Categories: Power BI
Tags:

AI Usage & Impact Analytics Dashboard in Power BI: Complete Step-by-Step Tutorial

Artificial Intelligence has quickly become part of education, professional work, research, coding, writing, customer support, and everyday productivity. But simply knowing that people use AI is not enough. A much more interesting question is: How are people using AI, and what impact does that usage have?

In this Power BI project, I used a publicly available dataset covering AI usage among students and professionals and transformed it into an interactive dashboard called AI Usage & Impact Analytics.

The objective was not simply to create attractive charts. The real objective was to take a messy dataset, clean it properly, prepare it for analysis, identify meaningful KPIs, and then build visuals that tell a story.

The final dashboard allows us to analyze AI adoption, popular AI tools, usage purposes, productivity, satisfaction, time saved, AI spending, and differences between students and professionals.

This is also a good example of a complete Power BI workflow because the project begins with raw data cleaning and ends with interactive business-style data visualization.

Where Did the Data Come From?

The original dataset was obtained from Kaggle:

AI Usage and Impact on Students & Professionals : Download Here

The dataset contains information about both students and professionals and records how they use Artificial Intelligence.

The original dataset contains 300 rows and 19 columns.

Some of the important fields include:

  • User_ID
  • Age
  • Gender
  • Country
  • User_Type
  • Education_Level
  • Profession
  • Monthly_Income
  • AI_Tool
  • AI_Purpose
  • AI_Usage_Hours_Per_Day
  • Tasks_Performed
  • Monthly_AI_Cost
  • Productivity_Score
  • Accuracy_Rating
  • Satisfaction_Score
  • Time_Saved_Hours_Per_Week
  • Work_or_Study_Hours_Per_Day
  • Would_Recommend

The dataset is particularly useful because it contains both categorical and numerical data.

For example, AI_Tool, Country, Gender, and Profession are categorical fields, while AI_Usage_Hours_Per_Day, Productivity_Score, and Time_Saved_Hours_Per_Week are numerical fields.

That makes it suitable for a Power BI project involving data cleaning, KPI creation, segmentation, comparison, and relationship analysis.

What Was Wrong With the Original Dataset?

Before creating the dashboard, I did not directly import the raw CSV and start creating charts.

The original dataset had several data-quality problems.

There were inconsistent category names such as:

  • male
  • Male
  • M
  • MALE

Similarly, AI tools contained variations such as:

  • ChatGPT
  • chatgpt
  • Chat GPT
  • Chatgpt

Education also had multiple versions of essentially the same category:

  • Bachelor’s
  • Bachelors
  • Undergraduate
  • undergrad
  • BA/BSc

The dataset also contained missing values across many columns.

There were also 12 duplicate rows.

This is important because poor data quality can directly affect Power BI results.

For example, if ChatGPT, chatgpt, and Chat GPT are treated as separate categories, Power BI will display them as different AI tools. That would make the AI adoption chart misleading.

How We Cleaned the Dataset

The first step was removing exact duplicate rows.

The original dataset had:

300 rows

After removing the 12 duplicate records:

288 rows

However, there were still many missing values.

Instead of filling those missing values with arbitrary averages or assumptions, I decided to create a strict analytical dataset.

The requirement was simple:

If a record has missing information in any important field, remove that record.

This reduced the dataset from 288 rows to 116 complete rows.

Therefore, the final Power BI dataset contains:

116 complete users

with:

0 missing values

and:

0 duplicate rows.

For a portfolio-style Power BI dashboard, I prefer this approach because every record displayed in the dashboard contains complete information.

Standardizing the Data

After removing incomplete records, the categorical fields were standardized.

For example, gender values were converted into consistent categories such as:

  • Male
  • Female
  • Other
  • Non-binary

Education categories were consolidated.

For example:

Undergraduate, undergrad, Bachelors, and BA/BSc were standardized into appropriate education categories.

AI tool names were also standardized.

For example:

chatgpt, Chat GPT, and Chatgpt

became:

ChatGPT

Similarly, Gemini was standardized to Google Gemini, and Copilot was standardized to Microsoft Copilot.

The numeric columns were also converted from text into proper numerical fields.

This was particularly important for Monthly_AI_Cost, where some values contained $.

After cleaning, the $ symbols were removed and the field was converted into a numerical column.

Creating Additional Fields

I also created a few calculated columns to make the dataset more useful for Power BI.

