Size, Not Customisation the Next AI Race

Size, Not Customisation the Next AI Race

Jul 4 , 2026. By Birhanu Beshah (PhD) ( Birhanu Beshah (PhD), ( birhanu.beshah@aait.edu.et) is an associate professor in the School of Mechanical & Industrial Engineering of Addis Abeba University (AAU). )

As Artificial Intelligence (AI) adapts to each profession and country, Ethiopia risks remaining a consumer of intelligence built elsewhere rather than a producer of its own.

For decades, software developers chased a single ambition. They wanted to build one application that could serve millions of users. Artificial intelligence (AI) is now turning that logic on its head.

Instead of forcing users to adapt to the software, AI is adapting itself to individual users, organisations, and industries. The race in artificial intelligence is no longer about building the largest language model. It is about enabling the largest number of customised AI applications.

The first generation of generative AI brought powerful conversational assistants capable of answering almost any question. Millions of people now lean on AI to write reports, summarise documents, generate code, and support research. Yet the limits of general-purpose AI soon became clear. Every profession has its own language, as every organisation has its own procedures and every country its own regulations. A generic AI can assist everyone, but it cannot fully understand anyone.

That realisation has set off a global movement toward mass customisation. Technology companies have read the shift.

OpenAI lets users create specialised GPTs, while Google offers Gems through Gemini. These platforms allow individuals and organisations to build AI assistants with their own instructions, domain knowledge, workflows, and personalities, without advanced programming skills. The value of AI is shifting from the model itself to the ability to shape it for a specific purpose.

The development echoes the economic theory of mass customisation that management scholars introduced in the 1990s. AI is making that vision real, not for physical products but for intelligence itself. Intelligence is becoming configurable, with consequences that extend well beyond convenience.

Customised AI sharply cuts the cost of expertise. A university can build an AI tutor that knows its curriculum, and a hospital can deploy an assistant trained on clinical guidelines and administrative procedures. A customs authority can create an adviser fluent in tariff schedules, customs law, and inspection protocols. Law firms, banks, manufacturers, and farmers can all develop assistants fitted to their own working conditions. Each of these tasks was once the domain of scarce specialists, and a customised assistant puts that knowledge within closer reach.

For Ethiopia, the opportunity could be considerable. The country has adopted an ambitious digital transformation agenda, yet much of the available AI is trained primarily on global data and on foreign institutions. Customisation offers a practical answer by embedding local knowledge, regulations, languages, and institutional practices into AI systems.

Ethiopia also faces a strategic risk, as many of its institutions are becoming eager consumers of AI without becoming its producers. Employees use ChatGPT or Gemini on their own, but institutions rarely capture that experience to build organisational intelligence. The distinction is between a consumer who buys intelligence shaped by others and a producer who shapes intelligence around its own institutions.

The transition, therefore, requires more than technology; it demands organisational change. Universities should fold AI customisation into teaching, and institutions should encourage staff to build custom assistants. Government should promote affordable access to APIs, the connections that let software tap AI models, along with cloud infrastructure and local-language datasets, while setting governance rules for privacy, security, and accountability.

An equally promising route is collaborative customisation. Sector-wide AI platforms could enable universities, healthcare institutions, and government agencies to build shared foundations that individual bodies can then tailor to their own needs.

The race toward mass customisation of AI has begun. Ethiopia should not take part merely as a consumer of customised intelligence built elsewhere. It should build its own ecosystem of customised AI applications that serve its institutions, languages, and development priorities. In the emerging digital economy, intelligence itself is becoming a customisable resource, and those who learn to shape it will lead the next wave of economic transformation.

PUBLISHED ON Jul 04,2026 [ VOL 27 , NO 1366]