Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
Once upon a time, the basic rules of workplace etiquette were well understood.
Don’t take credit for someone else’s work. Don’t waste other people’s time by being late. Try to thank anyone who has done something to make your life easier. And so on.
The arrival of AI has added uncertainty to proceedings.
This month, Deloitte released a survey of 25,000 workers in the UK that it claimed was the largest study of workplace generative AI use ever carried out in a single country.
It showed that 63 per cent of us are using AI tools at work, mostly for three reasons: to search for information, write emails and summarise stuff.
The first of these poses few problems. The fact that ChatGPT can find studies and data in seconds that I would struggle to locate elsewhere helps me and offends no one.
Emails are another matter. AI’s view on how I might like to reply to a work email has improved since its chirpy early suggestions of “Love it!” and “Awesome!” — phrases I almost never use.
But the more sober versions it has come up with lately are still far from perfect. Too often they are cold, robotic or not quite what I would write. I never use them because I feel sure that anyone receiving them would suspect AI had had a hand in writing them, and might therefore think I was lazy, rude or both.
It turns out this is a reasonable assumption. “People can detect AI really easily,” says Myka Meier, founder of the Beaumont Etiquette firm that offers professional conduct training to large businesses across the globe.
Its relatively new AI etiquette courses are popular, she told me, and emphasise the importance of the personal, aka human, touch.
AI’s ability to summarise long documents also needs to be treated with caution, not least by managers.
I sympathise with time-starved bosses for whom AI summaries offer great temptation. But a friend recently showed me an email her manager had sent as part of a regular performance review.
A lengthy discussion the pair had held had been reduced to a few stilted paragraphs that reeked of AI and glossed over many of what my friend said were her most important achievements and goals.
The lesson here is clear. It is one thing to rely on AI summaries of lengthy transcripts or reports that only you are relying on, but another if you are condensing information of importance to someone else as well.
In that case, quietly using AI smacks of a lack of care and attention.
The same goes for another egregious breach of AI etiquette: sending AI-assisted work to a colleague, such as a reply to a customer or an assessment of a rival, without first checking it is all accurate.
This is a quick way to transfer work from yourself to someone else, and it has become alarmingly common. More than a third of US workers say they have been on the receiving end of colleagues’ workslop, which costs them more than three hours a month to fix and makes them understandably angry.
This raises another aspect of AI etiquette that does not get as much attention as it should: being transparent about using AI tools.
Burdening people with workslop is bad enough. Recording an online meeting without warning all participants is far worse.
Yet the share of meetings being recorded has been soaring. So too is the practice of sending an AI note taker to a meeting without alerting the humans gathered. It is not unknown for people to turn up to meetings to find themselves outnumbered by AI note-taking bots.
Again, a small amount of human communication goes a long way. If you explain beforehand that you have an unavoidable clash but would still like to use a note-taker if all agree, that may be acceptable.
If you don’t, you risk looking presumptuous and rude.
Ultimately, these rules are not hard to understand. They differ little from what has always been required. Before AI made bots so cheap and easy to use, would you have ever gone to a meeting room and left your phone on the table to record the conversation while you headed off for lunch? No. Me neither. Just because tech makes something possible does not make it acceptable. One day this will be obvious. In the meantime, we must muddle on as best we can.