When you were in school, was there always that one kid who never really did the reading they were supposed to?

(Maybe *you* were that kid?… no problem)

I had a friend who was “that kid”, and when it came time to turn in a book report, they’d watch the movie adapted from the book, write up a few pages of the plot, and turn it in.

That never ended with a good grade.

I’m talking about book reports based on movie adaptations because this is what AI *constantly* reminds me of.

Whenever I ask ChatGPT a question (and yes, even the most current version that costs a monthly subscription), the answer I get looks correct at first glance, but then on closer inspection, it’s just not quite right.

Chat GPT is like the tech version of the kid who watches the movie adaptation or reads the Cliffs Notes / Coles Notes of the book (for my Gen X and older readers) or reads the wikipedia summary of a book.

It’s technically correct, probably, but lacks depth and knowledge.

Pop Quiz, Hot Shot

I tested AI to see how it would fare if I asked it how to make a really easy chart in Power BI.

I picked a Bar Chart, since we all know what a Bar Chart is, and they are easy to make.

I tested the prompt “How do I make a bar chart in Power BI?” in ChatGPT 3 (the latest free version), Chat GPT 4o (the latest paid version that is billed as “Best for complex tasks”, and also the free version of Google Gemini.

The answers from all 3 of these AIs were *kinda* right, but were missing key steps and information AND are severely out-of-date in some respects. I expected more from Chat GPT 4o for actually correct instructions about how to make a very simple chart after paying for it, honestly.

Chat GPT 4o’s answer was almost word-for-word identical to the Chat GPT 3 version.

Here’s how the AIs did when I asked them “How do I make a bar chart in Power BI?”

Loading Data to Power BI:

The GOOD: They both started with loading/importing data.

The BAD: there is NO help regarding how data should be structured so you can make a bar chart. They both just assume you know, or it’s implying that ANY data can make a bar chart. Minor points to Google Gemini here for including a step of “Transform your data as needed using Power Query Editor” which, yes, is good to know, but it includes NO information about how the data needs to be structured. It even links to a website for this instruction, which you would assume have helpful instructions for transforming data into a form that you need for a Bar Chart.

My VERDICT: A little bit helpful, but missing a LOT of key information and help. Having an image or even a description of what kind of data structure is needed would go a long way.

Choosing the Bar Chart Visual:

The GOOD: They both recommend choosing the Bar Chart visual

The BAD: ChatGPT recommends choosing a ‘Clustered Bar Chart’ or a ‘Stacked Bar Chart’ depending on what kind of bar chart you want to create, with no help about when you choose a Clustered Bar over a Stacked Bar, or vice versa.

The UGLY: Chart choice often makes the difference between a good and bad chart. It would be nice to have an AI offer the pros and cons of each when saying “pick this or that”. While both kinds of bar charts DO allow viewers to compare categories of data, both are used for very different purposes. Also, it totally ignored the other bar chart type that comes with Power BI (100% Stacked Bar, which has a purpose unique from the other two).

My VERDICT: It knows that we’re making a bar chart (big whoop), but is totally ignoring the context that has to be known to make a GOOD Bar Chart.

Add Data to the Bar Chart

The GOOD: Both AIs correctly say to use a category field and a numerical field to make a bar chart.

The BAD and the UGLY: Both AIs are producing their answers from a version of Power BI that is at least 2 years old (as of now). The answers are very out-of-date. They both say to add a category field to the ‘Axis’ and a numerical value to ‘Values’. The problem is that these places where fields are added haven’t been named this for a couple years. They are called “Y-Axis” and “X-Axis” now.

This may not seem like a big deal for you, but it’s VERY important and confusing to people just starting out with Power BI. If the instructions don’t match reality, everything gets more frustrating.

The EXTRA UGLY: Both AIs note that you can add another category field to the Legend space in Power BI, but you almost NEVER want to do this if you want your chart to be easily understood. Adding a legend category splits all your bars into multiple bars and creates an interpretive mess of a bar chart. ONE Bar per category, okay? If you feel the urge to do this, email me first and ask me if you should. I’ll say NO! but give you better alternatives.

My VERDICT: The fact that even the paid “advanced” version of ChatGPT has data that’s 2 years out of date is a HUGE problem, because it makes their instructions technically wrong.

Formatting

The GOOD: Both AIs say “customize the appearance of your chart using the ‘Format’ pane.

The BAD: There is NO information about how to format to make a good bar chart that easy and accessible for users.

My VERDICT: Total FAIL. Good formatting can make ALL the difference between a good bar chart and a bad one, and in fact is the most important part of a Bar Chart. A great descriptive title, accessible colors, eliminating clutter.. these are all the bare minimum for Bar Charts, and AI doesn’t mention these at all.

OVERALL VERDICT

In summary, AIs provide the barest minimum of instructions about how to create a Bar Chart in Power BI, and if followed verbatim (with a lot of interpretation on your part) will give you a Bar Chart, but it will be a Bar Chart that sucks.

Vague, incomplete and out-of-date instructions is what we get from AI.

Seeing as how AI right now is basically an amalgamation of all the info online (which is good, bad, and ugly) it makes sense that their Power BI Chart instructions are horrible. Power BI instruction online is overwhelmingly inaccessible, user-unfriendly, and kind of obnoxious, assuming that there are some things you should just KNOW *before* opening Power BI.

Present company excluded of course.

Unfortunately, this incompleteness and vagueness is not limited to Bar Chart instructions from AI. I’ve asked them how to make many kinds of Charts, and the results are almost always bad or even completely wrong or impossible, like the AI has never actually used Power BI before.

This is because it hasn’t used Power BI before. It’s just saying stuff that *sounds* correct.

A lot like a kid writing a book report when they haven’t read the book.