Copilot might soon be able to tell users if an email is good or bad, according to patent

What if we told you that Copilot might soon be able to understand subtext in an email you’re sending or receiving, so that you can send the perfect follow-up? You wouldn’t have to spend hours thinking about the a proper answer, Copilot could just do that for you.

Sounds exciting, right? Well, you should know that Microsoft is indeed designing or developing something of the sort, because we found a recent patent that describes such a technology. The paper, suggestively called Second Chance Message Enhancements, describes a technology that provides advanced feedback for messages you are writing or have sent in a messaging app.

It works in two main scenarios:

For Draft Messages, the system gets the text when you start typing a message. The system analyzes the message’s content using unique models or rules if certain conditions are met (like a long enough message or specific keywords).

Based on the analysis, the system gives the message a score that indicates how good or bad it is. If the score is high or low enough, the system provides detailed feedback to help you improve the message.

For Sent Messages, the system gets the text after you send a message. The system again uses unique models or rules to examine the message’s content and gives the sent message a score indicating how good or bad it is.

If the score is high or low enough, the system sends you an alert in the messaging app. If you interact with the alert, the system provides detailed feedback to help you understand and improve the message.

In both cases, the system uses feedback scores to decide whether to give you more detailed, advanced feedback. This helps you write better messages whether you’re still drafting or have already sent them.

We have reviewed the patent. The paper uses mockup images of Copilot, the said technology, so this is definitely about the AI model, as you can see in the features image above.

It would be interesting if Copilot ends up being able to identify the tone and subtext of each email. Microsoft experimented with an AI model in 2023, Project Rumi, that was capable of using the web camera to analyze the user’s facial expressions and react accordingly, to exceptional results, so this is more than feasible. Copilot is already capable of drafting an email, however, you need to tweak it with specific prompts to make it write a proper follow-up.

This technology would automatically make Copilot detect the tone and help you craft an email specific to each case. In this regard, the AI model might actually become useful, and the paper provides several examples: from dealing with difficult conversations at work to getting suggestions on how to ask for a raise or a promotion.

What do you think about it?

You can read the full paper here.