Of all the various AI technologies, generative AI (gen AI) has arguably captured the most excitement among organizations in the business world and the public sector. Its ability to automate activities across a range of functions—customer interactions, content generation, and computer coding, for example—promises to reshape how much work gets done and boost performance.
McKinsey research has estimated that the application of gen AI to 63 use cases could generate global annual economic value worth between $2.6 trillion and $4.4 trillion, adding 15 to 40 percent to the value we previously estimated that other AI technologies, such as machine learning, advanced analytics, and deep learning, could unlock. In Gulf Cooperation Council (GCC) countries, the same gen AI use cases could generate between $21 billion and $35 billion a year, on top of the $150 billion that other AI technologies could deliver. To put that into perspective, gen AI could be worth 1.7 to 2.8 percent of annual non-oil GDP in the GCC economies today (Exhibit 1).
Recognition of that potential and the likely disruption it will unleash across sectors has prompted a surge in gen AI investments and partnerships in the GCC region, both by governments keen to position their economies as global leaders in the technology as well as by private organizations. (For examples of some of the high-profile ones, see sidebar “Companies and governments chase gen AI rewards.”) But looking more broadly across the region, how well-positioned are organizations to reap the benefits of gen AI?
To gauge the answer, McKinsey partnered with the GCC Board Directors Institute to survey 140 executives from eight different sectors in the region and conducted a series of interviews (for more details, see sidebar “About the research”).
The conclusion is that while many organizations are investing in gen AI, few have begun to scale its implementation and extract value from their investments. A small number—those we call the value realizers—stand out from the crowd, however, already generating more than 5 percent of their earnings from gen AI. The survey suggests these are also the ones that have made the most progress in rewiring their organizations to build the capabilities required to adopt gen AI. There are lessons therein for others looking to unlock its rewards.
The state of play
Many GCC organizations are taking prompt action to capture the surge in value that gen AI offers. Almost three-quarters of respondents say gen AI is already being used to some extent in their organizations. Many have also developed a gen AI strategy and road map and have directed budgets to areas where gen AI is likely to have the most impact. But on all fronts, value realizers are pushing harder.
Exhibit 2 shows the potential value gen AI could deliver by sector in the GCC economy. The value potential is the biggest by far in the energy sector, where annual gains of between $5 billion and $8 billion are possible, followed by capital projects and infrastructure and financial services.
Many GCC organizations are taking prompt action to capture their share of this. Almost three-quarters of survey respondents said their organizations were already using gen AI in at least one business function. That compares with 65 percent of respondents in a similar global McKinsey survey conducted earlier this year. And 57 percent of GCC respondents said their organizations were investing at least 5 percent of their digital budgets on gen AI. That compares with 33 percent globally. Moreover, half of the GCC survey respondents said their organizations had drawn up a road map to implement priority use cases more broadly—again a higher proportion than in the global survey, where the figure was 26 percent.
Chasing the value
According to the survey, GCC organizations are focusing their early efforts on functions that McKinsey research shows are likely to deliver the most value—namely marketing and sales and software engineering and IT (Exhibit 3).
One chief information officer (CIO) pointed out that while it might have been tempting to start applying gen AI use cases across all functions, the wiser move was to start in sales and marketing, where leading global companies had proved gen AI’s value. The chief technology officer (CTO) of a large education organization was also led by the evidence. “It made sense for us to start with software engineering because of the proven use cases in code generation and bug detection,” the CTO said. “We’ve since quickly realized value.”
Value realizers push harder
Despite these generally encouraging statistics, there are significant differences when we separate out the value realizers, which are defined as those that have adopted gen AI in at least one business function, develop their own proprietary models or significantly customize publicly available ones, and generate more than 5 percent of their earnings from gen AI. Ten of the 140 survey respondents reported working within organizations that fulfilled all three criteria. And these respondents indicated just how determined value realizers are in the pursuit of gen AI value. Consider the following:
- Strategy: Eighty percent of value realizers reported having an integrated road map to introduce priority gen AI uses. That compares with 48 percent of respondents from other organizations.
