Should You Actually Press Send?
AI has become incredibly good at helping us write messages. It still hasn't learnt when to tell us not to send them.
Many of us have become accustomed to the ease of AI. How it finishes our sentences. How it smooths a rushed thought into something polished. How it turns "I'm annoyed" into "I wanted to follow up on this."
We love its efficiency. As someone who uses AI heavily myself, it's difficult to go back — especially when we're trying to move faster than ever, with as little delay as possible.
But there's an area we don't usually think about when we're moving that fast. And that's exactly what we miss.
Researchers have a name for this. It's called automation bias: our tendency to trust an automated system's output simply because it came from the system, and to stop applying our own judgement once it has (Parasuraman & Riley, 1997). It was first studied seriously in aviation, where investigators found that pilots and crews sometimes deferred to what the automation told them even when their own senses were saying something different (Mosier et al., 2001). We've quietly handed that same trust to a chatbot.
Just because a message has been professionally written and made to sound acceptable… should it be sent?
Should These Be Sent?
"Hi team! A quick reminder to please enjoy all hot or strong-smelling lunches in the main kitchen area rather than at your desks. Because our workspace has limited air circulation, lingering food scents can be distracting for surrounding teammates. Thanks for helping us keep the desk area comfortable for everyone!"
"Hey Al! Quick favor, would you mind heating up and eating hot meals in the breakroom/kitchen instead of at the desk? The ventilation in this section of the office is terrible, and food smells just hang in the air around our desks all afternoon. Thanks so much for understanding!"
If your answer was yes… then this is exactly the problem.
"Strong-smelling." "Limited air circulation." Neutral, on paper. But ask which lunches actually trigger these emails in most real offices, and it's rarely bland pasta. It's curry. It's kimchi. It's the food that reminds someone of home.
"Smelly lunch" policies are one of the most common, and most rarely examined, ways an office quietly asks people from certain backgrounds to eat like they're apologising for themselves.
That's not an accident, and it rarely feels like a big deal in the moment. Many offices have outdated policies in place that were written long before any tool could tell them the policy itself might be the problem: policies that quietly create a poor workplace culture and make people from different backgrounds feel uncomfortable, or singled out. The Equality and Human Rights Commission has written about exactly this link: fair treatment at work isn't only about the big, obvious things, it's built from moments like this one too. They can't truly be themselves at work, while others can.
In an attempt to create consistency or uniformity, we neglect the plain fact that we're all different. Trying to flatten a workplace removes the dynamism that comes from the usual melting pot of an office. CIPD's own research on inclusion makes a similar point: belonging isn't built by asking everyone to fit the same mould — it's built by making room for the fact that they won't. We can't expect everyone to be the same.
The lunch email and the Slack messages below aren't really about lunch, or about Slack. They're the same move, wearing two different outfits: take the way one kind of person experiences the world, call it "neutral" or "professional," and quietly ask everyone else to make themselves smaller to fit it. Diet and culture in one. How a brain works in the other.
Read these again, but this time, imagine the person on the receiving end thinks differently to you. Not worse. Differently.
Should These Be Sent?
"Hey [Name], I noticed you got super absorbed in fixing those minor code edge-cases today, but the main sprint ticket is still untouched. You really need to practice better prioritization and not get distracted by rabbit holes. When you get side-tracked like that, it slows down the whole team's velocity. Can you promise to just stick strictly to the high-priority task list tomorrow?"
"Hi [Name], thanks for sending over your project updates. To be honest, your updates in Slack are always a bit scattered and jump all over the place. It takes the team extra time to piece together what you mean. Going forward, please structure your thoughts sequentially before messaging us. Think through the entire plan from A to Z first so you aren't sending piecemeal info."
The answer is NO.
"Rabbit holes." "Scattered." "Structure your thoughts sequentially before messaging us."
To a neurotypical manager, that's just direct feedback. To someone who processes and communicates differently, it can read as being told the way their brain works is a problem to be corrected. The British Psychological Society's neurodiversity guidelines and CIPD's neuroinclusion research both come back to the same point: workplaces are usually built around one default way of thinking, then act surprised when anyone who doesn't fit that default struggles to belong. Neurodiversity isn't a deficiency to manage around. It's been shown to be a genuine competitive advantage when organisations actually build for it, rather than quietly correcting people out of it (Harvard Business Review, 2017).
For the most part, most messages like these appear neutral. Typical feedback from senior leadership. Nothing dramatic. But that's exactly how it stays invisible: this kind of message can create real toxicity for some members of a team while looking, to everyone else, like Tuesday.
Most tools like Copilot won't question whether the message needs to be sent at all. They'll just format it in the most professional, easiest-to-read way possible. Efficiency gets optimised for. Humanity doesn't.
And because of that efficiency, we lose our humanity — but at the same time, we've never actually been told it wasn't okay. It's time for a rewiring. And a bit more reflection.
Most people will avoid complaining about messages that make them uncomfortable at work, because they don't want to draw attention to themselves. CIPD's research on employee voice describes exactly this pattern: people staying quiet not because there isn't a problem, but because raising it feels riskier than absorbing it. So it looks like there isn't a problem. There is one. It's just been accepted as normal.
Google's own internal research into what makes teams effective found that psychological safety — the sense that you can speak up, be yourself, or make a mistake without being punished for it — was the single biggest factor separating high-performing teams from the rest (Project Aristotle). Every message like the ones above chips away at that, one Slack notification at a time.
After the training, everyone feels like they've fulfilled their duty and ticked the box. But are they truly being intentional? Are they being consistently trained to be mindful, day in and day out, or did it end the moment the session did? It's easy to have blind spots, if you've never been told that you're wrong. And most of us don't get told. Not by a colleague. Not by a manager. And never by the tool that just made the message sound so reasonable.
This Is Why Tonely Exists
Not to write the message for you. Not to soften it into corporate-speak so it slides through unnoticed. Just to be the pause that automation quietly removed: one second, before you hit send, to ask whether the version of you that wrote it fast is the version you'd want someone else to receive.
AI has become incredibly good at helping us write messages. It still hasn't learnt when to tell us not to send them.
So I ask you — should you actually press send on that message?