I hope this article finds you well.

Did that make you cringe, ever so slightly? In the decades since the very first email was sent in 1971, the technology has become the quiet infrastructure of white-collar work.

Email came with the promise of efficiency, clarity and less friction in organisational communication. Instead, for many, it has morphed into something else: always there, near impossible to escape and sometimes simply overwhelming.

Right now, something is shifting again. The rise of generative artificial intelligence (AI) technologies, such as ChatGPT and Microsoft Copilot, is increasingly allowing people to offload the repetitive routines of tending one’s inbox – drafting, summarising and replying.

My colleagues in the ARC Centre of Excellence for Automated Decision Making & Society found 45.6% of Australians have recently used a generative AI tool, 82.6% of those using it for text generation. A healthy chunk of that use likely includes email.

So, what happens if we end up fully automating one of the staples of the white-collar daily grind? Will AI technologies reduce some of the friction, or generate new forms of it? Dare I ask – are we actually about to get more email?

Why the printer isn’t dead yet

Soon after the advent of email, some voices in the business world heralded the coming end of paper use in the office. That didn’t happen. If you work in an office today, there’s a good chance you still have a printer.

In their 2001 book, The Myth of the Paperless Office, Abigail Sellen and Richard Harper show how digital tools rarely eliminate older forms of work. Instead, they reshape them.

Sellen and Harper show how paper use didn’t disappear with the rise of email and other digital communication tools; in many cases, it intensified. The takeaway isn’t that offices failed to modernise, but rather that work reorganised around what these new tools could do.

In this case, paper persisted not only out of habit, but because of what it affords: it is easy to annotate, spread out, carry and view at a glance. This was all too clunky (or impossible) to perform via the digital alternatives.

At the same time, email and digitisation dramatically lowered the cost of producing and distributing communication. It was far easier to send more messages, to more people, more often.

Circling back to today

Will AI be different? If early signs are anything to go by, the answer is: not in the way we might hope.

Like earlier waves of workplace technology, AI is less likely to replace existing communication practices than to intensify them – but at least it might come with better grammar and a suspiciously upbeat tone.

Some new AI tools offer to manage your inbox entirely, feeding into broader privacy concerns about the technology.

At this moment, what a lot of these products seem to offer is not an escape from email, but a smoothing of its rough edges. Workers are using AI to soften otherwise blunt requests, modify their tone or expand what might otherwise be considered too brief a response.

Rather than removing the need to communicate, these tools offer pathways to make a delicate performance easier.

What email is actually for

Email, like many forms of communication, is as much about maintaining everyday relationships as it is about the transfer of information.

At work, it’s often about signalling competence, responsiveness, collegiality and authority. “Just looping someone in” or “circling back” are all part of our absurd office vocabulary, a shared dialect that helps us navigate hierarchy, soften demands and keep things moving – all without saying what we really think.

If AI lowers the effort required to produce these signals, it won’t necessarily reduce their importance, but it could unsettle things in rather odd ways.

If more people use AI to draft emails they don’t particularly want to write, we end up with a game of bureaucratic “mime”: everyone performing sincerity and quietly outsourcing it, and no one entirely sure how much of their inbox was actually written by a human.

The labour of email was never just about crafting sentences. It’s always been the scanning, the sorting and the deciding. AI doesn’t remove this burden. If anything, it amplifies it.

When everything arrives polished, everything looks important. That points to a deeper question for the future of work: if AI can perform responsiveness, why are we generating so many situations that still require it?

Looking forward

What would a workplace look like if email wasn’t the default solution to every coordination problem? Perhaps fewer performative check-ins, “just touching base”, “looping you in” or “following up on the below”. More clearer expectations about what actually requires a response, and what doesn’t.

Email, like paper, is likely to persist for good reasons. It is simple, flexible and universal. It allows things to be deferred, revisited, forwarded and quietly ignored.

But if AI is going to change any of this, my hope is that it makes visible how much of this is ritual, how much is habit, and how much has long been unnecessary.

And if the machines are happy to keep saying “hope this finds you well” to each other, we might finally have permission to stop.

Daniel Angus receives funding from the Australian Research Council through Linkage Project LP190101051 'Young Australians and the Promotion of Alcohol on Social Media'. He is a Chief Investigator with the ARC Centre of Excellence for Automated Decision Making & Society.

