For more crisp and insightful business and economic news, subscribe to The Daily Upside newsletter. It's completely free and we guarantee you'll learn something new every day.
Microsoft wants to help make emails to that one annoying client sound a little less harsh.
The company is seeking to patent a system for "automatic tone detection and suggestion" in messages. Basically, this helps a user determine the unintended attitude that their messages may give off, whether it be too curt and aggressive, too wordy, or downright inappropriate.
"Sometimes while creating content, the user may be unaware of the emotional attitude carried by their content," Microsoft noted in its filing. "Furthermore, while some users may notice that the emotional tone carried by their content is inappropriate, they may find it challenging to change the language to convey a proper tone."
Here's how it works: Using a machine learning model, this system detects a user's tone within a section of text, taking into account the "content environment" that said text is placed. For example, a content environment can include an email, an instant message, a social media post or a document. The system then makes suggestions on whether a user should modify or rephrase that text depending on the tone and the environment. These suggestions may show up as a sidebar of the content environment's user interface.
Microsoft notes its system can catch an array of tones, including angry, accusatory, disapproving, optimistic, forceful, encouraging, egocentric, unassuming and surprised. For example, if you send an email to your coworker saying, "your proposal is a terrible idea," the system will flag that as being too aggressive and accusatory.
The company says in its filing that conventional tone detection and rephrasing services don't take into account the environment that the content is in, making them less intuitive to when informal versus formal language is acceptable.
Microsoft's interest in AI-based productivity tools is far from unprecedented. On the patent side, the company has filed for plenty of AI-powered workplace and monitoring tools, including one that monitors if you're working off the clock.
More publicly, the company has touted its upcoming 365 Copilot, which integrates large language models with its flagship apps like Word, Excel, PowerPoint, Outlook and Teams. Microsoft is also reportedly looking to integrate AI into its built-in Windows features, like Photos, Snipping Tool and Camera.
But there are several reasons a tool like this may not fly. First, communication norms differ between cultures and generations, and can be more granular at the organizational and individual level. What the AI may deem as an angry email could just be the blunt communication norms built into a company's culture. Alternatively, some companies may be more casual with one another than others. (For example, as a member of Gen Z, I speak in a much more sarcastic and casual manner than some of my colleagues. Would this system call me out if I said "slay" in an instant message?)
At the end of the day, the effectiveness of this model will differ for every organization that uses it, because AI can only understand context to a certain extent. (FYI, this is an issue that comes up in speech recognition contexts a lot.)
Another consideration: This system has a feedback mechanism to optionally train the model to understand organizational communication norms. If a company has a particularly aggressive or toxic manner of speaking with one another, the system could take that and replicate it further.
And finally, this adds another layer of monitoring that employees simply may not want. Similar to Microsoft's patent to track off-hours work, while the intention may be a good one -- in this case, aiming to make communication smoother -- it could make employees uncomfortable to have an AI model that watches as they type.
