ChatGPT is a popular AI virtual assistant that you can use to perform a variety of functions, including:
In its first iteration (GPT-3), ChatGPT had no functionality to upload files. Now, you can upload files for ChatGPT to ‘read,’ including PDFs, images – and Excel files.
Excel files are useful for formatting and editing data. Yet, if you’re looking to complete data-related tasks (such as summarising, extracting and visualising data), ChatGPT offers a fast and user-friendly approach.
ChatGPT generates information quickly, automating manual Excel-related tasks. Its latest iteration, GPT-4o, boasts an average latency of just 0.32 seconds, making it 17 times faster than GPT-4. DataCamp noted that GPT-4o’s latency is similar to a human conversation, making conversing with ChatGPT about your Excel data easier.
In this article, we’ll explore how to use ChatGPT to ‘read’ Excel files, how it works and how to make the most of ChatGPT’s ever-increasing functionalities. By the end, you’ll clearly understand what ChatGPT can and cannot do, and the various ways of automating Excel-related tasks.
To use ChatGPT, you’ll need to create an account. You can use an existing Apple, Google or Microsoft account or email address. If you already have an account, simply log in.
When you log in, you’ll see this interface:
Uploading to ChatGPT’s interface is easy – just click the arrow icon (this will appear when you start typing in the box). You can then upload an Excel file from your computer or device, Google Drive or Microsoft OneDrive. You will be able to upload up to 20 Excel files at once.
(Note: ChatGPT’s free plan only allows you to upload up to three files daily. However, their paid version allows users to upload 80 files every 3 hours).
ChatGPT now has the functionality to generate Excel files. When you hover over the blue ‘Download’ text, it will automatically download the file straight to your device (the expand button next to the text opens the Python analysis).
Its ability to quickly generate a downloadable Excel file makes it a helpful tool for completing various Excel-related functions.
You can submit a prompt to make ChatGPT ‘read’ the Excel file in the bar containing the paperclip icon. Once it’s read the Excel file, you can hit the button with the arrow pointing upwards, which will submit the prompt and the files. ChatGPT will then usually complete your request within a few seconds, though larger (or multiple) files may take longer.
Once ChatGPT has ‘read’ the Excel file, you can use it to perform several functions, including:
Let’s break these down:
You could prompt ChatGPT to perform specific data cleaning tasks like:
ChatGPT’s ability to quickly analyse large volumes of data makes it an effective tool for identifying and correcting outliers or errors. You can upload Excel files and ask it to identify anomalous data, including why it has marked the data points as anomalous.
You can also ask it to generate an Excel file with the anomalous data flagged, which should look something like this, depending on what it detects as anomalous:
When uploading Excel files, prompt ChatGPT to produce a summary of the data. Unless you specify otherwise, the summary will take the format of the extracted data, with brief summaries in italics.
As data extraction specialists, we’ve been working alongside technology like ChatGPT for years. Though ChatGPT can be a helpful tool for extracting data quickly from sources like PDFs, its tendency to hallucinate means you’ll need to review outputted critical data carefully.
If you need to extract data from a PDF into Excel, prompt the LLM to ‘extract the data from this PDF into a downloadable Excel file’.
ChatGPT uses DALL-E to generate images from text or Excel files. These images are then downloadable. Try a prompt like:
You can also customise the image’s colours or switch to an interactive format. In the interactive format, you can modify the image prompt in real time, adjusting the image until it’s ready to download.
The quality of ChatGPT’s response will depend on the quality of the prompt. The phrase ‘prompt engineering’ might be a LinkedIn cliché, but it is also an important skill for navigating virtual assistants. Although, according to one of our software engineers, it’s more of an art than a science.
If you’re looking to fulfil a more advanced function than simply asking ChatGPT to ‘read’ a file, it might be helpful to position the prompt. What does that mean?
It simply means prefacing a prompt by telling ChatGPT to adopt a persona. For example, you might start the prompt with ‘you are a data analyst’.
We’ve found that positioning the prompt yields a better quality of response, i.e. including more detail. Consider a few other personas that you might ask ChatGPT to adopt:
Though it might take extra time, qualifying the prompt will pay dividends. For example, if you add multiple file types, being clear about each type will prevent ChatGPT from generating confused outputs. Consider the following scenario:
Bad version of a prompt: ‘Read the Excel and fill out the PDF’.
An improved version: ‘Read the data from this Excel file and use it to generate answers to the questions in this PDF form. Each answer should be about 50 words’.
Asking ChatGPT to complete several tasks at once generally means the quality of its responses will be lower. Although it might take longer, it’s better to ask one prompt at a time while maintaining a clear link in the prompt to the original prompt.
Instead of leading ChatGPT towards certain assumptions (e.g. ‘Find all the mistakes in this Excel file’), consider the most unbiased wording (such as ‘Review whether there are mistakes in this Excel file’). Guiding ChatGPT towards specific conclusions can increase the likelihood of generating inaccurate or fabricated information.
ChatGPT is built on a large language model (LLM). LLMs can read natural human language in context. Unlike traditional Excel-reading tools such as Optical Character Recognition (OCR), ChatGPT can understand an Excel file within its internal context (i.e. how each data point relates to the others in the file) and the external context (e.g. how the data relates to industry trends, economic conditions, etc.).
Have you experienced an error? Not to worry; ChatGPT can often experience errors, and there are specific steps you can take towards resolving them. Let’s explore a few common scenarios.
ChatGPT’s free version has an upload limit of three files (per day). Consequently, you may need to upgrade to ChatGPT Plus for $20 to upload more files, which offers extended limits.
