You select a small group of cells in a spreadsheet, hold your breath, click the chart tool, and...the graph is incomprehensible. With some research and a bit of practice, you could probably fix it. Just like the last time you needed a chart. And the time before that. If making charts is so easy, why does it always take so long? 

Data visualization is finicky. Tick the wrong graph setting or highlight the wrong numbers, and everything breaks. Or maybe you’re not even sure what to select to begin with. You’ve got hundreds of rows and dozens of columns of data, so you fuss around with the settings hoping to stumble across something worth charting. Well, now that OpenAI’s Code Interpreter plugin is publicly available, you never have to Google “which axis is x” again.

Give ChatGPT a database, spreadsheet, or even just an image of some semi-organized information, and ask for help in plain language, laymen’s terms. In seconds, you’ll have a chart, graph, or data summary. Heck, you can even ask for the visualization to be interactive if you want. Like so many recent AI developments, the main limitation is your imagination. So first, let’s talk about some things you can do with this incredible tool:

• Get answers to analysis questions based on several variables
• Clean up messy data sources and merge two databases
• Use natural language prompts to turn data into visualizations
• Ask open-ended questions about trends and insights hiding in CSVs, images, and PDFs

The implications of this are huge for any business intelligence use case, but especially for digital signage. Grab a recent spreadsheet, or use one of the example databases we link to below, and let’s see just how easy it is to master Code Interpreter and get your data ready for the big screen.

How to enable ChatGPT’s Code Interpreter

OpenAI introduced ChatGPT plugins in May, 2023, to let the chatbot access information outside of its carefully guarded digital playground. The initial list of plugins ranged from OpenTable’s restaurant-booking tool to Zapier’s automation platform. It wasn’t until July that Code Interpreter saw its alpha release, and even then, enabling it requires a few extra steps. 

First, you need a $20/month ChatGPT Plus subscription. Create or sign in to your OpenAI account. Then, log in and click “Upgrade to Plus.” Select the "Upgrade Plan" option, input your payment details, and complete the checkout.

Once you’ve subscribed to ChatGPT Plus, head to the Settings menu and look for the "Beta features" section. Then, hit the "Code interpreter" toggle. Now you can use the plugin when chatting with ChatGPT. Just hover over the GPT-4 toggle at the top of the screen and select a Code Interpreter chat.

Keep in mind that while we’re mostly going to cover using Code Interpreter for data analysis, it’s more than that. It can also host and run Python scripts from inside the chat interface. It just so happens that one of the best use cases for that is interacting with data and visualizations. Here are some examples.

Ask questions about your data and its properties

The first and most basic experiment is dropping a CSV file into the chat and asking a basic question. If you don’t have any data ready to work with, grab a free Airtable account and export one of their templated databases that comes with dummy records. We uploaded the Sales CRM Template and asked, “How many proposals does Sales currently have out?”

You can watch Code Interpreter essentially have a conversation with itself as it parses through your question. This is really helpful because it lets you see where things might go wrong. For example, it may have noted that it based its answer on the Expected Close Date field instead of Status. You’d see what went wrong and tell ChatGPT how to fix it. 

While that was a pretty simple request, there’s nothing stopping you from adding multiple variables to a single question. You could ask, “What was the win/loss ratio of our deals in May compared to your win/loss ratio for all of 2020?” As long as you know how to phrase what you’re looking for, you can get an answer in seconds. 

Although we tend to recommend more automated solutions for digital signage content, this “manual” solution makes it exceptionally easy to generate interesting and timely updates to share with your team. That is, as long as you have decent data to work with. 

Clean up messy data and merge databases

When you’re adding new records and entries every day, it doesn’t take long for your spreadsheet or database to get ugly. Maybe you need to add new fields or rename others, which breaks some third-party automation, and all of a sudden, you have a thousand records stored in a separate file. Messy data is unavoidable. Or, at least, it was. 

Now, you can use Code Interpreter to clean things up for you. No tedious manual reformatting necessary. You could, for example, upload two similar but slightly different databases from Data.gov: Active Businesses in LA and Business Owners in Chicago. Both of them have information about registered companies, so let’s merge them into a single data source with a single field for business names. 

We probably didn’t need to create such a detailed prompt, but it’s a good habit to provide as much context as possible when working with Code Interpreter. As long as you tell it which columns in each source overlap, that’s more than enough context for it to get to work.

