Seventy-one percent of 1,666 surveyed firms operating in emerging markets said they used artificial intelligence, but most adopters had not yet moved beyond experimentation or limited use, according to an International Finance Corporation survey.
The findings show how quickly AI tools have entered business operations, while highlighting the gap between trying the technology and integrating it into the way a company works. The survey focused on venture-capital and private-equity-backed businesses with significant operations in emerging markets. It does not represent all companies in those economies.
Among firms using AI internally, 57% were experimenting or using it in a limited way, according to the World Bank account of the IFC survey. Just 10% of internal adopters had embedded AI across core processes.
Trying a tool is different from changing a process
AI can support activities ranging from drafting and data analysis to customer service. A company may give employees access to a tool without changing how work is assigned, checked or measured. Deeper integration can mean connecting a system to an established workflow, defining who reviews its output and deciding what happens when it is wrong.
That difference matters when executives assess returns. A demonstration can show that a tool is capable of completing a task, but it does not establish whether using it across a team saves time, improves quality or just shifts work to someone else.
Related research has also examined the choices workers and employers face when training AI systems. The survey also distinguished between internal applications and AI used in products or services offered externally. Among external adopters, 45% were experimenting or using the technology in a limited way, while 19% described AI as central to their business.
Training is part of the investment
Among the surveyed AI adopters, 43% said they deliberately upskilled existing employees to support adoption. Nine percent primarily hired specialized talent and 1% primarily outsourced.
Those categories do not mean a business cannot use more than one approach. They do suggest that existing staff play a substantial role in putting tools to work. Training can include knowing when to use a system, checking its output, handling confidential information and escalating mistakes.
For a smaller company, the cost of that work may be less visible than a subscription or software contract. Managers assessing a proposed deployment can account for the time needed to prepare data, test results and update procedures as well as the price of the tool.
Reported benefits need careful interpretation
Most adopters reported positive outcomes. The World Bank said 93% cited a positive effect, with 54% reporting productivity improvements. Those are firms’ own assessments, not independently measured gains, and the survey does not establish that AI caused the reported improvements.
Results are also shaped by the sample. Venture-backed and private-equity-backed firms may have different access to funding, skills and technology from a typical small business. The figures should not be read as an adoption rate for every business in emerging markets.
For managers considering their next step, a useful test is to define one measurable task before expanding access. A company can compare completion time, error rates and the amount of review required, then revisit the decision as the process changes. The survey suggests that many firms are still at the stage where those practical questions matter more than the simple fact that they have tried AI.
