ChatGPT’s new Code Interpreter plug-in adds powerful data science abilities to the popular chatbot.
Sit down and have a chat with your donor data. That’s what using ChatGPT’s Code Interpreter feels like.
Many fundraisers have a love/hate relationship with data.
Data analytics is the solid foundation for our work as fundraisers and the key means of measuring our successes. But working with data often feels frustrating, difficult, and disempowering. The tools that we use — donor CRMs, spreadsheets & accounting software — require investments of time and training to use effectively.
But maybe it doesn’t have to be that way.
In recent months we’ve seen the appearance of AI-powered assistants for creating art & images (Midjourney, Dall-E), and texts (ChatGPT, Bing) that can reduce complex skill-based tasks into simple text requests.
What if we had an AI-powered chatbot for data analytics?
What is Code Interpreter?
Code Interpreter is a ChatGPT plug-in that adds-in skills for reading and analyzing data, generating insights and visualizations and even programming code. It runs in the same interface and responds to plain text instructions, same as ChatGPT.
Here’s a walkthrough of a few basic data tasks:
Start by uploading a data file with recent donation transactions (see important points below about properly sanitizing your data before uploading) and ask for a simple summary.
Request a simple chart
Then a more complex chart:
Then a more complex data analysis.
Code interpreter is able to design and run number-crunching processes, and also offer explanations about the results. It’s like you’ve invited a helpful fundraising data assistant to pull up a chair and join your conversation — ready to answer your data questions, run analyses, and generate charts and graphs on the fly.
It’s a refreshing break from the grind of report-building in CRM apps or Excel.
Here are some typical questions that Code Interpreter can answer:
- How many donations did we receive from (XX channel or location) this month?
- What is the average donation amount from our last three fundraising campaigns?
- Is the average donation amount trending upward or downward?
- How did the response rate through Facebook compare with our display ads in the last quarter?
Code Interpreter is also capable of more advanced data science functions, including writing and executing Python code for machine learning models.
The rise of data chatbots?
So far, Code Interpreter isn’t much more than a demo. It doesn’t have the data processing and reporting capabilities to match an actual CRM or analytics suite, and won’t replace mainstream data intelligence products anytime soon.
But as we’ve seen with ChatGPT, progress in the practice of prompt-engineering has unlocked advanced capabilities in text chatbots, and this may also be true with data chatbots. ‘Prompting for Data Analytics’ could become another in-demand job skill.
We can expect other chatbots like Code Interpreter to appear soon, likely within our familiar fundraising CRMs. This could be good news for fundraisers who struggle to get satisfactory data intelligence from their current tools.
Data chatbots may also help broaden the tent for data literacy by making hands-on data work accessible to people who don’t have analytics skills — and might even make working with data fun.
cautions about data chatbots
There are a few caveats:
- Despite its impressive skillset, Code Interpreter isn’t a data scientist. It can’t tell you what questions to ask, or what meaning to attach to its data outputs. What it can do is make it easier for you to explore different ways of looking at your data.
- For Code Interpreter to work, you need to give it some data. It will accept uploads of up to 500 Mb, and it can run analyses across multiple files (if they have some data connection between them). How you select and prepare your data files for upload will determine the kinds of data questions that Code Interpreter can answer.
- Data security should be a top concern anytime you are transferring data, especially to cloud-based systems. While OpenAI promises full data privacy and will not hold your data past 30 days, you should always ensure your data has been scrubbed of any fields with personal information — name, address, phone, email, postal code, etc.
- Be aware of bias in your data. For fundraisers, remember that your CRM data is a historic snapshot. It’s a picture of how your campaigns performed in the past, and may not match the priorities you have for the future (reaching new audiences, new campaigns, mobilizing new resources).
Code Interpreter is only available for ChatGPT Plus users right now ($20/month).