‘That's great! Good work!' The AI spambots taking over email

Telegraph

04-05-2025

An estimated 350bn emails are sent each day. British office workers receive as many as 10,000 per year, according to one study by Warwick University. Your typical journalist would probably see five times as many.
Inboxes are bursting – and now AI bots are being added into the mix.
AI chatbots, writing tools and automatic replies are flooding the email products used by hundreds of millions of office workers.
With the emergence of tools such as ChatGPT, Google, Apple and Microsoft have started piling their email products with AI widgets in an attempt to outmanoeuvre their rivals and drive new growth. To do so, they must convince their users and customers that these tools are actually worth using.
From Apple's Writing Tools AI to Google's Gemini prompts, tech giants are seeking to make boring emails more productive and convenient for internet users. But glitches and AI hallucinations – where the bots make up facts – have threatened to dent trust in the technology.
Communications experts, meanwhile, are sceptical that shoehorning AI into emails will make workers as productive as big tech claims.
'When someone realises they are interacting with a chatbot, especially when it wasn't disclosed, it can really backfire,' says Andrew Brodsky, assistant professor at the University of Texas at Austin and author of Ping: The Secrets of Successful Virtual Communication.
Difficult to disrupt
Technology giants have long attempted to reinvent email, which is one of the earliest forms of digital communication, pre-dating the World Wide Web.
However, despite the rise of corporate chat tools such as Slack, sending emails has been difficult to disrupt. The basic format remains strikingly similar to what it looked like in its early years and it has remained essential to almost all white-collar workplaces.
AI prompts in email are not entirely new. In 2017, Google added Smart Reply to Gmail, which can offer up one or two line answers such as 'that's great' or 'many thanks'.
However, these answers have grown more sophisticated thanks to the advent of large language models (LLMs): technology that allows AI chatbots to provide detailed answers to prompts in plain English.
On Gmail, for instance, users can choose to bring up the company's Gemini chatbot and ask it questions. They can ask it to draft an email in different styles, making the text more formal, longer or shorter.
Gemini can also summarise the contents of long conversations or email threads. Microsoft's Copilot, meanwhile, can generate emails or provide 'coaching' on writing style to users.
Apple also offers AI-generated Smart Replies in its Mail app and has a product called Writing Tools that can rephrase and draft emails for users.
In an advert for its AI technology, Apple showed its bot rephrasing an off-the-cuff email from a worker to his boss, changing 'holler back' to a more professional 'please let me know your thoughts, best regards'.
The world's tech giants pushing these tools claim these AI functions will save people valuable time. A 2024 study of 1,300 Microsoft 365 users found those using Copilot saved 11 minutes per day – or the equivalent of a whole week each year.
In one case study, Google claimed that staff at one customer, billionaire Mark Cuban's online pharmacy business Cost Plus Drugs, were saving up to five hours per week using Gmail's prompts.
Not all users are convinced, however. On Microsoft's community forum, users have raged against Co-pilot, with one calling it a 'flop in AI productivity'. On Reddit, another says it 'struggles with even the simplest tasks'.
Brain rot
The Telegraph's own use of Gemini in Gmail had mixed results.
Often, the bot would seek to fill out snappy emails with verbiage and replace plain speech with jargon. Asking Gemini to add a calendar reminder for a call at 8am Thursday, based on an email chain, inexplicably resulted in a diary note for 11am Friday.
However, other tasks, such as digging out useful information buried in old emails, yielded useful results and reminders.
All instances of Gemini in Gmail warn users that the answers it gives could be incorrect. 'Double-check it,' its built-in chatbot warns.
Apple's addition of Apple Intelligence to its Mail app has been met with some alarming reviews from users. Several have reported that the AI falls for obvious phishing or spam emails, sometimes labelling them as 'priority'.
Jeff Hancock, a professor of communication at Stanford University who has been advising companies on their use of AI, says the reaction to AI tools among workers has been mixed. Bosses are eager for their workers to use them to boost their productivity, but many workers find the technology frustrating.
'There have been these big investments in AI products. [Companies] have been handing it to employees without any sense of what they should be doing with it,' he says.
'We are not seeing as much uptake as we might have thought.'
Many workers worry the AI bots do not sound authentic, or create extra work as they are forced to triple-check drafts.
What's more, writing experts question whether the tools are making people more efficient or their emails more meaningful.
Brodsky, of the University of Texas at Austin, says AI can be useful for brainstorming for those who struggle with writers' block or helping to edit 'low-stakes' emails.
But for many workers, realising a colleague is trying to speak to them with AI 'signals to me that the person does not care', he adds.
There are also concerns that outsourcing writing to an AI bot could lead to a drop in critical thinking skills. A study of the use of AI by 319 knowledge workers from Microsoft and Carnegie Mellon, published in February, found that 'confidence in AI is associated with reduced critical thinking effort'.
The report added that reliance on automation could leave our brains 'atrophied and unprepared'.
John Warner, author of More Than Words: How to Think About Writing in the Age of AI, says: 'We may be engaged in an active de-skilling of ourselves.'
If an email requires a well-thought-out answer, turning to an AI chatbot to draft it is probably a mistake, he says. 'Writing is thinking, and if thinking is part of what a response should do, then automating it is probably a mistake.'
A Google spokesman said: 'Our goal with integrating AI features, including Gemini, into Google Workspace is to help users save time, communicate more effectively, and focus on what's most important.
'We're excited about the potential of generative AI to assist with tasks like drafting emails, summarising long threads, prioritising messages and suggesting contextual replies.'
Google also pointed to a Harris Poll that found 82pc of US knowledge workers aged between 22 and 39 were using AI at work, and 70pc had used it for tasks such as drafting emails.
Apple and Microsoft were contacted for comment.