The AI Workplace Paradox: What Becomes of the Office of the Future?
Updated: 20 hours ago
AI is impacting on the workforce, but are organisations about to discover that the relationship between people, productivity and the workplace is not as linear as they might believe?
1. The great AI workforce experiment
Across the corporate world, organisations are cutting jobs, restructuring, moving work offshore and investing heavily in AI. A compilation of just 19 major companies caught up in the 2026 restructuring wave accounts for approximately 100,000 to 110,000 jobs affected.
Some jobs are being directly replaced through automation, while others are disappearing as organisations redirect investment towards AI, restructure their operations or pursue cheaper ways of delivering work. Whatever the mechanism, AI is increasingly influencing decisions about the future workforce.
And there's something enormously attractive to organisations about replacing human labour. Humans are expensive. They have expectations, need managing, take holidays, get sick, disagree with management, demand pay rises and occasionally leave for a competitor.
AI, by comparison, promises something close to the perfect employee. Available around the clock, endlessly scalable, increasingly capable and, potentially, considerably cheaper. It's easy to see the attraction, although treating ordinary human needs as inefficiencies to eliminate raises uncomfortable ethical questions.
For organisations contemplating the workplace implications, the obvious conclusion seems straightforward. Fewer employees mean fewer desks, fewer private offices, fewer meeting rooms and less demand for all the other work settings that accommodate employees. Less office space means lower property costs.
Except I'm not convinced it's going to be anything like that simple.
COVID taught us how unpredictable major disruptions can be. Few would have anticipated that a global pandemic would send second-hand car prices soaring, disrupt supply chains, empty CBDs, accelerate online shopping and fundamentally change where millions of people worked. Some consequences were obvious. Others seemed completely unrelated to the original event.
And before we've fully resolved the workplace consequences of COVID and hybrid working, AI is introducing a potentially even bigger disruption. This time, we're not simply changing where people work. We're changing what people do, how they work, who or what they work with, and potentially whether they're needed at all.
2. Is AI the golden egg of the next industrial revolution?
The ambition behind AI isn't simply to automate individual tasks. It's increasingly about automating enough of a job that the human performing it is no longer required.
But a job generally involves a combination of doing and thinking.
AI is already highly capable at producing documents, analysing data, writing software, researching subjects and completing repetitive processes. The thinking component, particularly contextual judgement, complex decision-making and dealing with unusual situations, has traditionally been harder to automate reliably.
The question is how long that distinction will remain relevant.
It's tempting to argue that humans will always be required to check AI's work, but that feels like a rather short-term defence of our employment prospects. AI agents are already being used to review the outputs of other agents, and multi-agent systems are becoming increasingly sophisticated. Although they can still make mistakes and reinforce each other's errors, today's limitations shouldn't automatically be treated as tomorrow's constraints.
Which makes the apparent enthusiasm for removing junior employees particularly interesting. Why does the redundancy conversation so rarely begin in the executive suite?
It's not difficult to imagine AI undertaking significant elements of strategy development, financial forecasting, resource allocation, performance reporting and management decision-making. In fact, experimental AI executive systems are already being developed.
One example is Open Executive, an open-source project from Sente Labs designed to automate elements of C-suite functions, including strategic analysis and operational decision-making. There's an interesting backstory. Open Executive was reportedly developed by people who had lost their jobs during an AI transformation. Their response was to develop technology capable of automating executive functions, turning the same logic of replacing humans with AI towards the C-suite. But the people behind Open Executive are now running an AI business themselves. Could their own technology eventually replace them too? The cycle of redundancy might simply keep turning.
Running a company involves legal accountability, leadership, relationships and consequences that software cannot simply assume today. But the same argument about automation gradually replacing the components of a job could apply to executives as much as anyone else.
Perhaps those leading the charge towards AI replacement should be careful what they wish for. Once the technology is considered capable of replacing employees, there's no particular reason the logic should stop at the boardroom door.
Governments rely heavily on employment-related taxation. If significant numbers of taxpayers are replaced by digital workers, what happens to that revenue?
A similar issue is playing out in the electric vehicle market. EV drivers generally don't pay fuel excise, creating a long-term challenge for a taxation system developed around petrol and diesel consumption.