- Business functions: Value realizers apply gen AI to a broader range of functions. It is the value realizers, and only the value realizers, that identify gen AI value pools in functions such as risk management, legal and compliance, product and/or service development, and human resources. “Deploying gen AI into our product research and development has revolutionized our innovation process, helping us improve solutions while reducing production costs,” the head of R&D at an energy and materials company said.
- Performance measurement: With a gen AI strategy in place, value realizers are far more likely than others to track progress in implementing it. Seventy percent of respondents from value realizers said their company had a system in place to measure and track key performance indicators, compared with 39 percent of other respondents. Such systems help companies adjust resources as needed across business functions in pursuit of value.
- External partnerships: Value realizers are more likely to partner with external service providers to bolster their efforts as they begin to adopt and scale gen AI. Eighty percent of respondents from the value realizers said such partnerships were in place, compared with 59 percent of others.
Value realizers’ drive might be explained by the fact that they are concentrated in sectors with some of the biggest potential value gains—energy and materials, financial services, and advanced manufacturing—giving them more of an incentive to adopt the technology. But whatever the impetus, their efforts appear to be reaping rewards.
The capabilities that matter
While many organizations have initiated gen AI projects, the survey also suggests that many might struggle to scale and capture value from them unless they undertake a fundamental rewiring of the organization.
While it is clear that many GCC organizations are keen to harness gen AI’s potential, most remain at an early stage of adoption. Embedding gen AI at scale will require a fundamental rewiring of the organization. McKinsey research shows this entails mastering certain capabilities in five areas: technology, data, talent, the operating model, and risk management. With this in mind, we assessed how well prepared GCC organizations are to implement their gen AI strategies. The results suggest they have a considerable way to go, while those already realizing value shed light on the path to follow.
Technology
Organizations have three model options when it comes to implementing gen AI. They can use off-the-shelf, publicly available models; they can customize these tools with proprietary data and systems; or they can (at much greater expense) develop their own foundational models from scratch. The choice can be dictated by a range of factors such as budget, security concerns, language needs, and the value that any organization might reap from the model. There is a growing view, however, that we are moving from a binary world of “build versus buy” to one that might be better characterized as “buy, build, and customize,” where the most successful organizations are those that construct ecosystems that blend proprietary, off-the-shelf, and open-source models.
In the GCC, a greater proportion of organizations customize models or develop their own compared with the global cohort, according to survey responses (Exhibit 4). Interviewees suggested this could be because many GCC organizations have been relatively slow to adopt gen AI but are now keen to catch up by learning from the experience of leading global organizations, which have proven the value of customization and proprietary models.
Whatever the models chosen, however, several practices help organizations create value from them.
- Defined processes that continually update and improve models. A company deploying gen AI into its customer care capabilities, for example, needs to regularly update the chatbot’s training data to make sure it meets user needs and can better handle complex inquiries.
- Technology and tooling that optimize development workflows and machine learning production. The streamlining and automation of repetitive tasks means models can be deployed faster with fewer errors, while giving developers the freedom to focus on higher-value tasks.
- Continuous testing and validation embedded in the model release process. This helps ensure that bugs and other issues are caught and addressed early and that user needs are met, which can lead to faster delivery of high-quality products.
In all these areas, value realizers lead the field (Exhibit 5).
Data
Data management is pivotal to scaling gen AI within an organization. Good data management includes the following:
- A comprehensive data strategy. To pursue their gen AI plans, organizations need to create a detailed strategy to acquire the necessary data and transform it into a usable format, and then find the right people to implement such a strategy.
- A centralized data repository. A centralized repository provides a single source of data that informs all models. The repository makes it easier to combine and process data, and all decisions are made from a consistent, single source of truth.
- The development of modular components. Modular model components—the data pipeline or feature selection, for example—function independently and so can be reused in other models.
Value realizers are stronger in all three areas than their peers, according to respondents (Exhibit 6).
