Using AI to supercharge spreadsheets
Stan Marion on cutting down time-consuming tasks like translation, product description creation, and lead scoring.
Published 2024-12-11
Stanislas Marion is on a mission to make spreadsheet users more productive.
He’s the CEO of Talarian, a company focused on building simple, powerful tools for Google Workspace and Microsoft Office. And he’s passionate about enhancing productivity within the environments we already know and use.
I recently sat down with him for an hour where he demonstrated how one of his products, GPT for Work, can turn spreadsheets and documents into powerful automation tools.
If you’ve ever struggled with repetitive tasks in Excel or Google Sheets, or you want to explore how AI can streamline your data workflows, Stanislas’s workshop is one not to miss.
Here’s a breakdown of the tools and techniques he shared in this eye-opening session.
Automating translations at scale
One of the simplest yet most powerful features of GPT for Work is its ability to handle bulk translations. Imagine you have a product catalog in English and need to launch in multiple international markets. Translating hundreds or thousands of product descriptions manually would take days, but with Stanislas’s tool, it happens in minutes.
Here’s how the translation feature works:
Step 1: Select the column to translate.
In the demo, Stanislas showed how you can select a column containing product descriptions (or any other text) in your spreadsheet. For example, you might have a catalog of bicycles with descriptions in English.
Step 2: Choose your target language.
In this case, Stanislas selected French as the target language and used the tool to translate several rows instantly. But you can select any language you need. "You can translate entire product catalogs to different languages at scale, which is perfect for businesses expanding into new markets,” Stanislas explained.
Step 3: Customize the translation style.
Unlike basic tools like Google Translate, GPT for Work allows you to add instructions to customize the translation. You can specify a tone, target audience, or specific phrasing preferences, making the output much more tailored.
For example, Stanislas mentioned:
“If you want the translation to sound more formal or more casual, you can add these instructions, and the AI will adjust the output accordingly."
Step 4: Run it at scale.
Once you’ve confirmed the initial translation looks good, you can run the tool on all rows in your spreadsheet. Stanislas showed the tool translating hundreds of rows in less than a minute, demonstrating just how powerful this feature can be for large datasets.
Extracting structured data from unstructured text
Another standout feature Stanislas demoed was the ability to extract structured data - like prices or product weights - from unstructured text. This is especially useful when you’re working with catalogs or datasets that don’t include all the information in an easy-to-use format.
Here’s the step-by-step process:
Step 1: Select the unstructured text.
In the demo, Stanislas worked with a catalog where the product descriptions included weight and price mixed into a paragraph of text. For example, “This bike weighs 9.1 kg and costs $599.”
Step 2: Set your extraction parameters.
Using a simple prompt, you can instruct the AI to extract only the numerical values for weight and price. Stanislas showed how the initial extraction included unnecessary text, but with a refined prompt, he was able to extract clean data that’s ready for analysis. "This is critical because if you don’t clean the data properly, you can’t run any calculations or sorting," he noted.
Step 3: Apply the extraction across all rows.
Just like the translation tool, you can run this extraction across your entire spreadsheet, making it possible to clean up thousands of rows of data in minutes. This feature is particularly helpful for businesses managing large inventories or working with vendor-supplied catalogs.
Creating product descriptions with custom AI prompts
For eCommerce managers who need to write product descriptions, Stan’s tool can automatically generate descriptions based on key product details - like manufacturer, model, and price.
Here’s how:
Step 1: Set up your product data.
In the demo, Stanislas worked with a simple spreadsheet containing columns for the product’s name, manufacturer, weight, and price.
Step 2: Write a custom AI prompt.
Using the tool’s prompt box, Stanislas wrote a custom prompt to generate a product description, referencing specific columns in the spreadsheet. "If you leave it open-ended, the AI might return descriptions in Markdown format, which isn’t ideal for spreadsheets," he warned. "So always specify the format you want."
Step 3: Run the prompt across multiple rows.
Once you’re happy with the prompt, you can apply it across all rows, generating unique product descriptions for every item in your catalog. Stanislas demonstrated the power of adding variety to these descriptions by using multiple output variations. This ensures that the descriptions don’t sound repetitive across different products.
Lead scoring and CRM enrichment with AI
For sales teams, Stanislas also showed how GPT for Work can automate lead scoring based on multiple inputs like company size, industry, and previous interactions.
Here’s a step-by-step breakdown of the lead scoring process:
Step 1: Set up your lead data.
Stanislas worked with a lead spreadsheet that included columns for company size, industry, and previous engagement.
Step 2: Write a scoring prompt.
Using the tool’s custom prompt feature, Stanislas wrote a prompt to assign each lead a score from 0 to 100 based on the available data.
Step 3: Add the next actions based on scores.
Once the leads were scored, Stanislas took it a step further by using the AI to recommend follow-up actions - such as scheduling a demo or sending a follow-up email - based on the lead score. "You can even use this tool to suggest the next steps for each lead, which takes a lot of the guesswork out of lead management,” he told me.
Security and data privacy concerns
One important topic Stanislas addressed during the workshop was data security. Given that many businesses handle sensitive information, it's crucial to know how these tools manage your data.
How secure is GPT for Work?
Stanislas explained that GPT for Work uses APIs from trusted AI providers like OpenAI and Anthropic, which adhere to strict privacy policies. The data is encrypted both at rest and in transit, and the company is ISO 27001 certified, ensuring high standards for data protection. "We do not train models on your inputs or outputs, and all data is encrypted using Google Cloud’s infrastructure,” he confirmed.
Stanislas Marion’s workshop provided a detailed look at how AI can be used to automate tedious tasks in spreadsheets, whether you’re managing eCommerce catalogs or scoring sales leads. From translations and data extraction to lead management, the workflows he demonstrated offer huge time savings and improved accuracy.
Want to learn more?
Catch the full recording of Stan’s workshop (available to Pro Ben’s Bites members) to see his AI-powered spreadsheet workflows in action.
Like this post? You may be interested in these: