Meta’s Muse and OpenAI’s Dots represent the newest generation of AI personal assistants, built to manage daily tasks such as reserving restaurant tables, purchasing groceries, and handling payments on users’ behalf. Since its debut in September, Muse has been downloaded 9 million times across the U.S. and Canada, with 1.8 million daily active users, according to market tracking. Meta’s leadership, including chief AI officer Alexandr Wang, has emphasized global expansion, describing the ambition to extend these tools to billions worldwide.
While major technology firms pursue worldwide adoption, smaller companies in emerging markets are already assessing how these agents perform outside Western frameworks. In South Africa, ManyHands has created Manny, a WhatsApp-based assistant capable of arranging grocery deliveries, scheduling appointments, and processing payments. The project faces significant obstacles, particularly around language and cultural adaptation. South Africa recognizes 11 official languages, and founder Anton Moulder has noted the difficulty of training an AI to communicate effectively in all of them.
Language barriers extend beyond mere translation. Large language models, which underpin these AI systems, deliver weaker performance in non-English contexts, potentially leaving English speakers with superior access. Payal Arora, a professor specializing in inclusive AI cultures at Utrecht University, characterizes the issue as one of systemic power imbalances. Without localized language support, these agents risk imposing Western communication norms rather than accommodating regional styles. In many Global South contexts, meaning is shaped by familial, communal, and informal networks—dimensions that current AI models may overlook.
Cultural and language gaps limit AI reach
Usha Raman, an independent AI policy researcher, has cautioned that unadapted agents could reinforce Western cultural and economic dominance. Meta has acknowledged the necessity of local language integration and regional service compatibility, though scaling these solutions globally remains unproven.
The proliferation of AI-driven activity also threatens to overwhelm already strained internet infrastructure in developing nations. These agents don’t merely interact with users; they query websites, process transactions, and relay information, creating a surge in network demand. Research from Cisco indicates agents generate up to 450% more traffic per task than human users. Cloudflare’s data shows a 1,700% increase in daily agent requests over the past year.
In regions with ample network capacity, such as East Africa, the impact may be negligible. However, in countries with weak connectivity, including Niger, Burkina Faso, and eastern Congo, the strain could be severe. Kyle Spencer, executive director of the African IXP Association, highlights physical vulnerabilities like road construction damaging fiber cables or undersea cable failures. Even cloud-based agents could fail to complete tasks or reach users for approval if connections are disrupted.
Cost and connectivity block Global South access
The digital divide in Africa remains pronounced: only 36% of Africans had internet access in 2025, compared to 92% of Europeans and 88% of North and South Americans, according to the International Telecommunication Union. Urban Africans are 2.6 times more likely to be online than rural residents. While Muse is currently free, competitors like Dots charge $100 per month, creating a financial barrier for most users in the Global South, as noted by Jane Munga, a fellow at the Carnegie Endowment for International Peace.
Local businesses would also require significant adjustments to engage with AI agents. To interact effectively, they must ensure websites are bot-accessible, maintain digital inventories, and support automated payments, demands that small, offline vendors may find impossible to meet. If these systems operate from servers abroad, local regulators may lack the tools to oversee their operations.
For AI agents to succeed in markets like South Africa, they must accommodate the dominant informal economies. Many transactions in the Global South occur outside formal banking systems, relying instead on cash, barter, or mobile money platforms like M-Pesa. While Manny integrates with WhatsApp Pay, a widely used service in Africa, broader adoption would require agents to handle fragmented payment methods. Small vendors often depend on cash or local digital wallets lacking standardized application programming interfaces. Without seamless interoperability, agents risk excluding the very populations they aim to assist. Experts emphasize that even routine tasks, such as confirming delivery times or processing refunds, could fail if underlying payment infrastructure isn’t bot-compatible.