An automated tool to police offensive language

A startup has developed an AI tool that can copyedit written text with an eye toward specific corporate style and potentially offensive language.

Why it matters: The shift to distributed work means employees are spending more time communicating internally and externally in writing, even as norms for what's acceptable in workplace speech keep evolving.

• AI tools powered by improving natural language processing (NLP) can help automate that oversight — even if in practice it can feel a little intrusive.

What's happening: San Francisco-based Writer's AI tool scans through written text of all kinds — emails, Slack messages, blog posts — to identify not just grammatical or spelling mistakes, but language that clashes with a company's preferred style or could come across as potentially biased.

• "We think of ourselves as the single source of truth for language that's been approved strategically by a company," says May Habib, Writer's CEO.

How it works: As a user writes, Writer will highlight problematic or weak language and offer possible alternatives.

• Write a message along the lines of "Is that the best you can do?," and Writer will identify it as potentially passive-aggressive.
• Use a term like "blacklist," and it can get flagged as potentially biased.

Details: Customers — which include large firms like Deloitte — can tweak Writer's out-of-the-box style guides to their specifications, says Habib.

The big picture: Writer is one of a number of automated writing assistants on the market, including Grammarly and Textio — a sign of both the growing importance of written communication in the digital age and the growing ability of machines to do some of that work for us.