
  1. 22

    Sample data option (working)

    "Upload files or load default" lets you try the workflow with no data of your own.

    Time to value

  2. 23

    Google Sheets link (working)

    "Link Google Sheet" covers the most common data source without a download.

    Flexibility and efficiency of use

  3. 24

    Step counter (working)

    "Workflow 1/2" tells you how far you are.

    Goal-gradient effect

  4. 25

    Plan beside the chat (working)

    The steps stay visible on the right while you work, step 1 highlighted.

    Visibility of system status

07

Workflow complete

Julius returns a bar chart, marks both steps done and offers to keep going.

Julius: Workflow complete
  1. 26

    Nudge to a better chart (working)

    "Want better charts? View in interactive mode" points to a richer output right under the result.

    Progressive disclosure

  2. 27

    Interactive or static (working)

    A toggle lets you switch chart type without re-running anything.

    User control and freedom

  3. 28

    Filler summary (not working)

    "This chart effectively illustrates the frequency of the most prevalent words" restates the chart instead of an insight.

    Framing

  4. 29

    Run next step after the end (not working)

    "Run next step" still shows while the panel says "All steps completed".

    Consistency and standards

  5. 30

    Completion state (working)

    "All steps completed" with a green check closes the loop.

    Visibility of system status

08

Explore Workflows

The workflow library has its own page with search and categories.

Julius: Explore Workflows
  1. 31

    Search first (working)

    A large search box fits people who arrive with a specific chart in mind.

    Discoverability

  2. 32

    Categories with counts (working)

    "Data Visualization 32 workflows" shows where the depth is.

    Discoverability

  3. 33

    Mixed taxonomy (not working)

    "Bar Chart Maker" sits next to broad topics like Statistics, a keyword page posing as a category.

    Consistency and standards

  4. 34

    Community-made (working)

    "Created by our community" signals the library grows without Julius writing every template.

    Social proof

09

Category page: Bar Chart Maker

Public category pages, like julius.ai/category/bar-chart-maker, list workflows with real chart thumbnails.

Julius: Category page: Bar Chart Maker
  1. 35

    Keyword URL (working)

    /category/bar-chart-maker matches what people type into Google.

    Discoverability

  2. 36

    Output as thumbnail (working)

    Each card shows the chart it makes, so you browse by result.

    Show, don't tell

  3. 37

    Run counts don't match (not working)

    Quick Visualization shows "1307 runs" here and 44267 in the app.

    Consistency and standards

  4. 38

    Named creators (working)

    "Alysha.G" and a verified badge credit community authors.

    Social proof

10

Category page: Data Visualization

A broader category page in the same template, a gallery of chart types.

Julius: Category page: Data Visualization
  1. 39

    SEO filler intro (not working)

    "Dive in and start visualizing your data today!" closes a generic paragraph written for crawlers.

    Signal vs noise

  2. 40

    Gallery of chart types (working)

    Bar, correlation matrix, line, pie: you pick the picture you want and the workflow comes with it.

    Recognition over recall

  3. 41

    Shareable preview links (working)

    Each card links to its own /workflows/preview page, so a template can be sent around.

    Network effects

Takeaways from Julius

  1. Package good prompts as named workflows with an example output and visible steps, so new users start from a result, not a blank box.
  2. Offer sample data next to upload so people can see the workflow work before trusting it with their own file.
  3. Publish templates as keyword category pages to turn your library into an acquisition channel, and keep stats consistent across both.