Stay informed with free updates
Simply sign up to the Artificial intelligence myFT Digest — delivered directly to your inbox.
On a recent flight to France, I got talking to the woman next to me and ChatGPT came up. She worked as a bank clerk, and said she now used the chatbot to write the majority of her correspondence with colleagues and customers.
She explained that writing had never been her strong point, and the app was able to generate text that captured what she wanted to say better than she could. I joked that she would be using it to write messages to friends and family before she knew it, and she rather sheepishly replied that she already had, asking for help with a recent birthday text.
Her unease suggested that she felt what she was doing was in some way wrong. Yet like many others, she is now using generative AI to produce vast swaths of personal and professional communication, unbeknown to those who receive these AI-generated messages.
While companies explore how to leverage generative AI in ways that are profitable and transformative, a quiet revolution is already under way among their staff. Lawyers, bankers, doctors and many other professionals are now regularly using chatbots to write to colleagues, customers, clients and patients.
Company executives appear to underestimate how extensively their employees are using the technology.
Last October, a survey by Deloitte on uptake in Europe found almost half (47 per cent) of respondents familiar with generative AI had used it for personal activities, while a quarter (23 per cent) had used it for work. The report also revealed that a quarter of employees used models that they paid for themselves.
This is within two years of ChatGPT’s release, an app that costs $20 a month for the Plus version that comes without caps on usage. Last month, China’s free rival, DeepSeek, displaced OpenAI’s ChatGPT at the top of the app stores. Its popularity could lead to even wider adoption of generative AI.
Deloitte’s survey, which polled 30,000 consumers and employees across 11 countries, reported that 63 per cent of staff members utilising generative AI said that their company either encourages (44 per cent) or allows (19 per cent) its use for work purposes. But nearly a quarter (23 per cent) said their organisation did not have an AI policy.
This lack of clarity has led to a rise in shadow usage that means some companies have little to no idea what their employees are doing. While there are hopes AI can improve productivity, it will come at a cost if an absence of training and oversight means inaccurate information is communicated between staff or to clients. Confidential details could also be at risk if workers use unsanctioned tools that do not adhere to security and privacy standards.
Research by the European Centre for the Development of Vocational Training has found that one in seven adult workers in European labour markets now work with digital tools that complete tasks for them, with 22 per cent of employees using AI to recognise, translate, transcribe or generate text.
Most media organisations, including the FT, have policies stating that journalism cannot be created by AI. But what about communication with colleagues? Is it OK to use generative AI to draft an email to a manager asking for a pay rise, or to write a team member’s performance review?
There is also a risk that employees who use generative AI have too much faith in the technology. In the Deloitte report, 70 per cent of users said they would trust generative AI summaries of news articles. The figure was 64 per cent on personal matters like tax returns or benefits. Another UK-focused survey found that 36 per cent of users mistakenly believed AI was always factually accurate.
The output of AI tools is improving all the time, but errors are still common and increasingly difficult to spot. Last year, technology company Factal noted that AI models regularly drop journalistic attributions from summaries, such as removing “police say” from news stories. For a lawyer reviewing a case, omissions like this could be crucial.
And what of the way this technology is being rapidly woven into our personal lives? Does it matter if a birthday note is penned by AI, or is it still the thought that counts? Is a world where chatbots assist those who struggle to turn thoughts into words better than one where less is written, or is the erosion of personal voice something we should guard against? It is clear that more companies must provide outlines that explain how their employees can use AI to communicate. As a society, we should do the same.
Read the full article here
