Embrace and extend Excel for AI data prep

Sometimes the best IT solution is the one you already have. Not always, of course: cloud infrastructure, for example, tends to provide much more flexibility and choice than private data centers. Unless it’s hey, in which case you’ll argue that a private data center is the way to go.

The key, as my colleague David Linthicum has stressed, is not to indulge in a “buzzword-oriented architecture,” where companies could “spend twice as much to modernize a workload that didn’t really need to be containerized, all because someone wanted to put containers on his CV.”

The problem is not the containers. Or cloud. EITHER [insert hot tech du jour here]. No, the problem is applying industry buzzwords to a business problem rather than letting the business problem dictate the solution.

Given how frantic companies are to apply the magic dust of machine learning to their business challenges, machine learning and artificial intelligence (ML/AI) is an area worth thinking about. Given the relative scarcity of ML/AI talent today, it’s worth looking at how you can best utilize the talent your company already employs, rather than praying that you can hire a data scientist to magically discover insights in your data. A better approach might be to make better use of the world’s most popular data tool to prepare data for machine learning models. Yes, I’m talking about Excel.

See past the ChatGPT hype

New advances in artificial intelligence are opening up opportunities for millions of people to start creating content of all kinds through machine learning, from code to art. Since its public release in November 2022, ChatGPT has grabbed headlines around the world and spawned a flood of commercial applications, along with many examples of abusive ChatGPT comments, fears of cheating on tests and exams, and more.

Google has released a Chrome extension called GPT for Sheets, which allows users to manipulate data with conversational language; Microsoft says it will integrate ChatGPT into all of its products, with Bing first. Microsoft recently invested $10 billion in OpenAI, the creators of ChatGPT.

But as exciting (and sometimes disappointing) ChatGPT apps can be, there’s a much more mundane and promising approach to machine learning already out there.

Excel riders, start your ML engines

I’ve written before about Akkio, a machine learning company that combines no-code and artificial intelligence., and how the Democrats turned the tool into a money printing machine in the 2022 election cycle. Akkio has launched Chat Data Prep, a cool new machine learning platform that allows users to transform data using ordinary conversational language. The technical term is natural language processing, but the less popular way to think of it is that it can transform the way Excel users work and allow them to accept the promise of AI much more easily.

It is estimated that 750 million people around the world use Excel. Microsoft CEO Satya Nadella has hailed Excel as the company’s most important consumer product. Turning Excel into a powerful machine learning tool could go a long way toward making machine learning something that everyday business employees can finally take advantage of.

“One of the things we were trying to figure out was how to build all the transformations you need on your data to use AI, even in our simple no-code ML platform,” Akkio co-founder Jonathan Reilly said in an interview. “Then we realized that we could use ML to accomplish this task. No organization wants financial planners spending their time importing, exporting, and manipulating data; they want them to focus on what the data tells them.”

Akkio’s new feature allows users to simply type in conversational language to make changes to their spreadsheet data. Leveraging AI and extensive language models, the platform interprets user requests and makes necessary changes to the data. It’s surprisingly easy. See for yourself in Akkio’s online demo (not closed).

The power of data for people

Why does this matter? You may be paying data scientists six figures to put your data to work, but most of their time is spent on data transformation, aka data wrangling. This is the technical process of converting data from one format, standard, or structure to another, without changing the content of the data sets, in order to prepare it for consumption by a machine learning model. Data preparation is the equivalent of cleaning work, although it is incredibly important work. The transformation increases the efficiency of business and analytics processes, and enables companies to make better data-driven decisions. But it is difficult and time consuming unless the user is familiar with Python or the popular SQL query language.

For example, there are several steps involved, starting with data cleansing (converting the data type and removing unnecessary characters). Here’s a hypothetical example of transformations someone who knows SQL or Python could do to harmonize multiple data sets for use in a machine learning model:

Convert year of birth to “Age”

Subtract the current year from Year_Birth.

Transform the customer enrollment date (“Dt_Customer”) to “Enrollment_Length”

It is similar to the above but with the addition of extracting the year part of the date function.

Transforms currency (“Revenue”) into numbers (“Revenue_M$”)

This involves four steps:

1. Clean the data by removing the characters “, $”.
2. Substitutes null value to 0
3. Convert string to integer
4. Reduce the numbers to the million-dollar format, which helps visualize the distribution of data.

And so on.

Not many of the three-quarters of a billion Excel users have even these basic programming skills. But any of them could write a simple request in plain English and Chat Data Prep will do the heavy lifting of data transformation. It even provides a preview of your results so you can verify that the result is what you wanted. Akkio claims that Chat Data Prep results in a 10x reduction in the time it takes to prepare data for analysis. With Chat Data Prep, users can change the format of dates, perform time-based math, and even fix messy data fields with a simple conversational command.

Making data analysis more accessible, efficient, and accurate is one of the mundane magic tricks that AI is increasingly making possible, quietly and behind the scenes. ChatGPT will grab the headlines, but your Excel users may be doing the heavy lifting of machine learning transformation within the enterprise.