Could governments eventually tax AI labour or reward organisations that retain and augment human employees rather than replace them?
The debate is already moving in that direction. In October 2026, independent Senator David Pocock proposed a 5% tax on revenue from large Australian data centres, costed by the Parliamentary Budget Office. It targets the infrastructure rather than the digital workforce itself, but it demonstrates how governments are beginning to reconsider who should pay for the economic consequences of AI.
The golden egg might not remain quite so golden once governments begin adjusting the rules.
3. What if the AI workforce costs more than the human workforce?
Imagine replacing a fully loaded $150,000 employee with AI, only to discover that the technology ultimately costs $150,000 or more to operate.
Enterprise software, computing, integration, cybersecurity, specialist support, governance, maintenance and ongoing development all carry costs. There may also be substantial expenditure on human supervision, particularly while AI systems remain unreliable in complex or sensitive applications.
McKinsey's August 2026 research found that in one banking customer-service application, human oversight accounted for 70% to 75% of an AI agent's variable running costs, while processing tokens represented just 20% to 25%.
But AI is becoming dramatically cheaper. Stanford's AI Index measured a more than 280-fold decline in the cost of GPT-3.5-equivalent performance between November 2022 and October 2024. Research published by Epoch AI in September 2026 found that the cost of achieving a given level of AI performance had continued falling by approximately 47% per quarter since 2023.
Cheaper AI processing doesn't necessarily mean lower overall infrastructure expenditure. Demand for computing capacity is growing, along with the land, electricity, cooling and water needed to support it. Australia has plenty of land, but suitable power and water infrastructure are another matter. Elon Musk is even exploring the possibility of putting AI computing infrastructure into orbit.
And not every AI investment will succeed. Gartner has predicted that more than 40% of agentic AI projects will be cancelled by the end of 2027 due to escalating costs, unclear business value or inadequate risk controls.
Replacing human labour with digital labour isn't automatically the same as reducing operating costs. AI might deliver enormous benefits through increased output, faster delivery or entirely new services, but that's a different business case from simply eliminating people to save money.
4. The AI conductor: More free time or more work?
Historically, much of knowledge work has been relatively sequential. You start a task, complete it and move on to the next. AI creates an opportunity to work very differently.
Imagine conducting an orchestra, except the musicians are ChatGPT, Claude, Copilot, Grok and an assortment of specialist AI agents.
You brief ChatGPT to undertake some research. While it works, you instruct Claude to produce a report. Meanwhile, Copilot is analysing a spreadsheet, and another agent is preparing a presentation.
But AI isn't simply a digital workforce waiting for instructions. It can also become an intellectual sparring partner, challenging assumptions, suggesting alternatives and helping people develop ideas they might never have reached independently. Its value isn't always in what it produces, but in how it can expand and support human thinking.
The introduction of increasingly autonomous agents, such as Anthropic's recently announced Claude Duet, takes this further. These systems are moving beyond responding to individual prompts towards undertaking activities, interacting with software and continuing to work while the human is occupied elsewhere.
But I can imagine two very different outcomes for the working day.
Scenario one: AI creates breathing space.
The technology undertakes labour-intensive activities, freeing employees to think, reflect, collaborate, make decisions or simply take a break. Perhaps workplaces will need more relaxed environments, comfortable informal settings and spaces that encourage creative thinking.
Scenario two: AI creates the hyper-productive employee.
Every time one AI starts processing, the employee launches another task. Then another. Then another, effectively “super-setting” their working day in an exhausting productivity drive.
Instead of reducing the workload, AI expands it. The human becomes a relentless conductor, moving between multiple agents, reviewing outputs, correcting errors and continuously feeding new activities into the system.
Interestingly, UC Berkeley research published in February 2026 observed employees using AI taking on more activities, working faster and extending their working hours. The technology created additional capacity, and people filled it with more work.
Reality will likely sit somewhere between these two extremes, with the balance varying between employees, roles and organisations.
These two scenarios could create very different workplace requirements. One may encourage more reflection, creativity and collaboration, while the other could require environments that support sustained concentration and intensive interaction with AI agents. An employee might need both types of settings throughout their working day. AI could make organisations considerably more productive while simultaneously making the working day more demanding for the humans who remain.
