Why I’m Never Going to Let AI Write My Emails

Need some help writing your emails? Through the wonders of AI and Large Language Models (LLMs), you can now get messages composed on your behalf in Gmail, in Apple Mail, in Outlook, and in many other email clients. Most of the time, the AI option pops up straight away, ready to give you however much assistance you need.

The pitch is that you can offload the drudgery of dealing with email to AI, and move on to other tasks that may be more interesting and important. Anecdotally, I’ve spoken to quite a few people who now use AI chatbots in this way. But it’s not something I’m ready to embrace, and I don’t think I ever will.

These are my reasons, which may or may not resonate with you, though I haven’t mentioned the issues of energy use and copyright violations that hang over the use of AI more generally. You can keep asking if I want some help in Gmail, Gemini, but I’d rather switch you off altogether.

I don’t want to forget how to write

Writing’s pretty easy, really—most of us can do it from an early age without too much trouble. Writing well is harder, but you don’t have to be a best-selling author to fire off a few emails. So is there any harm in using AI for some low-level email composing? It may be quicker and more convenient, but I’m not sure it’s actually beneficial.

As author David McCullough once said: “Writing is thinking.” The skill of being able to choose the right word to put in front of the previous one gets the cogs of the brain moving, and forces some thought about what’s being said. Word choice and sentence structure matters, even on the shortest and most banal of emails.

I don’t want to sit down at a laptop one day and find myself struggling to compose a few lines of text. Is that far-fetched? Perhaps not, based on reports from those who’ve already tried farming out emails to AI. “Is it difficult trying to get that thought in your head translated to an email?” asks Google. Well, yes, it is, and that’s sort of the point.

People deserve a human response

It’s fair to say a lot of us get an excessive amount of email (if you don’t, consider yourself lucky). Chances are that plenty of your incoming email will be from people you don’t know personally, but no matter the sender and the recipient in an email conversation, I think human responses are worth the effort.

If all of our emails—arranging work drinks, applying for jobs, discussing a project—are written by AI, then we’re heading for piles and piles of machine-written missives that lack any kind of nuance or personal touch. Imagine a group email chain where every response sounds the same, irrespective of who sent it.

Even if I’m writing a simple “no thanks” email, if I’m communicating with another human being, I’m of the opinion that they deserve a response that has come straight from me. This is more of a principled stance than anything else, but I’m sticking with it.

AI writes a lot of generic slop

An AI email that sounds like a lot of other AI emails.
Credit: Lifehacker

Get AI to write a thank you note to someone who hosted an event you recently attended, for example, and you’re going to get a rather generic spiel that’s the averaging out of countless other thank you notes. It’s going to be bland, impersonal, and forgettable.

I can see the temptation to use AI to compose an important email—applying for a job, maybe, or appealing against a company decision—but your message is likely to end up reading like the algorithm-processed, mass-produced text that it is. You’re going to sound like everyone else, basically (see the previous point).

You could get AI to do a draft and then edit it, but if I started down this path, I could see myself making fewer and fewer edits to my messages, out of laziness or habit.

I don’t trust AI to get the details right

AI still makes lots of mistakes, though the chatbot developers don’t tend to mention them much. If you’re drafting an email about a new project pitch, a family get-together, a customer inquiry or whatever it is, there’s no guarantee that an AI will get all the details right.

The more important the content of the email, the more important this becomes. The companies pushing AI-driven emails seem to be of the opinion that we can all be chasing business leads, organizing colleagues, and expressing heartfelt feelings over email with the help of AI, but I’m not convinced.

People make mistakes as well, but I’d rather trust myself than a black box of algorithms that aren’t even fully understood by the developers who code them. Does AI know the people I’m emailing, and the specific details they need? Of course not.

AI talking to AI is not a future I want

We may not even need to click “send” in the future.
Credit: Lifehacker

To paraphrase George Orwell, if you want a picture of the future, imagine your AI sending thousands of emails a minute to everyone else’s AIs, forever. At what point do we abdicate responsibility completely to chatbots, and just let them get on with it? I don’t even want to take a single step towards that.

Right now, not even the most enthusiastic AI fans are suggesting that we start sending AI-written emails out into the ether without checking and editing them first, but isn’t that the obvious next step? I can almost see the Google I/O on-stage presentation now—get Gemini to handle everything, for the ultimate productivity boost.

Preliminary studies already show that we forget almost everything we write using AI, which has worrying implications if we’re sending out important information that needs to be recalled later. It’s not a future I’m going to be signing up for, no matter how insistent the AI prompts get.