Last year, we surveyed firms in emerging markets backed by private equity (PE) and venture capital (VC) funds, which we refer to as frontier firms, and found that almost 40 percent were using artificial intelligence (AI). This year, the picture looks considerably different.
The International Finance Corporation’s (IFC) latest survey of 1,666 frontier firms with significant activity in emerging markets, shows that AI has moved into the mainstream: 71 percent now report using AI, whereas half of all adopters say they began using AI during 2025. As the World Development Report 2026 highlights, AI offers developing economies a pathway to boost productivity, but only if it is adopted, adapted to local conditions and used at scale. Frontier firms are typically the first to adapt global technologies to local contexts, and pave the way for other firms to follow.
Adoption, however, remains shallow for many firms. We distinguish between internal adopters, which use AI in their own operations, and external adopters, which build it into products or services for customers. More than half of internal adopters (57 percent) and nearly half of external adopters (45 percent) are still experimenting with AI or deploying it on a limited scale. Far fewer have made AI central to how they operate: just 10 percent of internal adopters have embedded it across their core internal processes, and 19 percent of external adopters describe AI as core to their business model.
What prevents firms from moving further up the AI ladder? Skills constitute the most frequently cited constraint, ahead of other commonly discussed barriers such as costs or technical infrastructure.
When firms were asked about barriers to (further) AI adoption, more than four in ten non-adopters identified skills shortages as their main obstacle, compared with about one quarter of current adopters.
Perceptions of barriers also shift with depth of adoption: firms using AI more extensively are more likely to say their current level of adoption is appropriate. This could reflect selection, as more capable firms both deploy AI more deeply and face fewer constraints, or experience, as firms that successfully move from pilots to broader use develop clearer use cases and invest more heavily in AI as a result. Among shallow adopters, however, the same answer may partly reflect limited awareness of what deeper use of AI could offer their business.
The skills needed for AI adoption are not all the same. Firms require a spectrum of capabilities, ranging from basic AI literacy to advanced technical and managerial expertise.
At the foundation are employees who can incorporate AI tools such as ChatGPT, Claude, and similar applications into their daily work (e.g., analyzing customer feedback or creating marketing content). Most firms appear to have reached this stage: around 70 percent report that staff regularly use AI applications.
The next rung of the ladder involves the ability to integrate and deploy AI solutions within business processes in a way that changes workflows and client-facing products. Only about half of surveyed firms report having these capabilities. Fewer appear to have the expertise needed to develop and adapt AI models internally: just 30 percent indicate that they can build AI solutions in-house. For example, an HRtech startup in Latin America highlights that AI has enabled it to accelerate execution of the product roadmap, deliver more customized features for clients and embed new AI capabilities into product development. The challenge is no longer whether firms have heard of AI. It is whether they have the capabilities to move from experimentation to transformation.
Closing this gap is important because firms with in-house capabilities can adapt AI to their specific context and are more likely to embed it into products, workflows, and proprietary data pipelines. Without continued skills development, today’s adoption gap could evolve into a deeper divide between firms that simply use AI and those that leverage it to innovate and improve business outcomes.
The survey shows that many firms are actively investing in building AI capabilities. Nearly half of AI adopters (43 percent) report deliberately upskilling existing employees to support AI adoption. In contrast, relatively few firms report relying primarily on hiring specialized talent (9 percent) or outsourcing AI-related activities (1 percent).
Yet progress is uneven. About half of adopting firms do not appear to have a deliberate strategy for building AI capabilities. Some companies expect employees to do more within existing roles (24 percent), while others (23 percent) report making no significant changes to workforce skills.
Upskilling the existing workforce can build foundational AI skills, but it remains unclear whether training alone can build the capabilities needed to adapt AI to a firm’s operations and market context. Complementary approaches, such as hiring AI specialists and strengthening managerial skills, may be needed to drive the organizational change toward an AI-enabled or AI-native model.
Perhaps the most striking finding is that AI adopters are already seeing results. An overwhelming share (93 percent) has self-reported positive business outcomes. Productivity improvements are the most frequently cited benefit (54 percent), far outweighing gains in product innovation (15 percent) or other outcomes. Notably, firms that have adopted AI more deeply are more likely to report benefits related to innovation. This matters because product innovation, more than efficiency gains alone, is how AI is likely to drive firm growth.
The initial diffusion of AI in emerging markets suggests that frontier firms that invest in strengthening skills and organizational capabilities are better positioned to move beyond experimentation towards deeper adoption, adapt AI to local and firm-specific contexts, and translate it into widely shared productivity gains and innovation.
Multiple stakeholders have a role to play. Firms need to continue investing in workforce skills to support AI adoption and adaptation, but also complement existing talent with advanced technical expertise to enable deeper integrations. Investors can help portfolio companies by sharing expertise, identifying promising use cases, and supporting access to talent and capabilities. Policymakers can strengthen human capital development and the innovation ecosystems that underpin technology adoption and development.
Together, these efforts will help determine whether frontier firms, and the wider population of firms around them, climb the AI ladder or stall on its lower rungs.
Source : World Bank