5. FTE + FTEA: When your workforce is no longer entirely human
For decades, organisations have measured their workforce using headcount and Full-Time Equivalent (FTE) figures. These underpin budgets, organisational structures, resource planning and office accommodation requirements.
But what happens when a team of ten employees manages fifty AI agents undertaking significant volumes of work? Is it still a ten-person team?
Perhaps we'll need another workforce measure: FTEA, or Full-Time Equivalent Agents.
An organisation might eventually measure its productive workforce as a combination of human FTE and digital FTEA. An agent doesn't automatically equal one employee, and in some circumstances its capacity could be considerably greater. An AI agent might operate 24 hours every day rather than the eight hours of a conventional working day.
Any meaningful FTEA measure would therefore need to reflect actual productive output rather than simply counting deployed agents. Nevertheless, it challenges how we currently think about workforce size.
A small business with five employees and a sophisticated digital workforce might outperform a traditional organisation employing fifty. What would that mean for the corporate hierarchy, management structures and the allocation of resources?
If organisations increasingly have a workforce comprising humans and AI agents, perhaps HR will become Human and Artificial Resources.
Who recruits the agents, measures their performance, upgrades their capabilities, manages access permissions and retires them when they become obsolete? Could HR departments eventually manage hundreds of digital workers alongside their human counterparts? Or will HR agents manage other agents, including those responsible for traditional human resources activities?
Imagine a future CV that focuses less on years of experience and more on the productive capabilities an individual brings to the organisation. Perhaps employees will have something resembling a computer-game character profile, accumulating specialist agents and capabilities that effectively level them up.
Two candidates might possess identical qualifications but dramatically different AI-enabled capabilities.
Of course, employers would need to resolve questions of cybersecurity, intellectual property, data protection and whether employees can bring their personal AI agents into corporate systems. But the concept challenges the traditional value of experience as a measure of capability.
Could the future employment question become less about what experience you have and more about what you and your AI workforce can accomplish?
6. The AI exoskeleton: Is the corporate ladder becoming obsolete?
During numerous pre-COVID Activity Based Working (ABW) transformations, I encountered an interesting pattern when introducing desk sharing into organisations where employees traditionally had assigned workstations.
Some of the strongest resistance didn't come from older, long-serving employees, as many people might expect. It came from ambitious graduates.
These younger employees understood the value of sitting near experienced colleagues, building relationships, overhearing conversations and learning through observation. They were competing to progress through their organisations and recognised that losing their guaranteed proximity to influential, knowledgeable colleagues could affect their career prospects.
They weren't necessarily resisting desk sharing. They were protecting their access to knowledge and opportunity.
Then COVID arrived, followed by widespread hybrid working. Even when graduates attended the office, there was no guarantee that the experienced colleagues they wanted to learn from would be there.
Research published in the Quarterly Journal of Economics in 2026 supports the importance of that proximity. Studying software engineers, researchers found that sitting near colleagues increased feedback and improved skill development, particularly among younger and less-experienced employees.
Now AI introduces another challenge. Stanford research released in August 2026 found that employment among 22 to 25-year-olds in highly AI-exposed occupations was approximately 19% below where it would have been had it followed employment trends among similarly aged workers in less-exposed occupations.
The researchers didn't find widespread economy-wide displacement, but the evidence suggests that reduced recruitment into some AI-exposed occupations is already affecting early-career opportunities.
If organisations stop recruiting junior employees, where will tomorrow's experienced leaders come from?
But perhaps we're approaching the problem through an outdated lens?
Traditionally, employees accumulate knowledge and experience over many years before reaching senior leadership positions. Yet young entrepreneurs have been bypassing the corporate ladder for decades. Some establish businesses straight out of university, while others use search funds to raise investment and acquire established companies, becoming CEOs alongside executives with twenty or thirty years more experience.
Now imagine what happens when AI dramatically expands the capabilities of those relatively inexperienced individuals.
Think of AI as an exoskeleton for the brain.
An inexperienced hiker might struggle to climb a mountain unaided, but an advanced physical exoskeleton could enable them to travel further, faster and with less effort.
A twenty-something might not possess the judgement or experience normally expected of a CEO. But give them access to sophisticated AI advisers, analysts, strategists and specialist agents, and perhaps their capability changes dramatically.
