Is AI About to Take My Job?

Kieren Sharma
Apr 13, 2025
5 min read

In our latest episode, we tackled the question that’s on everyone’s minds: is artificial intelligence about to take our jobs? This is a topic that came up frequently during our recent interviews with the public, highlighting the widespread anxieties surrounding AI’s growing capabilities.

Listen Now

Why Now?

The timing of this episode is crucial. As we move into 2025, we’re witnessing major AI companies rolling out new “agentic” AI models—systems that can plan, reason, and act without constant human oversight. Unlike standard AI (think facial recognition or old-school search engines) that tackles only single-step problems, agentic AI takes initiative: it not only understands and describes tasks but also decides on its next steps. It’s a progression that some believe is nudging us ever closer to Artificial General Intelligence (AGI), potentially more quickly than anyone previously predicted. To dive deeper into what AGI really means, be sure to check out our short stuff episode on the topic.

A Look Back: Lessons from Technological Unemployment

To shed light on the possible future impact of AI, we first turned our attention to history. We traced major waves of technological change—from the printing press and mechanised weaving to the steam engine, assembly lines, and the digital revolution—to see why these moments caused massive job shifts and also sparked new industries.

Decades ago, a Russian-American economist made a striking analogy: once steam engines rendered horses largely obsolete, it would only be a matter of time before technology outpaced human labour in a similar way.

“The human worker will go the way of the horse." - Wassily Leontief

Even the term “technological unemployment” was popularised by John Maynard Keynes nearly a century ago. History shows that people’s concerns about being “replaced by machines” have been around for generations.

Two Major Forces

Throughout history, technological advancements have brought about two significant forces that influence the world of work:

Substituting force
New technologies often displace certain human tasks and skills. For instance, the printing press replaced scribes, and mechanised textile machinery greatly reduced the need for hand-weavers.
Complementing force
Over the longer term, however, technology has repeatedly created new roles and resulted in higher productivity. Cheaper textiles, for example, gave rise to booming consumer demand, factory expansions, and new jobs in related sectors like machine maintenance and product distribution.

Yet, as we discussed, the benefits are not always evenly shared. Economists warn of a “hollowing out” effect: many middle-skill jobs vanish while high and low-skill roles grow, widening the gap and challenging the workforce to keep up.

Agentic AI: The Game Changer

We then focused specifically on agentic AI, differentiating it from standard AI. Standard AI operates as a single-step problem solver (e.g., facial recognition, song recommendation) where humans remain in the driver's seat. In contrast, agentic AI is a multi-step problem solver with a feedback loop, capable of acting without constant human intervention. Examples range from the humble Roomba to sophisticated AI chess players and new search engines.

The development of agentic Large Language Models (LLMs), capable of planning, browsing the web, and reasoning, marks a significant step. Early forms of this, like Auto-GPT and BabyAGI, were even developed by the open-source community in 2023. The release of “deep research" modes by companies like Google (with Gemini in December 2024) and OpenAI (with ChatGPT in February 2025) showcased the power of these models to conduct complex research independently, saving significant time and effort. This ability of agentic AI to act without human oversight is a core concept.

The potential of agentic AI has led to serious discussions, even prompting an open letter in 2023 calling for a pause in large-scale AI development. Furthermore, prominent figures like Mark Zuckerberg predict that by 2025, AI could start replacing mid-level programmers and software engineers.

Why AI is Different

Looking back at the printing press or the steam engine, humans ultimately found new tasks that only they could perform—or, at least, tasks the technology of the time couldn’t handle. AI, however, is evolving to do more and more tasks once considered uniquely human. That includes the new jobs created by the complementing force; in theory, AI could also learn these roles just as quickly.

Moreover, agentic AI models can learn by doing, fix their own errors, and run continuously without needing breaks. This velocity of improvement sets them apart. We also discussed Moravec’s paradox, which reveals that AI struggles with tasks humans find simple (such as motor skills, basic perception, and face-to-face interaction) while excelling at what humans consider incredibly difficult (complex calculations, large-scale data analysis). This mismatch makes job displacement far from intuitive.

Another element is energy cost. Unlike prior machines that quickly replaced labour for economic reasons, current large language models demand huge computing resources and energy. As a fun side note, OpenAI itself doesn’t expect to see positive cash flow until 2029, hinting that high operating costs may slow full-scale AI takeover...for now.

Is Your Job at Risk? A Rule of Thumb

We offered a simple framework: ask yourself three questions to gauge how vulnerable any given role might be to AI.

Is it easy to define the goal of the task?
If yes, the AI knows exactly what “success” looks like.
Is it straightforward to understand or measure when that goal has been achieved?
Clear metrics (e.g. code compiling without errors, a completed report) make automation easier.
Is there a lot of data on the task for the AI to be trained on?
Abundant, high-quality data—like code repositories, research papers, or market reports—enables AI to learn quickly and reproduce strong results.

Applying these, we found roles in software coding, research & analysis, writing & content creation, and customer support & administration particularly exposed. Surveys from 2025 show about 90% of US workers worry about AI’s impact on their job security, and around 40% fear complete job elimination.

Looking to the Future: ACI, AGI, and Beyond

A major idea in this episode was Artificial Capable Intelligence (ACI), a term popularised by Mustafa Suleyman, CEO of Microsoft AI. He even suggests a modern Turing test: hand an AI $100,000 in seed capital and see if it can grow this to $1 million on its own. This underscores how deeply AI may soon integrate with (or entirely replace) human decision-making.

Digital vs. Physical

Though AI has made leaps in digital tasks, it remains limited with physical tasks. Humanoid robots still struggle with the flexibility, dexterity, and real-time adaptability humans take for granted. Yet, progress here is speeding up. If (or probably when) advanced robotics converge with agentic AI, a far broader range of jobs could fall into the “substitutable” column.

A “Deep Utopia”?

We also explored the concept of a future where machine intelligence solves all of our problems, leaving human labour largely unnecessary—a scenario dubbed “deep utopia” by Nick Bostrom. The question then becomes: What gives life meaning when work disappears? Some philosophers predict we’ll shift focus to pursuits we now see as leisure—art, exploration, community projects. Others note that certain roles, often summarised as the “three Ps” (priests, prostitutes, parenting), might resist automation because society inherently values the human touch.

The Journey Continues

AI’s potential to reshape the workforce may feel daunting, especially as agentic AI proves itself capable of handling increasingly complex tasks. Yet, if history has taught us anything, it’s that there's more than meets the eye when it comes to technological unemployment. While jobs will inevitably evolve or disappear, there could also be new opportunities on the horizon, especially for those who adapt by honing skills AI struggles to replicate.

We hope this episode clarifies the mechanics of agentic AI and the broader conversation around jobs, displacement, and what the future might hold.

If you enjoyed reading, don’t forget to subscribe to our newsletter for more, share it with a friend or family member, and let us know your thoughts—whether it’s feedback, future topics, or guest ideas, we’d love to hear from you!