A humanoid robot recently made headlines around the world for running a half-marathon and beating the human world record. Around the same time, an AI-powered robot defeated an elite human player in table tennis. What the robot lacked in experience, it made up for by reacting faster and more consistently than any person could.

These moments feel like milestones. Finally, it seems machines are stepping into one of the most human arenas – sports.

But while it is tempting to frame this as robots versus humans, sport robotics isn’t really about competition. It’s about how machines can learn to move, react and interact in dynamic, unpredictable environments – and what that means for human performance.

How do you train a robot to play sport?

Training a robot to play sport is fundamentally different from training a human athlete.

People learn through practice, coaching and experience, constantly adjusting to changing conditions. In sport science, this is often described as a tight coupling between perception and action. That is, seeing, deciding, and moving in one continuous loop.

Robots, by contrast, are trained using a combination of simulation, data and control algorithms. Engineers build detailed virtual environments where robots can “practice” millions of times. They learn how to track objects, predict motion and coordinate their bodies. Sometimes, motion analysis techniques are used to track athletes doing the specific movements the robot needs to emulate.

For fast-paced sports such as table tennis, the challenge is extreme. A robot must detect the ball, predict its trajectory and execute a precise movement within fractions of a second. This requires close integration between computer vision, machine learning and real-time control.

One of the biggest advances in recent years has been the ability to train robots in simulation and then transfer those skills into the real world – a process known as “sim-to-real”. Combined with rapid improvements in sensors and computing, this has dramatically accelerated progress.

We’ve seen similar developments in robot basketball and robot soccer, where systems have evolved from simply locating the ball to coordinating as teams, making tactical decisions and adapting to opponents.

Beyond entertainment

While robot athletes make for compelling demonstrations, their greatest impact will likely be behind the scenes where they can be used to train human athletes.

One of the central challenges in sport is designing effective practice. Athletes need repetition to build skill. But they also need variability to reflect real competition. Too much repetition becomes predictable; too much variability becomes chaotic.

Robotics offers a potential way to balance both.

A robotic training partner can deliver highly repeatable actions at elite intensity, while also introducing carefully controlled variation. For example, a robotic tennis server could replicate the motion of a world-class player while systematically varying ball speed, flight and placement.

From a sport science perspective, this creates what is known as a “representative learning environment”. The key benefit is it replicates the key perceptual and decision-making demands of elite competition, which is difficult for coaches to recreate in the training environment.

In our work, we’ve been exploring how robotics could support sports such as tennis, cricket and the football codes. The goal is to combine realism, repeatability, variability, and data to enhance skill development and link technique to outcomes.

Robots may also help manage training load. They can reduce the physical demands on coaches and training partners while still exposing athletes to high-quality game-like scenarios.

Beyond performance, there are opportunities for fan engagement. Interactive robots at live events or demonstrations of elite skills could offer new ways for audiences to experience sport.

Will robots ever be ‘great’?

Over the next decade, robots will likely become more agile, more robust and better able to operate in complex environments. Tasks that robots currently find difficult, such as running on uneven terrain and catching or throwing balls, will become increasingly achievable.

But even as robots improve, there are important limits.

Sporting greatness is not just about executing movements perfectly. It involves creativity, decision-making under pressure, and the ability to adapt in ways shaped by experience, emotion and context.

From a sport science perspective, elite performance emerges from the interaction between the athlete, the task and the environment. Robots can be engineered to perform specific tasks extremely well, but they do not experience this interaction in the same embodied, meaningful way.

This means robots may surpass humans in tightly defined challenges – such as bowling a cricket ball with perfect consistency – but they are unlikely to achieve greatness in the holistic human sense.

A new role for robots in sport

Rather than replacing athletes, robots are more likely to become part of the sporting ecosystem.

In the same way that video analysis and wearable sensors have transformed training, robotics offers a new tool for coaches and sport scientists. It enables practice environments that can be precisely controlled, repeated, and adapted to individual needs.

The real opportunity is not to build robot champions, but to better understand human performance, and help athletes reach higher levels.

Jonathan Roberts receives funding from the Australian Research Council.

Marc Portus does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.