That doesn't guarantee good leadership, but it could compensate for gaps that traditionally required decades of experience to overcome.
Perhaps AI won't simply eliminate the bottom rungs of the corporate ladder. It might enable some people to leapfrog the ladder altogether.
And working backwards, what does that mean for education? If AI becomes an increasingly powerful professional exoskeleton, will traditional qualifications and business education need to change? Perhaps the future MBA will increasingly resemble an MBAi, where understanding how to orchestrate artificial intelligence becomes as important as traditional management knowledge.
Which brings us back to those ambitious graduates. AI might put an expert adviser on their laptop, removing some of the advantage of sitting beside an experienced colleague. But the relationships, judgement and opportunities gained from working alongside people may become more valuable, not less.
7. AI and the Future of the Workplace: Will Fewer People Mean Less Office Space?
If an organisation reduces its workforce by 20%, the obvious assumption is that it can reduce its workplace by a similar amount. In practice, the relationship between headcount and space has never been that straightforward. Hybrid working, desk sharing, attendance patterns and the physical limitations of buildings all influence how much accommodation an organisation actually needs.
AI and the future of the workplace aren't simply about how much office space organisations will need. What if the remaining employees need a fundamentally different type of workplace?
Consider the AI conductor we discussed earlier. If employees increasingly manage multiple AI agents, perhaps the conventional desk with two monitors will no longer be sufficient. They may need more immersive environments, multiple screens displaying the progress of different agents, sophisticated voice interfaces and the ability to interact with several digital workers simultaneously.
We could be heading towards something resembling Tony Stark's laboratory in Iron Man, with JARVIS and an assortment of AI assistants responding to instructions, presenting information and undertaking work in real time.
Voice interaction with AI is becoming increasingly common. Now imagine fifty employees in an open-plan office, each talking to several AI agents. Apart from the obvious noise and distraction, what happens when one person's AI overhears instructions intended for somebody else's? Or when two agents respond to the same instruction? Unless the technology can reliably identify individual voices and intended recipients, we could have a rather chaotic workplace.
And we're not simply talking about employees issuing occasional voice commands. People will increasingly hold extended conversations with AI, brainstorming ideas, challenging recommendations, asking questions and working through complex problems. What might previously have been a silent activity at a computer could become a continuous spoken dialogue. That's a fundamentally different acoustic challenge for workplace designers.
Headsets, directional microphones and speaker recognition will undoubtedly help, but acoustic privacy could become considerably more important. And it's not just about noise. Employees may be discussing commercially sensitive information with AI systems that colleagues or visitors shouldn't overhear.
After spending decades removing office partitions to encourage people to communicate with each other, we might find ourselves putting them back so people can communicate with machines.
The return of the cubicle? That would be quite an ironic outcome for an industry that has spent years advocating open-plan workplaces, Activity Based Working and collaboration.
But perhaps the AI-enabled workplace won't be a cubicle at all. It might be a sophisticated individual cockpit, an enclosed immersive room or a shared AI command centre where employees monitor and direct digital operations through interactive displays. Think less traditional office and more SpaceX mission control.
There is, however, an obvious alternative to building these expensive environments in corporate offices.
Working from home.
Why travel to the office to sit in an acoustically isolated room talking to AI agents when you could do exactly the same thing at home? Imagine a CEO's home office designed around a Stark Industries aesthetic, complete with immersive displays, voice-controlled systems and a digital team working around the clock.
For some knowledge workers, the home could become the ideal environment for conducting AI agents, while the corporate office becomes somewhere they visit primarily to interact with other humans.
The costs of these increasingly sophisticated home workplaces may, as with home working today, fall largely on employees, sometimes supported by employer contributions. The organisation potentially reduces its property footprint, but the physical workplace and its associated costs haven't disappeared. They've simply shifted location.
And then there's the metaverse.
Remember when we were all going to abandon physical offices and meet as avatars in virtual workplaces? The vision generated enormous excitement, but widespread adoption never materialised.
Perhaps the metaverse arrived too early, or perhaps we misunderstood its most useful application.
Instead of using virtual reality to replicate meetings between humans, what if the metaverse becomes the place where humans meet their AI workforce?
