Most everyday jobs do not actually need the most powerful, cutting-edge AI models, argues Jovan Kurbalija in his blog post ‘Do we really need frontier AI for everyday work?’. While frontier AI systems dominate headlines with ever-growing capabilities, their real-world value for routine professional tasks is often limited. For many people, much of daily work remains simple, repetitive, and predictable.
Kurbalija points out that large parts of professional life, from administration and law to healthcare and corporate management, operate within narrow linguistic and cognitive boundaries. Daily communication relies on a small working vocabulary, and most decision-making follows familiar mental patterns.
In this context, highly complex AI models are often unnecessary. Smaller, specialised systems can handle these tasks more efficiently, at lower cost and with fewer risks.
Using frontier AI for routine work, the author suggests, is like using a sledgehammer to crack a nut. These large models are designed to handle almost anything, but that breadth comes with higher costs, heavier governance requirements, and stronger dependence on major technology platforms.
In contrast, small language models tailored to specific tasks or organisations can be faster, cheaper, and easier to control, while still delivering strong results.
Kurbalija compares this to professional expertise itself. Most jobs never required having the Encyclopaedia Britannica open on the desk. Real expertise lives in procedures, institutions, and communities, not in massive collections of general knowledge.
Similarly, the most useful AI tools are often those designed to draft standard documents, summarise meetings, classify requests, or answer questions based on a defined body of organisational knowledge.
Diplomacy, an area Kurbalija knows well, illustrates both the strengths and limits of AI. Many diplomatic tasks are highly ritualised and can be automated using rules-based systems or smaller models. But core diplomatic skills, such as negotiation, persuasion, empathy, and trust-building, remain deeply human and resistant to automation. The lesson, he argues, is to automate routines while recognising where AI should stop.
The broader paradox is that large AI platforms may benefit more from users than users benefit from frontier AI. By sitting at the centre of workflows, these platforms collect valuable data and organisational knowledge, even when their advanced capabilities are not truly needed.
As Kurbalija concludes, a more common-sense approach would prioritise smaller, specialised models for everyday work, reserving frontier AI for genuinely complex tasks, and moving beyond the assumption that bigger AI is always better.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