ChatGPT recognized that there were multiple ways to tackle this problem and gave us some options. Granted, it threw some fairly technical data analysis terms at us, but this is the same chatbot that kids are using to do their homework. Just ask it to dumb down its explanations.

The important takeaway here is that we ended up with a new downloadable CSV file of the merged databases. Beyond the typical question-and-answer back and forth, ChatGPT can now give you a file to work with outside of the chat interface. Slap a profile photo of an overeager junior analyst next to the responses, and you could fool most people into thinking they were working with a real person.

Transform data into visualizations

At the end of the day, you probably don’t want another CSV. You want a graph or chart that shows trends in the data, which you can then put on your digital signage. And that’s a piece of cake when all it takes is a simple, natural language request. Imagine that you’re the floor manager for a manufacturing plant. You’ve got a spreadsheet full of workplace safety data (like this one) and want to see if incidents tend to happen more often on certain days of the week.

Upload the safety data and ask for a graph of accidents based on which day of the week they happened. Obviously, you could request a specific type of graph if you have an idea of what you want. Or, you could simply describe what you want to visualize with using any technical charting or analysis vocabulary. For example, you might say, “I want to see if certain incident types are more likely to occur on certain days of the week. Can you show me in a graph or chart?”

An updated graph is not all that it gave us. Code Interpreter also added below the updated graph that “It seems that ‘Vehicle’ incidents occur more frequently on Friday compared to other days. Similarly, ‘Cut’ incidents appear to be more common on Tuesday and Wednesday.” With those insights, you could create digital signage playlists with safety warnings specific to each day of the week based on your existing data.

Ask open-ended questions about insights and trends

So far, we’ve been asking Code Interpreter for fairly specific outputs. We knew that we wanted to compare the sales win/loss ratio of May to the win/loss ratio of the whole year. Or to see which safety incidents were most likely to occur on certain days of the week. What do you do when you don’t have any ideas? When you have no clue what your data is trying to tell you? Arguably, this is the most powerful analysis feature of Code Interpreter.

This time, let’s give ChatGPT a spreadsheet with 1,000 rows of work order data (thanks to Contextures.com’s example data). We want to reveal any data-backed observation that might help us increase efficiency or profitability but have no idea where to start.

After explaining its interpretation of what each field means, ChatGPT went on to provide a whopping 900 words and six graphs worth of recommendations. Including “Some technicians, like Lopez and Cartier, are handling significantly more work orders than others. If the workload is too heavy for some technicians, you may consider redistributing the work more evenly to prevent burnout and maintain high-quality service. Alternatively, if certain technicians are handling more work due to specific expertise, you may want to invest in training other technicians in those areas to balance the workload.” 

From there, you could ask ChatGPT to write the JavaScript code to turn your Work Orders By Lead Technicians graph into an automated Chart.js visualization for your signage. Or build an automated dashboard for displaying training progress for individual technicians.

Keep in mind that you don’t have to give Code Interpreter a spreadsheet or database file. It did well with tables of data in PDFs in our tests. And it’s supposed to be able to use optical character recognition to identify numerical data in images like screenshots, but consistently failed when we tried. Undoubtedly that will get better with time.

Adding graphs and charts to your ScreenCloud signage

Once you have a visualization that you want to share with your team, it’s time to think about how you want to put it up on the screen. The first and easiest option is to simply upload a graph or chart as an image in ScreenCloud’s Media tab. Or maybe Code Interpreter helped guide you through the process of creating a visualization in your spreadsheet and adding it to Google Slides. Just add the Slides integration to ScreenCloud, and any time your spreadsheet is updated, you’ll see the changes on your signage.

The same goes for any app that generates embed codes. Turn Code Interpreter’s analytical insights into an Airtable view and embed that on your signage. If you know your way around Zapier and webhooks, you could also send data directly to ScreenCloud’s Application Data store and code basic visualizations with JavaScript and CSS.

Data analysis and visualization is great for boosting the effectiveness of internal communication efforts. Especially when you set up AI tools like ChatGPT to automatically post to your signage. If you’re lucky, you’ll never again have to check and uncheck random boxes in your chart settings to create something passable. Sign up for a free trial of ScreenCloud today and build something amazing!