These extended limits permit up to 80 file uploads every three hours on GPT-4o. However, during peak times, these limits may be reduced. For unlimited document uploads, ChatGPT Pro costs $200 per month.
Or, if you don’t want to pay, you’ll need to wait 24 hours for ChatGPT’s document limits to reset.
It might be awkward to copy and paste ChatGPT’s output, as it will retain all of ChatGPT’s formatting. Instead, you can prompt ChatGPT to structure its output into a copiable format.
If none of these solutions work, you may want to try the following three general troubleshooting tips:
By clearing your cookies, you can enable ChatGPT to set up its (own) cookies, which might lead to a better performance. Switching browsers and logging back in has also helped when I’ve received error messages.
ChatGPT rarely goes down, but the service status will indicate whether it’s an error with your system or ChatGPT.
ChatGPT isn’t the only AI-powered virtual assistant that can read Excel files. Let’s explore the alternatives.
Claude is a tool developed by Anthropic, a company founded by OpenAI’s ex-employees. Claude had a prosperous year, releasing the much-acclaimed ‘Claude 3’ family. Part of Claude 3’s capabilities include reading Excel files.
Claude is founded around a constitution that promotes responsible AI. When considering which AI tool is the most ‘ethical’, Claude is likely the best choice.
According to Vellum AI’s 2024 LLM leaderboard, OpenAI (ChatGPT) outranked Claude, coming first in Multitask Reasoning, as in Massive Multitask Language Understanding (MMLU). Some analytical prompts for Excel files invoke ChatGPT’s MMLU abilities – such as descriptive analysis and predictive forecasting – meaning ChatGPT might outperform Claude’s in the accuracy of these tasks.
Gemini (formerly Bard) is a tool released by Google. Gemini’s free version cannot read Excel files – only images. Gemini’s paid version, Gemini Advanced, can read Excel files, but at £18.99 a month, it might be too high a price tag for some users.
You can integrate Gemini Advanced into Excel. Users can leverage custom functions, such as =GEMINI ("Your Prompt"), making it easier to analyse data than manually uploading Excel files to its interface.
Though Gemini is rapidly improving, ChatGPT or Claude 3.5 Sonnet outrank its reasoning abilities. If you’re looking for convenience, Gemini might be a good option. It also embeds into various Google Workspace applications like Gmail, Google Docs, Google Sheets and Google Slides, essentially functioning as an AI assistant.
If you have a strong use case for using ChatGPT to read Excel – for example, for an enterprise – you may need to implement some form of automation. Automation describes the use of technology to achieve a task with no or minimal human input. Several automation tools exist that vary in technical expertise required to operate them.
There are a variety of automation software that creates workflows by connecting programs. In the below example, we use Zapier, and we’ve worked with Workato. Other options include:
You’ll need to log in to the software and enter a carefully selected prompt (e.g. ‘When an Excel spreadsheet is uploaded to Google Drive, calculate its financial ratios and download these ratios as an Excel spreadsheet’). The program will generate a proposed workflow automatically, which you can then customise.
Once a trigger action is activated (e.g. a file is uploaded to a certain repository), the automation tool will complete the process without user invention.
Generally, automation software is straightforward (to operate), making it an easy option for individuals or businesses with limited technical resources. You could consider using an AI if you have a high volume of Excel files where the output needs to be returned quickly.
Another way to automate functionality is to connect to ChatGPT’s API. Though connecting an API requires (some) technical expertise, they’re designed to convey information quickly and are scalable.
To use ChatGPT’s API, go to https://platform.openai.com/docs/overview and create an API key. You will then need to install OpenAI Python SDK and create a Python script to interact with the ChatGPT API.
The Python script will then do the following:
It may take technical expertise to refine the API’s performance and debug any issues. If you need to analyse or output Excel files on a large scale (e.g. hundreds of thousands or millions), consider using a specialist vendor, as their staff should resolve any technical issues quickly.
ChatGPT cannot read Numbers files, so you must export it in Numbers as an Excel or CSV (i.e. click on File < Export < CSV or Excel).
ChatGPT can sometimes read .xls (the older Excel format). If you’re experiencing issues with ChatGPT reading .xls files, you can convert them in Excel to .xlsx or use an online converter.
Yes! It’s easy to connect ChatGPT to your Google Drive or Microsoft OneDrive. Go to ChatGPT and sign in to your Google account. Then, when you click the plus icon, it will pull up a screen to add files from your Google Drive. The files will automatically be ordered based on the last modified version.
ChatGPT is a helpful tool for completing various Excel-related tasks. The best way to become proficient at using ChatGPT for these tasks is to use it repeatedly. Through continuous experimentation, it will become easy to optimise ChatGPT’s results.
However, checking the outputted data’s accuracy for important Excel-related tasks is essential. If want to use ChatGPT to ‘read’ multiple files, consider paying for a premium account, as their free version limits are not overly generous.
Finally, if you want to convert files from PDFs or images into Excel, try Evolution AI’s free product, Transcribe. Email hello@evolution.ai or book a demo to find out more about our services.
You may have heard these terms referred to in the context of ChatGPT:
Context window: The amount of text ChatGPT can process in one go. The context window of AI virtual assistants is constantly improving, meaning they will be able to read higher volumes of Excel spreadsheets simultaneously.
Multimodal AI: An AI system capable of interpreting and processing various forms of input, such as text, images, audio and video. ChatGPT’s multimodal model can interpret whether you upload an Excel file or a screenshot of one.
Natural Language Processing (NLP): a subsection of AI focused on comprehending language like a human. For example, NLP allows an AI model like ChatGPT to understand how the terms ‘Excel’, ‘spreadsheet’ and ‘file’ relate to each other.
