As of now, Global GDP growth remains at a modest 3.1 – 3.2% according to the IMF and World Bank. However, some expert opinions such as from Golman Sachs and McKinsey, estimate that AI could increase global GDP 7% or $7 Trillion over the next ten years.
Hidden within the data, the microeconomic landscape is undergoing radical, tech driven reorganisation. The question is, who are looking likely to be the winners and losers of this tectonic shift and what does it mean for the average person.
The Journey So Far
In 2024, US private investment into AI reached $109.1 Billion, equating to almost 9 times that of China. Fast forward to today and 2026 stands as a “year of reckoning” as we enter the period where the same investments are expected to start showing results on the bottom line.
The technology has moved forwards and we are seeing simple automation tasks mature into fully fledged “Agentic” AI. These systems are technically capable of sophisticated, autonomous planning and independent decision making which marks a shift into complex knowledge work.
Whilst this all sounds like things are going in the right direction, the truth is less straight forward. There are numerous examples of pioneering early adopters seeing initially strong results, only to later understand the limitations of the technology and be forced to adjust their strategies accordingly.
Software Development: The “Almost Right” Productivity Paradox
GitHub Copilot and similar tools have revolutionized developer throughput, with users demonstrating task completion speeds up to 81% faster than those working without AI assistance.
By 2025, approximately 84% of developers utilised AI tools, which now generate roughly 41% of all code globally. However, a 2025 study by METR identified a “productivity paradox” where developers expecting 24% gains where actually experiencing slowdowns in certain controlled situations.
The reason for these slowdowns was code that was deemed “almost right” resulting in solutions that failed during final testing. Because AI-generated code often looks correct on the surface, developers spend disproportionate amounts of time debugging and verifying output, work that 46% of developers now treat with active distrust.
Customer Operations: Klarna’s Efficiency vs. Brand Erosion
Klarna opted for an aggressive pivot to AI in an effort to drive radical cost cutting and automation of key areas.
They integrated an Open-AI powered assistant that effectively handles 2.3 million conversations per month, doing the work of 700+ full time agents. Whilst the system slashed resolution times from a baseline of 11-15 minutes down to under 2 minutes and its projected to save the company at lest $40 million annually.
However, analysts from Forrester noted that the company may have over-indexed on automation and it was leading to significant customer dissatisfaction. Many customers reported that the AI agent provided repetitive, generic answers and lacked the ability to navigate any complex inquiries. The company also found that they had lost critical institutional knowledge through layoffs and in early 2025 were forced to rehire human agents to handle the more nuanced conversations that AI could not handle.
Physical Logistics: Amazon’s Robotic Surge and the Human Toll
Amazon has effectively created a “Robotic Revolution” in warehousing, but the resulting efficiency has pushed human endurance to its limits. As of late 2025, Amazon has deployed 1 million robots across its logistics network. The impact was huge, in 2016 the average worker handled and average of 175 packages, in 2025 that number surged to 3870 packages per worker.
New systems like “Sequoia” allow inventory to move 25% faster than legacy sites, while generative AI models like “DeepFleet” have boosted robot coordination speed by 10%. However, the cost of all these “improvements” is felt on the human side of the process. Workers reported higher stress and fatigue levels with AI driven tracking and relentless scheduling creating productivity targets that triggered calls from lawmakers for mandatory impact assessments to protect workers from burnout.
The Geopolitical Divide: Winners and Losers
The AI power is not being distributed equally around the globe either. It appears to be dictated by a nations 4 C’s: Connectivity, Compute, Context, and Competency.
The U.S. remains the leader with 4049 data centres at present, this is nearly double that of the EU. China is aggressively investing to close the gap investing almost $140 Billion into its domestic semi-conductor industry to overcome the impact of Western sanctions.
India and the Philippines are having to evolve their service sectors into knowledge-based services as opposed to voice. In general, low-income nations hold less than 0.1% of the global data centre capacity. They face the “Inverse Balassa-Samuelson” effect, where AI-driven hyper-efficiency in wealthy nations makes the exports of developing countries less competitive.
Systemic Risks to the Outlook
Despite all of the best projections, history teaches us one thing, nothing is entirely predictable and the situation is no different for AI adoption.
Take the energy wall for example. AI power demands could reach 1500 TWh by 2030. To give context to the figure, that is the equivalent to the energy needs of Japan.
Labour displacement is another very big question mark, not just for production but also consumption. Whilst many new roles may be created due to the advent of AI driver operations, its estimated that 92 million jobs face being displaced by 2030 for the same reason. This alone would have vast implications for social stability if reskilling or other suitable solutions are not found.
Finally, the biggest risk of all comes from something many are speaking about but few in the line of fire want to admit. The chances of an AI bubble bursting are magnified as the promised gains from hyper investment are not realised. The longer the promise is the only thing being sold, the higher the chances become of a global .com style bust occurring.
There are many parallels with the .com bubble of the 2000s with companies making massive investments and then witnessing a diffusion lag as they wait for the tech to fully pay off.
In the late 90s, it took years for the internet to actually improve the bottom line. Today, we see a similar paradox: developers use tools like GitHub Copilot to work faster, but 66% of them spend their “saved” time fixing “almost right” code that the AI hallucinated.
We are seeing stock valuations that are propped up by as yet “unrealised gains” with investors again betting on future improvements that still have not shown up in GDP data, yet.
Does this mean AI is not the next big thing for humanity, almost certainly not. AI will drive exponential progress for humanity, the question is, have we bet the farm on a product that’s simply not ready yet.
Time will tell.