These days, it’s hard to come across a startup that isn’t related to artificial intelligence in some way.
Until a few years ago, AI was considered a part of technology. But these days, the term “technology” itself seems to be completely obsolete.
Companies used to call themselves “digital first,” but now they are “AI first,” and what this actually means has yet to be properly defined.
Until the release of OpenAI’s ChatGPT, AI startups typically worked on narrow, specialized use cases such as machine learning, computer vision, and predictive analytics that are typically built into enterprise software and rarely visible to end users.
With the launch of ChatGPT, a new crop of AI startups have emerged that are using the computing power of large-scale language models (LLMs) to build platforms and services on top of them. These so-called “wrappers” package models of ChatGPT for specific tasks, languages, or user groups, and have exploded across the Mena region.
“Rapper” has almost become a slur. Critics say these startups are vulnerable because they don’t own the core technology.
Generative AI and LLMs such as Google’s Gemini, Anthropic’s Claude, and ChatGPT have significantly lowered the barrier to entry and become the fastest way to go from idea to pilot within days, if not hours.
The cost is minimal and requires proper prompting and time rather than hiring thousands of engineers to build infrastructure and train your own models.
However, in some circles, the word “rapper” is almost derogatory. Critics argue that these startups don’t own the core technology, making them vulnerable if the LLMs they use move into their own niches or change their prices.
For some in the Middle East, it also highlights the more unpleasant and unfortunate reality that we continue to be consumers of AI innovation rather than pioneers.
One of the frequent criticisms of Mena’s startup ecosystem over the years is that it is a copycat ecosystem full of companies that copy successful models elsewhere and localize them, rather than building something truly unique.
Many believe that this indicates a lack of innovation, and they are not completely wrong. AI wrappers are the latest iteration of this follower mentality, plugging into what works in the US or Europe, translating it, localizing the user experience, and shipping it.
But just as super app Careem successfully localized Uber’s model for Mena, some AI wrappers can truly add value and become truly defensible when built on their own data.
Possible examples include legal research tools trained on GCC case law or logistics agents integrated into port operators’ systems. These are tools that can survive because they are tied to data and workflows that the underlying LLM provider does not have.
Read more:
Read more:
There are signs that the region can innovate at the deep tech level. Tarjama’s Arabic.AI is the world’s first B2B advanced LLM designed specifically for Arabic enterprise applications to understand the landscape in fields such as law, medicine, and business.
Meanwhile, the UAE’s Stargate’s partnership with OpenAI and G42 to build one of the world’s largest AI data centers signals a shift from using AI models to hosting and shaping them.
Saudi Arabia is doing something similar with Humane, which is backed by the sovereign wealth fund PIF, building a model of native Arabic speakers familiar with Islamic values.
Rappers will ultimately have a short game. Infrastructure development is a long-term battle that both Saudi Arabia and the UAE are engaged in. This region is currently built on top of ChatGPT, but we strive to avoid having to do so all the time.
Triska Hamid is a writer and angel investor focused on technology and startups in the Middle East.