Imagine managing ten AI agents on a project. Rather than switching between ten chat windows, you enter a virtual meeting room where each agent has a digital identity and appears as a member of your team. You brief them collectively, review their progress, hear their recommendations and watch them debate different approaches.
We may even discover that the metaverse is more useful for managing artificial colleagues than meeting human ones.
None of this necessarily requires a headset or an elaborate virtual world. Similar interactions could take place through desktop interfaces, augmented reality or technologies we haven't yet developed. But giving AI agents a visible identity and a shared environment could make managing increasingly complex digital workforces more intuitive.
An employee could spend an entire day collaborating without speaking to a single human being. They might be engaged in constant discussion, problem-solving and creative development with AI, yet physically occupy an individual workspace all day. Collaboration is no longer necessarily human-to-human, and workplace strategies may need to accommodate both.
Of course, this creates another contradiction. If AI increasingly undertakes the individual production work, perhaps the physical office becomes more important as somewhere humans gather to develop relationships, exchange knowledge, collaborate and maintain organisational culture. Activities that arguably become more valuable precisely because so much other work can be undertaken independently with AI.
The implications extend beyond furniture and floorplans. Increasing reliance on AI agents could also change workplace requirements for technology, connectivity, cybersecurity and access controls.
The paradox is that AI could substantially reduce the number of humans employed by an organisation without producing an equivalent reduction in the space, technology or investment required to accommodate those who remain.
The workplace of the future might be smaller. It might be more expensive per employee. It might be more immersive, more dispersed, more enclosed or increasingly virtual.
It certainly doesn't have to look like today's office, only with fewer desks.
8. What if our assumptions about AI are wrong?
Workplace Strategy has traditionally involved understanding an organisation's workforce, projected growth, ways of working, attendance patterns and accommodation requirements. Those measures remain essential, but AI introduces a new level of uncertainty into the relationship between workforce demand and workplace supply.
Scenario one: Imagine an organisation that decides to embrace AI aggressively. It eliminates 25% of its workforce, reduces its office accommodation, surrenders floors and celebrates the projected savings in salaries and property costs.
Twelve months later, the technology hasn't delivered what was promised. Productivity is declining, customer satisfaction is deteriorating, projects are falling behind schedule and operational costs are higher than anticipated. The remaining employees are struggling to manage the workload, while the AI systems intended to replace their former colleagues are proving less capable or more expensive than expected.
The organisation now needs to reverse course. It must recruit employees, rebuild lost capability and find somewhere to accommodate them.
What initially appeared to be an intelligent efficiency program has become an extraordinarily expensive exercise in correcting a premature decision.
Commonwealth Bank provides an interesting real-world example. In August 2025, it reversed plans to eliminate 45 customer-service roles after acknowledging that its AI voice bot hadn't made those positions redundant. Yet the experience didn't stop the bank pursuing hundreds of further job cuts in 2026, attributed to a combination of restructuring, automation and changing business requirements. Getting one AI workforce decision wrong clearly doesn't mean an organisation will abandon its broader transformation ambitions.
Scenario two: The opposite outcome is also possible. An organisation might retain its current accommodation, expecting workforce growth, only to discover that AI agents can absorb much of the additional workload without recruiting more people. Its office portfolio becomes progressively underutilised.
Scenario three: Or perhaps the organisation ends up with roughly the right amount of space, but entirely the wrong type of workplace. It planned for employees undertaking largely individual computer-based activities, only to discover that their roles now involve directing AI agents, collaborating with colleagues and making complex decisions in very different environments.
These scenarios expose the danger of treating workplace strategy as a property-cost response to an AI-driven headcount target. The starting point is the organisation's future Target Operating Model (TOM): how the business intends to operate, what work AI will undertake, what people will do and what capabilities the organisation will need. That operating model determines workforce requirements, which in turn drive the workplace strategy. The workplace is there to enable the TOM, not dictate it.
AI capabilities are changing rapidly, but leases, fit-outs and property commitments can last for years. Organisations need to test the assumptions behind their future operating model and consider alternative outcomes before making costly accommodation decisions that may be difficult to reverse.
The organisations that get this right won't necessarily be those that eliminate the most employees or surrender the most office space. They'll be those that understand how AI is changing their business before deciding what workplace that business needs.
