The business world is still coming to grips with the seismic potential of generative AI (gen AI). According to the McKinsey Global Institute, the technology could contribute $2.6 trillion to $4.4 trillion annually to the global economy. In addition, 50 percent of work activities could be automated from 2030 to 2060, a trend that gen AI is poised to accelerate.1 However, one does not need to look to the future to see the impact of the technology: gen AI is already being integrated into industries around the world, from pharmaceuticals to financial services.
The influence and velocity of gen AI will vary across regions. McKinsey estimates it could add $20 billion to $30 billion a year to the economy of Central America and the Caribbean by 2030, equal to about 5 percent of the region’s GDP.2 In particular, it could reshape financial services, agriculture, and business process outsourcing, among other industries. Along with its benefits, however, gen AI presents unique challenges and risks that regional leaders must address.
This article explores the impact, challenges, and risks associated with gen AI and highlights the steps companies in the region can take to unlock its benefits in a fast, sustainable, and responsible manner.
How gen AI could generate value in Central America and the Caribbean
Our analysis reveals that gen AI could create value for businesses in three main ways: increasing worker productivity, augmenting skills, and enabling innovation and growth.
Increase worker productivity
McKinsey research estimates that 60 to 70 percent of work activities could be exposed to some degree of automation through gen AI.3 By automating or assisting with repetitive, tedious work, gen AI could significantly boost worker productivity by freeing up employees to tackle more complex, creative tasks and reduce costs. For example, a leading bank in the region used gen AI to increase its developers’ productivity by 40 percent, accelerating its time to market and capacity for innovation. Similarly, various contact centers around the globe have deployed gen AI virtual copilots to assist agents in responding to customer requests, improving productivity by 15 percent, according to McKinsey analysis.
Augment skills
From interacting with bank customers to caring for hospital patients, gen AI can assist in activities such as insight generation and recommendations, ideation, and problem solving. These uses could not only allow employees to cover more ground but also elevate task quality. In the agricultural sector, for example, gen AI models will be able to analyze data on soil composition, weather patterns, and crop characteristics, providing farmers with valuable information (for instance, through chat) about optimal planting strategies, irrigation techniques, and pest control. Such insights could help increase the industry’s productivity and sustainability.
Enable innovation and growth
Gen AI could also fuel innovation and growth in the region by creating new products and services and empowering companies to address problems that they previously perceived as intractable. For example, in the financial sector, gen AI can improve risk models, which would allow unbanked individuals to join the financial system. In education, gen AI can facilitate the development of interactive and personalized educational materials, increasing access for students throughout Central America. And in all sectors, gen AI can harness unstructured data to inform decision making.
Addressing the challenges and risks of gen AI
While gen AI offers tremendous potential, Central American companies considering implementing this technology should be aware of several challenges and risks.
Workforce movements and changes in talent requirements
By 2030, based on McKinsey analysis, millions of workers in Central America will likely have to change occupations or roles because of gen AI. The jobs most likely to be affected by gen AI are those related to office support and customer service. In parallel, gen AI could spur demand in fields such as prompt engineering, data engineering, and data, jobs requiring new skills that are scarce in the current labor market (see sidebar “Typical roles in a generative AI center of excellence”). Workers in lower-wage jobs in support or back-office functions (such as call-center representatives or administrative personnel) are more likely to need to change occupations than those in higher-paid positions, and most may need to acquire additional skills to do so successfully.
This dynamic poses a significant challenge for Central America and could require close collaboration between the public and private sectors: less than half of the population have sufficient skills to use computers for basic professional tasks, and workers have demonstrated low adoption rates of virtual tools in recent years.4 The region must proactively address this risk by investing in training and upskilling programs that can prepare the workforce for this transition.
Availability and quality of data and information systems
Gen AI models require high-quality data to produce accurate and reliable results. Where traditional AI typically relies on structured data, gen AI draws on unstructured data such as text, images, and video, presenting new challenges even for companies with established AI operations. Further, managing heterogeneous information systems, including legacy tools and both on-premises and cloud initiatives, complicates integration and maintenance efforts.
In Central America, companies still struggle with data availability, quality, and storage capacity. Lack of data or proper information systems could put the region at a disadvantage and complicate gen AI adoption. As a result, some companies could have to make large investments to ensure the availability of diverse and representative data sets.
Ethical implications, model biases, reliability and accuracy, and intellectual property infringement
Gen AI models can reflect biases implicit in the data used to train them and inadvertently share private information embedded in training data. Organizations will need to implement guidelines to prevent biased outputs that could affect users and cause reputational damage. They will also need to implement regulations to protect the privacy of individuals. Since most organizations in Central America are still in the initial stages of data governance and AI models, these actions are even more crucial.
Gen AI can also be used to create highly realistic but fake images, videos, or text (known as deepfakes), making it difficult to determine the authenticity of content. These capabilities pose a significant risk of misinformation and manipulation, which can lead to social unrest, political instability, and reputational harm.
When such events occur, assigning accountability can be extremely difficult because of the wide range of stakeholders—from component developers and data providers to cloud service providers and model suppliers. For example, if an AI model generates an incorrect medical diagnosis that leads to the death of a patient, which party would be responsible?
Data used to train gen AI models could also present significant intellectual property (IP) risks, including the infringement of copyrighted materials, trademarks, patents, or other legally protected assets. When using a supplier’s gen AI tool, organizations will need to safeguard their data to protect their IP and the IP of companies in the region. Policy makers could provide clear guidance by establishing frameworks to determine ownership rights and create accountability for the outputs from gen AI.
