Dispelling a terrible thing that Grammarly wants you to believe
I have been served a LOT of ads for Grammarly recently. They enrage me!
The most common scenario proposed in AI writing tool ads is the idea that you will be struggling to send a work email and could use an AI writing tool to perfect your wording. They don't tend to dig into the reason why you might be worried about this - they don't mention any cognitive difficulty with writing, though they do often hint at a work pressure for why you'd be very concerned with the work you're putting into a document like an email or a school paper.
I'm sure there are a lot of terrible bosses out there who are so horrible that every email you send is high-stakes. There are also a lot of jobs out there were sending high-stakes communications is part of the core skillset. But these aren't really the situations the ads are depicting - instead, these ads are trying to sell LLMs as a tool for sending routine business communications. They are trying to make it look natural to fuss over each email so neurotically that you gotta plug it into Grammarly before sending it. They soft-pedal whatever anxiety would drive someone to use an LLM to generate a routine business communication, then make the AI tool look like like a source of relief.
These ads are all embracing and even encouraging a kind of anxiety - like, anxiety-disorder level anxiety - that I find extremely dark and distressing. Check out this ad where a woman nervously slides a "tone slider" around on a three-sentence email and discusses the drawbacks of having the wrong tone - too friendly! too formal! - for the shortest, simplest possible communication:
I've worked office jobs for the last fourteen years, and none of those jobs required me to have perfect email composition. In fact, I spent the first five years of my career working extremely closely with many people who were not completely fluent in English. I was sometimes even in the minority in the workplace as a native English speaker. My coworkers sent short, oddly-worded, inexpertly-composed emails all the time, and nobody gave a single shit. These workers were not hired for their ability to speak or write English - they were hired to facilitate business meetings, negotiate with other companies, create art, write code, or produce videogames. Nobody expected them to write "perfect" emails.
What's more, my coworkers were not particularly hung up on the idea that they should be sending "better" emails. They had already hit the bar necessary for efficient communication. They could ask for what they needed from me, and I could understand it. They were communicating fearlessly at whatever text quality bar they were capable of achieving.
I find it extremely dark and horrible that these LLM ads are suggesting: hey, instead of communicating your thoughts naturally and fearlessly, why not get a little afraid? Why not get a little more neurotic? Plug each of your emails into this tool, and we'll find all those secret flaws you can't even see in your text. We'll make it appropriate for the world. You need us to mediate your thoughts, or they won't be fit to share.
I write for a living, but not everyone does. Not everyone's job description requires them to be communicating in "perfect," AI-curated English. There's a massive difference between communicating your goals and needs in a way that someone can understand, and communicating "perfect" text, finessed through a forest-burning AI via UI designed to make you believe that you have avoided some kind of serious, email-based interpersonal crisis.
But the bar is way the fuck lower than these ads suggest it is. I have not worked for a single company where this extreme level of anxiety over routine business emails between non-writer staffers could ever be justified. Even at jobs full of native English speakers, the expected communication quality bar was flexible enough for almost everyone to simply communicate fearlessly and efficiently in their own voice. I'm thinking of the various engineers I've worked with burning data center power to gain some marginal "improvement" to their fucking emails, and the idea makes me sick. And it makes me even sicker to think of everyone I know who spoke English as a second language feeling like they need to feed every communication through an AI tool in order to talk to their coworkers.
Because those tools aren't just going to prey on their anxieties - they're also going to intervene in their vernacular. You don't just communicate with your coworkers over email. You talk to them out loud, too. When you get to know someone, it's important to learn their particular way of communicating - their favorite idioms, their favorite buzzwords - so that you can communicate your needs and goals in any context, including verbally, where an AI cannot help you. The goal is never to communicate perfectly, but to negotiate for outcomes. If you're inviting an AI tool to intervene in that relationship, you are gonna fuck that process up and waste an incredible amount of time.
When you use an LLM to perform word substitution on something you wrote, then it's changing the meaning of what you wrote. Even if your concern about your self-expression is so intense that the idea of communicating in your own raw vernacular is terrifying, I personally would rather hear you choose the words yourself. Most of the time, we are gonna be using those words when we call one another, or talk to one another in person, or negotiate for a common outcome when there's been a misunderstanding. If you're constantly sending me heavily-ornamented LLM emails full of words you don't fully understand and would never have chosen yourself, then you're wasting everyone's time. I want your thoughts, not Grammarly's.
And let's be honest: if your workplace does consider these basic work emails so high-stakes that you and your coworkers all feel that you need to be sending each other LLM-generated emails, then something outrageously rotten is going on at your job. If you are all sending one another messes of text full of words that you don't fully understand and would never have chosen yourself, then you're not communicating at all.
The nightmare of a workplace that devolves into LLMs chattering at and summarizing one another's work to avoid humans actually communicating is probably an exaggeration, but I'd guess it's not by much. I've already heard enough about game devs engaging in this kind of behavior.
Specifically, I was told a story about a game dev team where one department refused to answer questions from another department, and instead urged them to run LLM summaries on "the docs". Of course, all the questions their colleagues wanted to ask them were about things not covered in the docs, or about emergent edge cases - the exact kind of thing that real humans need to talk about in their own words.
If your workplace is like this... run! Your bosses are fucking stupid.
And if you're writing sloppy emails or slack messages at maximum speed, communicating your real thoughts in your own vernacular at exactly the text quality necessary to get your work done and keep your coworkers happy... then you are perfect and good, and I love you. Grammarly wants to give you a fucking anxiety disorder. I hope they shit themselves and die.