Kan AI befria oss från mejlkorgens tyranni?

New tools automating inbox management could ease a loathed productivity burden. But do they work?

Maani Safa used to spend hours every Friday evening catching up on hundreds of unread emails from the previous week.

But now the digital marketing chief executive has slashed that time to less than 30 minutes, thanks, he says, to an AI-powered workflow.

First, a customised ChatGPT bot scans his inbox to find and summarise important messages, suggesting a draft reply for each. Safa then uses those suggestions to dictate responses to another AI tool, Whispr Flow, which transforms them into neatly formatted, tonally appropriate emails.

“It’s genuinely transformative,” Safa says of the system, which he designed himself. “[It has] become an extension of me, not something that just auto-generates garbage.”

Ubiquitous, time-consuming and generally unloved, email is fast becoming a test bed for AI’s much-touted productivity-boosting power. In the past year, dozens of products have launched that promise to streamline inbox management, and in doing so reclaim hours of time now sucked into dull and often fruitless administrative activity.

In January, Google rolled out its Gemini AI to all of Gmail’s multi-billion-strong user base, calling it a “personal, proactive inbox assistant to help you manage your life”. It can summarise long email chains, polish or draft messages and create to-do lists from emails, reminding users about tasks or invoices. Microsoft has also integrated its Copilot assistant with Outlook, helping users prioritise their inbox and schedule appointments, among other features.

Colette Stallbaumer, general manager of Copilot, says AI can free workers from the “drudgery” of administrative tasks. A Microsoft study last year found the average employee received 117 emails and 153 Teams messages a day, while EU data shows the majority of university-educated workers spend at least half their day on digital communications.

“Knowledge work has reached peak inefficiency,” Stallbaumer says. As well as lengthening office hours to an “infinite workday”, this deluge means the day feels fragmented and chaotic, Microsoft says. The average time between interruptions by a meeting, email or message during core work hours was just two minutes. “The volume and fragmentation of digital communication is just leaving people overwhelmed . . . the opportunity that we have is for AI to really help with that.”

OpenAI and Anthropic have both introduced integrations with email providers that allow chatbots to answer users’ questions such as “when did my manager last email me?” or “what are my most urgent unread messages?” They also offer AI “agents” that can take control of folders within the user’s computer or their internet browser to carry out a wide range of tasks, including email.

Meanwhile a wave of start-ups are creating AI-powered email productivity tools. At least 85 companies focused on AI email assistants — including general tools, such as new products Superhuman or Read AI, and those focused on specific fields, such as sales — have attracted total funding of $581mn to date, according to Dealroom.

Among the largest is Fyxer, a London-based start-up which has raised some $40mn, and grew out of a founder’s previous business which linked people with human executive assistants.

CEO Richard Hollingsworth says this has given the company a close understanding of what professionals need from an email assistant. “One of the reasons we haven’t seen a product really take hold . . . is because folks who are building these tools are building them for other tech lovers,” he says. Instead, the company wants to target traditional industries that are “really dominated by email”. Estate agent Knight Frank was an early customer.

Katie Cameron, an independent estate agent in Chester, says Fyxer has been invaluable in helping her juggle nine inboxes across the three property businesses she runs, automatically slotting emails into categories like “FYI” and “Awaiting reply” and taking knowledge from previous conversations into account to suggest replies, create calendar invites and prioritise her inbox.

“It learns as it goes, which is cool — when I first started using it, it would always say ‘Cheers, Kate’, now it says ‘Thanks, Kate’,” she says. “Sometimes I am guilty of letting it think for me but it saves me hours a week.”

Natalie Mackenzie, chief executive of BIS Services, which supports people with brain injuries or disorders, is also a fan of the technology and has encouraged her team to adopt Microsoft’s Copilot to manage their busy inboxes — checking grammar, tone and automatically copying in people based on keywords.

Its ability to polish language helps junior staff feel confident about the tone of messages, but she tells her team not to trust AI summaries: an inaccurate overview of an incident involving a client, for example, could be dangerous. “Being in a very human industry where it’s very much about empathy, we can’t rely on it too much.”

Others are reluctant to trust the tech at all. Mark Tehrani, a cyber security start-up founder and academic, abandoned his experiment with giving ChatGPT’s “agent mode” access to his inbox after it deleted an important message by mistake. OpenAI acknowledged that its ChatGPT agent was in its early stages and “can still make mistakes”.

Business email writing coach Kim Arnold warns that recipients can also feel “disrespected” by obviously artificial messages. “We may be able to send out an email in half the time, but . . . that might be creating problems if the quality isn’t there,” she says.

Already, the ease of mass producing emails is making many people’s inboxes much harder to deal with. “Frictionless, fast, automated email has actually existed for a long time before AI, but we call it spam.”

Still, for many in the tech industry, email is simply a stepping stone to the real goal: autonomous AI personal assistants. Will McKeon-White, a technology analyst at Forrester Research, a US tech-focused research group, says this requires AI to be integrated fully into entire computer systems rather than specific apps.

“It needs the ability to actually execute on whatever it is that you’re asking . . . whether that’s create a calendar event, a reminder, a note, look up something on the web, go retrieve a song or help somebody understand what is in their email,” he says. “Humans usually do not understand [technical] limitations . . . so you want it to have the ability to just do more.”

What is less clear, he says, is if there is demand for such agents. Many people are wary of giving AI access to everything on their device. Still more are cautious about handing over something as personal as communication to a bot.

Even Copilot’s Stallbaumer says technology should augment emails rather than fully taking over. And while Mackenzie is keen for her team to use AI, she does not use it to manage her own inbox. “My PA’s for that,” she says. “I still like the human touch.”

©The Financial Times Limited 2026. All Rights Reserved. FT and Financial Times are trademarks of the Financial Times Ltd. Not to be redistributed, copied or modified in any way.