How organizations in the region can capture the value of gen AI
In a recent McKinsey survey of more than 100 organizations with annual revenues of more than $50 million, 63 percent of respondents said one of their top priorities is implementing gen AI in their organization.5 However, 91 percent of these respondents do not feel prepared to do so responsibly and sustainably. Central American companies can pursue seven strategies to rewire their organizations and accelerate their gen AI implementation journey.
1. Define a long-term strategy and reimagine domains from end to end rather than focusing on use cases
Gen AI has significant potential to redefine certain business domains along the value chain (for example, marketing for a B2C company or operations for a manufacturer). However, the ease of deploying it can tempt organizations to apply it to isolated use cases, resulting in a minimum viable product (MVP) and proof-of-concept paralysis from a lack of sponsorship, impact, or both. To harness gen AI for a domain such as marketing, customer service, or finance, companies should consider taking an end-to-end view. Through this exercise, organizations can pinpoint use cases that would enable a domain’s transformation and develop a long-term strategy that includes organizational and cultural changes, capability building, and investments (see sidebar “How generative AI could reimagine customer service”).
2. Select a lighthouse domain to demonstrate the transformational potential of gen AI and the feasibility of its application
To achieve a transformation with gen AI and build momentum, organizations will need the buy-in of key stakeholders. They can gain that buy-in by showing how the technology can elevate the business and generate value. As with traditional AI, this goal can be achieved by focusing on lighthouse domains: areas or business functions that have the potential to produce substantial value with the assistance of gen AI, are visible to the rest of the organization, play an important role in the business strategy, and have sufficient technological maturity to facilitate the implementation of gen AI (exhibit). The appropriate selection of lighthouse domains will vary by the type of organization. At a bank, for example, a lighthouse domain could be customer acquisition or collections.
Once the lighthouse domains and gen AI use cases have been identified, we recommend identifying one or two use cases that are easy to implement and launching a proof of concept to quickly test and refine them before scaling to adjacent use cases. By focusing on quick wins that deliver significant results, companies can build support and expand the transformation, leveraging the multipurpose nature of gen AI.
3. Build a cross-functional governance team committed to gen AI
Gen AI requires a deliberate and coordinated approach to balance its impact with the accompanying risk. We recommend forming a cross-functional group of leaders—for example, leaders from data science, engineering, legal, cybersecurity, marketing, design, and other business functions. This group could not only help identify and prioritize the highest-value use cases but also serve as a control tower, capturing impact, building capabilities, and executing a coordinated, safe implementation across the organization.
4. Create an integrated technology ecosystem with adequate data
Large language models are just the tip of the iceberg; other essential factors are required to successfully implement gen AI use cases. A clear technology and data strategy designed to generate enterprise value and competitive advantage is also critical. Each use case has its own requirements for an organization’s technological infrastructure. Successful implementation requires key elements, such as a robust cloud infrastructure, a front end to enable user access to the technology, and traditional AI models in production, among others. Companies may also need to process new sources of unstructured data, design a tagging structure that properly organizes this information, build infrastructure that enables access to new sources of data, and establish governance for data quality and security.
5. Develop the required talent and skills
Deploying gen AI to create business value requires companies to develop their technical capabilities and current workforces. Top management should assess prioritized use cases and ensure the organization has the necessary skills, which can encompass technical roles and a mix of talent in engineering, data, design, risk, product, and other business functions.
In addition to hiring the right talent, companies must train and develop their existing workforces.6 Gen AI applications have intuitive prompt-based conversational user interfaces, but users still need to optimize their prompts, understand the limitations of the technology, and know where and when they can integrate the application into their workflows. Users should also be trained to identify and mitigate the risks of gen AI.7 Business leaders in Central America could give clear guidelines on the use of gen AI tools and provide ongoing education and training to enable employees to effectively use the technology.
6. Balance risk and value creation
The best use cases consider both the organization’s risk tolerance and the opportunity’s potential impact. For example, an organization might prioritize a less-valuable but low-risk use case such as creating initial drafts of marketing messages for an employee to review before sending. Notably, technology users and reviewers will need training to properly manage gen AI risks.8
In parallel, a company could work on higher-value, high-risk use cases, such as a virtual agent that interacts directly with customers and addresses their requirements. Implementing this use case would require developing a robust guardrail system, which could take additional time and resources, to safely capture the use case’s value.
7. Forge strategic partnerships
Organizations don’t have to develop all the applications or models themselves; they can partner with gen AI vendors and experts to move faster.9 Similarly, they don’t need to create all their risk management guardrails from scratch but can instead collaborate with third parties. Companies can tap cloud service providers for their computing capabilities. In general, organizations will need to analyze the requirements of each gen AI application, how vital each application is to the company’s strategy, and how much they expect to differentiate their offerings through the use case. With this analysis, companies will be able to define their partnership and investment strategy.
Gen AI has immense potential to transform productivity across various Central American sectors, including agriculture, education, banking, and contact centers. However, its implementation comes with challenges for data availability and quality, technology requirements, workforce movements, changing talent requirements, ethical considerations, and risk management. By carefully navigating these issues, the region’s businesses can harness gen AI’s potential to drive economic growth, improve quality of life, and foster innovation in a safe and responsible manner.