Weekly AI Usage

The original dataset provides daily AI usage.

I created:

Weekly_AI_Usage_Hours

using:

AI Usage Hours Per Day × 7

This gives a weekly perspective on AI usage.

Productivity Per AI Hour

Another derived metric was:

Productivity_Per_AI_Hour

This compares the productivity score against daily AI usage.

It helps provide another analytical perspective rather than simply looking at productivity and usage independently.

AI User Segment

Users were also categorized into three segments based on daily AI usage:

  • Low AI User
  • Moderate AI User
  • High AI User

This creates an opportunity for deeper segmentation in future dashboard pages.

Importing the Clean Dataset Into Power BI

Once the cleaned CSV was ready, I opened Power BI Desktop.

From the Home tab:

Get Data → Text/CSV

I selected the cleaned CSV file and loaded it into Power BI.

Before building visuals, I checked the data types.

This step is extremely important.

Fields such as:

Age

Monthly_Income

AI_Usage_Hours_Per_Day

Monthly_AI_Cost

Productivity_Score

Accuracy_Rating

Satisfaction_Score

and

Time_Saved_Hours_Per_Week

should be treated as numerical fields.

Fields such as:

AI_Tool

AI_Purpose

Country

Gender

and

User_Type

should remain categorical/text fields.

Creating the Six KPI Cards

At the top of the dashboard, I created six KPI cards.

1. Count of Users

This card uses:

User_ID

with aggregation:

Count

The result is:

116

This tells us that the final analysis contains 116 complete user records.

2. Average AI Usage Per Day

Field:

AI_Usage_Hours_Per_Day

Aggregation:

Average

The dashboard shows:

3.07 hours/day

This means the users in the cleaned dataset report an average AI usage of approximately 3.07 hours per day.

3. Average Productivity Score

Field:

Productivity_Score

Aggregation:

Average

The dashboard shows:

5.19

This is the average reported productivity score across the 116 users.

One important point is that we should not describe this as 5.19 out of 10 unless the dataset documentation explicitly confirms that scale.

A safer interpretation is:

Users reported an average productivity score of 5.19.

4. Average Satisfaction Score

Field:

Satisfaction_Score

Aggregation:

Average

The dashboard displays:

6.03

This represents the average reported satisfaction score among the users in the cleaned dataset.

It provides a high-level indication of how satisfied respondents are with their AI experience.

5. Average Time Saved

Field:

Time_Saved_Hours_Per_Week

Aggregation:

Average

The dashboard shows:

6.12 hours

This means the users in this dataset report saving approximately 6.12 hours per week on average through AI-assisted activities.

This is one of the most practically useful metrics because it translates AI usage into a time-saving measure.

6. Average Monthly AI Cost

Field:

Monthly_AI_Cost

Aggregation:

Average

The dashboard shows:

9.27

This represents the average monthly AI cost recorded in the dataset.

Because the dataset description does not clearly establish a currency for this field, I would avoid adding $ or claiming that this is USD unless the original dataset documentation confirms the currency.

Visual 1: AI Tool Adoption

The first visual is a horizontal bar chart titled:

AI Tool Adoption

The category is:

AI_Tool

The value is:

Count of User_ID

This visual answers a simple question:

Which AI tools are being used most frequently by the respondents?

In the dashboard, ChatGPT is clearly the leading tool, followed by Google Gemini, Microsoft Copilot, Claude, Perplexity AI, and Grammarly.

This is useful because it immediately shows the distribution of AI tool adoption.

It also demonstrates why cleaning the AI tool field was important.

If ChatGPT, chatgpt, and Chat GPT had remained separate categories, the actual adoption of ChatGPT would have been artificially divided across multiple bars.

Visual 2: AI Usage by User Type

The second visual compares:

Students vs Professionals

using the User_Type field.

The chart shows the proportion of users belonging to each group.

This allows us to understand the composition of the dataset.

The dashboard currently shows a relatively balanced split between students and professionals.

This is important because the dataset is specifically designed to include both groups.

For deeper analysis, this visual can also be changed into a comparison of average AI usage hours per day between students and professionals.

Visual 3: AI Purpose Distribution

The third visual is:

AI Purpose Distribution

It uses:

AI_Purpose

and counts the number of users.

