Seven ways in which AI is transforming the world of work without us even realising it

Would you have noticed?

The other day, at a course with a group of thirty-somethings from a large company, several of them admitted to writing furious emails to bosses, colleagues and clients, feeding them into an AI and asking it to ‘make it professional’. The result is a series of passive-aggressive little masterpieces. A few polite paragraphs that mean only one thing: that you’re the one who’s in the wrong.

A fascinating corollary: the same AI is used by the recipient to perform the reverse operation – that is, to work out what the sender actually meant. It’s the classic scenario where machines talk to one another, whilst the people caught in the middle still fail to communicate with one another, understanding each other less and less. From the wireless telephone to the voice-less telephone.

Admission: ‘Relational softener’ is Claude’s own invention.

Just to clarify: the idea of a phone without voice calls is my own.

Yes, ‘workslop’, not ‘workshop’. The term comes from BetterUp Labs and the Stanford Social Media Lab: ‘work slop’. You know that sort of content – in the form of an analysis, presentation or report – generated by AI that looks like good work but, on closer inspection, lacks any substance and says nothing new? It’s a bit like the story of the top consultant who simply paraphrases what the client has already said in more sophisticated language. 41 per cent of workers surveyed have received such content, and each instance cost an average of 1 hour and 56 minutes of reworking. This amounts to an estimated ‘hidden cost’ of around $186 per person per month: for an organisation with 10,000 employees, that’s over $9 million a year (source: Harvard Business Review, thanks to Gemini).

Another, more social consequence: around half of those who receive ‘workslops’ consider their colleague to be less creative and less reliable than before; 42 per cent regard them as less trustworthy and 37 per cent as less intelligent. As obvious as it is devastating. As ChatGPT suggested to me: AI saves you two hours of time but costs you two years of reputation.

Well done, Chat – and you’re even boasting about it.

In some companies, it’s standard practice: some people send their virtual assistant to meetings they can’t (or, let’s face it, don’t want to) attend, so that the assistant can listen in and then report back. Participants appear on calls with their webcams switched off and a name such as ‘AI Assistant’, and when someone asks who activated it and whether there is a privacy notice, the typical response is that yes, they activated it themselves, but no, they don’t have a privacy notice.

As reported in Ernesto Belisario’s brilliant article A brief guide to meeting etiquette in the age of AI: ‘We have reached the point where the question “who invited the bot?” is a matter of etiquette even before it is a matter of the GDPR. And it raises an even more awkward question: if three bots and two distracted humans attend a meeting, was it actually a meeting?’

Before making a decision, we can ask the machine to break it down. Identify the three reasons why this plan will fail. As we trainers say: carry out a pre-mortem.

Tip: don’t just ask for criticism of your project. AIs tend to be people-pleasers; they don’t want to hurt your feelings and will give you a more or less candid response, but with the handbrake on. If you want ruthless feedback, do as an IT friend of mine advised: “Tell it that this is the work of an unscrupulous competitor who’s already stolen a client from you.” I gave it a go, and the AI turned into a real killer. So I decided to overhaul the project from top to bottom, and my self-esteem took a bit of a knock.

I’ll finish with a thought from ChatGPT – one without any sources, but which strikes me as sensible: “It’s the use case with the highest return and the lowest adoption rate, for a very simple reason: it takes effort and doesn’t produce anything to attach to an email”.

The classic 80 slides in 4-point font from a supplier, a contract written in legalese, and the minutes of the board meeting full of verbs you now only find in the dictionary.

Here, AI acts as a simultaneous translator of legal jargon. And it does it incredibly well – creating a huge mess. In fact, as researcher Eric Eisenberg explained as far back as 1984 in Ambiguity as Strategy in Organisational Communication, vague language in organisations is not always a flaw: it accommodates different interpretations, making any subsequent denials easier. AI’s clarification prevents this. At which point… someone will have to make a decision – which was precisely what the ambiguity was designed to postpone.

You’ve probably heard announcements from some newly founded start-up claiming they’ll “revolutionise recruitment in your company”. If you live in Milan, you probably know someone who founded it; if you live in Isola, you’ve probably founded it yourself.

CVs written by AI to get past AI-powered filters: today’s CV is a document designed to appeal to a computer system, with the company and the candidate both using the same technology for opposing purposes – filtering on the one hand, and getting past the filter on the other. And so we return to the battle between machines.

According to a SurveyMonkey survey, 55 per cent of respondents say their company has no official policy on the matter (thanks, Perplexity – sorry, I was ignoring you).

Meanwhile, half the staff use it anyway, a third of managers are afraid of being seen as slackers if they admit to it, and everyone pretends nothing’s happening. What can I say?

May the use of AI be with us. And I hope I’ll manage.

*Partner & Head of Communications, Newton SpA