It’s possible that AI could take 90%+ of jobs, but is it probable?
I was listening to Steven Bartlett's podcast recently (Diary of a CEO), and there's a common view emerging from guests like Mo Gawdat (ex-Google executive) and Dr. Roman Yampolskiy (Associate Professor in AI) making claims that AI will wipe out more than 90% of jobs in the next 5 to 10 years. This has become a common discussion topic with clients, colleagues, and friends, a hot topic for debate.
My view? It makes for a great headline and clickbait, but as much as this could be possible, I struggle to see being probable! Here are four reasons why I believe the doomsday narrative doesn't hold up for me and why the future of work will look more like evolution than extinction.
Most jobs aren't checklists and you will always need a human in the loop
Very few roles consist entirely of simple, defined tasks. Real jobs are dynamic and involve coordination, iteration, judgment and ultimately someone being accountable for the outcome.
- AI might draft code or answer queries, but who takes responsibility when bugs crash production or when a customer escalates in frustration? You cannot fire or sue the AI.
- Who takes responsibility when the task or code output is more complex than a simple prompt can manage, or when it needs out-of-the-box thinking to solve? You need a human in the loop.
Look at what has happened in the past year. Verizon rolled out a Google-powered AI assistant in its service centres, and it worked because it supported people rather than replaced them. Agents used the AI to surface information faster, cut call times, and even boost sales by nearly 40%. UK and USA based retailers are taking a similar path, adopting AI for call routing and post-purchase updates to reduce wait times and make agents more effective. In both cases, staff stayed in place and were shifted into higher-value work.
But not all experiments have gone smoothly. Commonwealth Bank of Australia (CBA) tried to cut dozens of call centre jobs by handing "simple" queries to an AI voice bot. Instead of lowering workloads, call volumes rose as frustrated customers escalated more issues. Overtime increased, managers had to jump back onto the phones, and eventually the bank reversed its decision, offering roles back to staff. It was a clear reminder that customer service is dynamic, often emotional, and not something you can hand off wholesale to a machine.
The same contrast shows up in software development. AI coding assistants can speed up routine tasks like documentation, boilerplate, and refactoring, sometimes by 20% to 50%. Junior developers see significant gains. But higher-order work like architecture, integration, and reviews still requires human oversight. You can argue to say that even Juniors could be replaced, but then who will become your intermediate to senior engineers in years to come. In fact, some recent field studies showed developers became slower when over-relying on AI without proper processes.
The pattern is clear: AI succeeds when it frees humans to do higher-order tasks. It fails when companies underestimate the dynamic and judgment-based nature of real jobs and try to remove people entirely.
Good paper to read on work done by the NBER is “Tracking firm use of AI in real-time: A snapshot from the business trends and outlook survey”, by Bonney et al; https://www.nber.org/papers/w32319
The Economy Can't Afford a Job Apocalypse
Let's imagine the 90% scenario. If unemployment ever hit even 20% to 30%, the system would buckle. Customers without income stop buying. Taxpayers disappear while demand for welfare surges. Governments cannot sustain that imbalance for long.
This is why most economists believe that large-scale automation forces the economy to rebalance. New industries, roles, and services spring up. The World Economic Forum estimates that while 83 million jobs could be displaced by 2027, around 69 million new roles will also emerge. Goldman Sachs projects AI could increase global GDP by 7%.
In short, remove too many people and the system collapses. Keep humans in play and the economy adapts. Plus being a company that uses Human enhanced with AI to deliver exceptional outputs will be more competitive than AI only companies (low cost vs quality) – would you go to an AI doctor or an Doctor that uses AI?
Good paper to read is one by Nobel laureate Daron Acemoglu “What do we know about the economics of AI?”, https://news.mit.edu/2024/what-do-we-know-about-economics-ai-1206
Change Always Moves Slower Than the Hype
Even when the technology is ready, adoption is never instant. Companies need to re-platform systems, clean up data, upgrade processes, train people, and align culture. That's change management, and it's slow and messy. Most organizations are still struggling with the basics, which means the full benefits of AI take years, not months, to materialize.
Economists call this the Productivity J-Curve. Big investments in new tech often show up as costs first, and only later as productivity gains once the right foundations are in place. It was the same with electricity, computers, and the internet.
We've seen the hype before. Remember when online shopping was meant to kill malls in five to ten years? Two decades later, e-commerce is huge, but malls are still here. Or take blockchain, which was going to replace all banks in a similar timeframe. The promise was revolutionary, but adoption has been incremental at best, with banks still firmly in place. I recall people saying six years ago that 80% of all new cars would be electric and self-driving by 2025. The list goes back as far as time.
AI is no different. Headlines race ahead, but systems, processes, regulators, and people take much longer to catch up. That's why sweeping predictions of job wipeouts almost always overestimate how fast the future arrives.
Good paper to read on this is “The Rise of Industrial AI in America: Microfoundations of the Productivity J-curve(s)” by McElheran et al.
AI Expands Human Capacity
Everyone's job most likely requires significantly more hours than are available to do their job effectively. If I wrote down all the activities I should do in my role to deliver the best/highest outcomes, I would need two to five times more working hours in the week. For example, more time on sales, communication, and writing LinkedIn thought pieces could easily be more than double my current hours alone.
So even if AI automated half my tasks, I would not be redundant. I would finally have capacity to tackle the higher-order work that usually gets squeezed out: strategy, problem-solving, innovation, building relationships, etc.; or capacity to add more commercial outputs for the business. For most professionals, the issue is not a lack of tasks; it's too many. AI simply helps us catch up.
That's exactly what shows up in the productivity studies: time saved on grunt work turns into more quality time for design, decisions, and customers. The real story is not "job killer" but "capacity builder." AI clears the clutter so people can lean into the higher-order work that actually moves the business forward.
Good research to read is PWCs press release “AI linked to a fourfold increase in productivity growth and 56% wage premium, while jobs grow even in the most easily automated roles”.
Acknowledging Where My Argument Might Fall Short: Speed, Scope, and Breakthrough AI
I acknowledge the counter arguments that I might be underestimating how fast AI is advancing or cherry-picking examples that support my case. Fair points worth addressing.
Yes, we're seeing remarkable breakthroughs in generative AI, and the prospect of artificial general intelligence (AGI) could theoretically accelerate displacement beyond historical patterns. Transportation, retail, and manufacturing are already seeing significant automation: autonomous vehicles are being tested at scale, self-checkout is expanding rapidly and factories have been automating for decades. Governments could establish sovereign funds and invest heavily in state-driven industries that harmoniously integrate AI, robotics, and additive manufacturing, creating a society where we can have anything we want without worrying about work (this being Mo Gawdat's key point in the podcast).
These are valid points, and there's a probability that I'll look back at this article wondering how wrong I got it. The questions for me still remain: ‘how quickly, to what extent and how likely is this to happen?’.
I am keen to hear your thought on the likelihood of the above counter points – please share your comments below.
The Real Takeaway
AI will transform work, no question. Some jobs will indeed disappear, but more likely, the majority will evolve, and entirely new ones will emerge. But I deeply believe it's not the doomsday scenario some headlines suggest. Time will tell.
At the very least, in the short to medium term, the real winners will be the businesses and people who treat AI as a partner, using it not to replace humans, but to help them do higher-value and more meaningful work.
So the question isn't "Will AI take your job?" It's "How will you use AI to do your job better and not make yourself obsolete?"
What are your thoughts?