The categories include:

  • Coding
  • Writing
  • Studying
  • Research
  • Content Creation
  • Data Analysis
  • Customer Support
  • Office Work

In the current dashboard, Coding is the largest purpose, followed by Writing, Studying, Research, and other activities.

This visual tells us why people are using AI, rather than simply which AI tool they use.

That distinction is important.

Two people might use the same AI tool but for completely different purposes.

Visual 4: AI Usage vs Productivity

This is one of the most analytical visuals in the dashboard.

It is a scatter chart using:

X-axis: AI_Usage_Hours_Per_Day

Y-axis: Productivity_Score

Each dot represents an individual user.

The trendline helps us understand the general relationship between AI usage and productivity.

The dashboard shows an upward relationship: users with higher AI usage tend to appear at higher productivity scores.

However, this should not be interpreted as proof that AI usage causes higher productivity.

A scatter plot shows an association in this dataset.

It does not establish causation.

The symmetry shading around the trendline also helps visually emphasize the spread of observations around the expected trend.

Visual 5: Productivity & Satisfaction by AI Tool

The fifth visual compares two metrics for each AI tool:

  • Average Productivity Score
  • Average Satisfaction Score

This is useful because productivity and satisfaction are not necessarily the same thing.

A tool might have users who report strong productivity improvements but relatively lower satisfaction.

Another tool might have high satisfaction but a smaller productivity impact.

The visual therefore lets us compare the perceived effectiveness and user experience of different AI tools.

This is more informative than simply showing which tool has the largest number of users.

Visual 6: AI Usage vs Time Saved

There was initially a mistake in the proposed visual where AI_Experience_Years was suggested.

That field does not exist in the actual CSV, so it should not be used.

The correct visual in this dashboard is:

AI Usage vs Time Saved

The fields are:

X-axis: AI_Usage_Hours_Per_Day

Y-axis: Time_Saved_Hours_Per_Week

The legend separates:

Student and Professional

This visual asks a very practical question:

Do people who use AI more each day report saving more time each week?

The dashboard shows an upward relationship between daily AI usage and reported weekly time savings.

Again, this indicates an observed relationship within this dataset, not proof of causation.

Adding Interactive Filters

One of the strongest parts of the dashboard is the filter panel on the right.

I added slicers for:

  • User_Type
  • Gender
  • Education_Level
  • Country
  • AI_Purpose
  • Profession
  • AI_Tool

This turns the dashboard from a static report into an interactive analytical tool.

For example, selecting:

User Type → Professional

allows the user to analyze professionals separately.

Selecting:

AI Tool → ChatGPT

allows us to see how the dashboard changes when focusing only on ChatGPT users.

Similarly, selecting a particular country or education level can reveal differences in AI adoption and impact.

What Does the Final Dashboard Tell Us?

The completed dashboard provides several important observations.

First, ChatGPT is the dominant AI tool among the respondents.

Second, AI is being used for a wide variety of activities, but coding is the most prominent purpose in this dataset.

Third, users report an average of approximately 3.07 hours of AI usage per day.

Fourth, the average reported time saved is 6.12 hours per week, showing that respondents associate AI with meaningful time savings.

The productivity scatter plot also shows an upward pattern between AI usage and reported productivity.

Similarly, the AI usage versus time-saved chart shows that higher daily AI usage generally appears alongside higher reported time savings.

But these relationships should be treated as observational insights, not causal conclusions.

Final Thoughts

This project demonstrates that building a Power BI dashboard is not simply about dragging fields onto charts.

The most important part happens before visualization.

We started with a 300-row dataset containing duplicates, inconsistent categories, text-formatted numerical values, and missing information.

After cleaning and removing incomplete records, we created a much more reliable dataset containing 116 complete records with zero missing values.

From there, we created KPIs, standardized dimensions, calculated additional fields, designed analytical visuals, and added interactive slicers.

The final dashboard tells a clear story:

Who is using AI → Which tools they use → Why they use AI → How frequently they use it → How productive they feel → How satisfied they are → How much time they report saving.

That is ultimately what makes a Power BI dashboard useful.

It is not the number of charts on the screen.

It is whether the dashboard helps someone understand the data and make sense of what is happening.

Disclaimer

This dashboard is based on a Kaggle dataset and represents the responses contained within that dataset. The results are observational and should not be interpreted as proof of causal relationships between AI usage, productivity, satisfaction, or time savings.