AI impacting jobs in 2025 (and beyond)
The release of the 2025 Future of Jobs Report by the World Economic Forum triggered press to raise concerns about the impact of AI on jobs. CNN’s headline following the report read: “41% of companies worldwide plan to reduce workforces by 2030 due to AI” and the bold statement “AI will be a job killer”.
It is worth noting that in comparison with previous editions, the 2025 Future of Jobs Report presents more conservative numbers on the impact of AI on jobs. Being exposed to the technology, surveyed companies have adjusted their expectations and demonstrate in 2025 a more nuanced view on the impact from AI on tasks becoming obsolete or subject to change.
On the other hand, the WEF report obviously could not cover the latest technology improvements in AI. From the report: “….. the capacity of GenAI to substitute a human in executing specific skills, based on an assessment by GPT-4o of its own ability …”.
With the introduction of Large Reasoning Models, eg. OpenAI’s O3, we enter the next level of Generative AI capabilities, does that make the perspective on AI impacting jobs worse?
Let’s step back and overlook this question from various angles:
In publications we can read OpenAI’s O3 demonstrated an impressive 87.5 percent score, which is comparable to human performance at 85 percent threshold in the ARC-AGI benchmark. We have come to the point that for complex tasks, organisations could consider the use of an AI system, and as such we have to reconsider impact on jobs. However in doing so, we should take into account that for an AI system to outperform a human, high compute processing is required, which translates into a 2 or even 3 digit cost per task (O3 test report). Based on this we can conclude that for now, reasoning models are too expensive to consider for operational use that would replace jobs.
Reducing CO2 footprint from AI
The more capable LRM models come with the wave of high compute. The World’s tolerance for heavy compute solutions is rapidly changing. Other technologies (eg. Advances in AI-focused chips) need to be commonly adopted first in order to drive down the CO2 footprint levels that society and companies are willing to accept. The creation of these chips comes with its own en
vironmental considerations that will need to be addressed. With these challenges, it is likely to take a while before reasoning models will be consumed at scale, and impact jobs.
Despite the sometimes impressive improvements of new models released, the basic issue remains unchanged: models hallucinate. With the release of Large Reasoning Models like OpenAI’s O3, there was hope that hallucination would become a minor issue. Unfortunately, this is not the reality. The fact that we have found ways to compensate hallucination, does not necessarily change trust levels in users. For fault intolerant use cases, we simply cannot accept mistakes from a computer and we don’t want to use AI without human validation. In other words: AI and end-to-end automation do not get along well for these use cases, and the primary use of AI remains to augment people in their job. Most systems are likely to remain with a human in the loop for the foreseeable future.
Companies struggle to adopt AI
Recent surveys confirm: about 75% of the companies struggles to scale AI and generate value from AI. Only 25% of the organisations are able to scale and generate an impact for jobs. These studies report failing business cases and companies state of AI readiness to be hurdles to overcome. Recognizing that a successful approach to AI requires a company to think big and adopt a more strategic and holistic approach to AI is a start, but really making that change is something that takes time. It remains to been seen how companies will catch up in AI adoption in 2025.
The increased demand for AI skills is creating dilemmas on how to secure the skills needed to deliver successful AI implementations. 63% of respondents (Future of Jobs Report ) identified reskilling their workforce as a major barrier, with an estimated 59 in every 100 roles expected to require upskilling by 2030. At the same time, a whopping 70% of employers intend to hire in these new skills. With years of neglect in developing AI related skills, there is already a major talent shortage with top professionals in the field able to command eye-watering salaries. With the time and investment needed to build strong AI skills and combine them with industry experience, all signs point towards major struggles to hire and retain the right talent in the near future. Employers may need to take on the burden of upskilling or find other ways of sourcing skills whilst the job market catches up with the demand.
Take away
It is hardly a surprise, the impact of AI on jobs is not dictated by the state of technology or the pace of the improvement of AI. It depends on the ability of companies and society to adopt, catch up and do something useful with it. This article touches on common challenges, struggles and priorities for companies and industry, and they are not that easy to tackle. To add, all surveys referenced in this article, reveal that companies being exposed to the technology have adjusted their expectations downwards and demonstrate a more nuanced view on the impact from AI and their ability to adopt AI. We like to think that these insights will help to shape your own conclusion about the impact form AI on jobs in 2025 (and beyond).
info@aimetimpact.nl
+31 6 52 03 32 70
KVK: 96361662
BTW: NL005206141B72