Interviewees make clear just how much work many companies will have to do to improve their data management, as data is often stored in different silos across the organization, different business functions use different tools and platforms, and it’s unclear who owns what data. The CTO from the education organization, for instance, noted that data from the 20 to 25 apps it was running across its operations was sitting in so many different silos that it was hard to draw any meaningful insights from it or collaborate across functions. “Hence the decision to build a centralized platform to consolidate our data, which meant first identifying the data we had and then structuring it,” the CTO said. The goal, the CTO explained, was to create a single interface that benefited all users. Parents, for example, can use a single app to register students, pay fees, or see students’ course work.
Yet GCC organizations face other data challenges too. Data privacy regulations sometimes dictate where and how data should be stored—particularly information about individuals or data that might have national security implications. This can make it difficult for organizations to bring all their data together and therefore analyze it effectively. And for the time being, at least, cloud service providers are not well established throughout the region. This can pose both technical and economic barriers to organizations that want to adopt gen AI but that manage their own IT infrastructure, as leveraging large sets of data and training gen AI models requires significant tooling and computing resources. Cloud providers provide off-the-shelf solutions.
Talent
Talent is one of the most critical capabilities when it comes to deploying gen AI effectively within an organization. Organizations could benefit from the following:
- A dedicated gen AI leader. The appointment of a single senior person (someone from the C-suite or board) to propel gen AI initiatives is a well-recognized means of speeding its adoption. Interviewees point out that it ensures both ownership of the task and accountability.
- A well-defined talent strategy. Organizations need people across functions with relevant technical skills, but recruiting and retaining that talent is a major concern for many GCC organizations. Hence the need for a well-crafted talent strategy—one that not only offers competitive benefits but growth opportunities too, crafting for each employee a personalized development program.
On both counts, value realizers lead the way by a considerable margin (Exhibit 7).
Regarding talent recruitment and attraction, it seems that even a good talent strategy might not suffice, given the ongoing need to grow tech talent in the GCC region. Overall, 48 percent of those who said their company had a talent strategy in place also noted that recruitment and retention remained a challenge, though that figure fell to 20 percent among respondents from value realizer organizations. The CIO of one large energy company that offers sizable incentives and invests heavily in training and development programs said he still faces the threat of talent being poached.
The size of the talent challenge differs by country and region and, as such, governments are considering various initiatives. Saudi Arabia and the United Arab Emirates (UAE), for example, have both launched programs to train workers on gen AI. Regardless, organizations could benefit from following the example of some of the value realizers that conduct extensive programs to teach their current technical and business employees gen AI skills. Such programs can have added benefits. The change management leader of a large UAE organization said running weekly workshops to educate employees about gen AI technology had not only helped to develop gen AI skills but also demonstrated how the technology complemented attendees’ work. That, he said, had proved pivotal in building enthusiasm for adopting gen AI.
The operating model
A center of excellence, organizational agility, and change management programs can be key enablers when it comes to scaling gen AI initiatives.
- Center of excellence: A center of excellence, where a single, centralized team works to coordinate the implementation of gen AI across the whole organization, makes sense for organizations in the process of building their capabilities. It helps focus resources on key use cases and drive the transition in an organized way from initial experimentation to production and, ultimately, widescale adoption. A critical task for this central team will be to develop and establish protocols and standards to support scale, ensuring that teams can access models while also minimizing risk and containing costs.
- Agility: Well-defined agile processes and standards promote collaboration, iterative improvements, and ultimately higher productivity when it comes to delivering gen AI solutions, though funding and budgeting processes need to be agile too to keep pace.
- Change management: Most respondents recognize that to scale and capture value from gen AI they will have to win the support of the workforce, which likely means implementing a change management program. A powerful change management program often includes a communication plan that conveys to everyone the importance of gen AI, offers skill-building opportunities, and ensures role models are demonstrating the new ways of working.
Value realizers are considerably more likely than other organizations to have adopted these features in their operating models, according to the survey results (Exhibit 8).
Risk management
While GCC organizations recognize the benefits of gen AI, they’re also aware of the associated risks—data management risks such as breaching privacy, bias, and intellectual property (IP) infringement; model management risks such as inaccurate output or lack of explainability; and cybersecurity security and incorrect use risks.
Of these, cybersecurity was identified as the most relevant risk to their organization by the highest share of respondents (66 percent), followed by regulatory compliance (52 percent) and inaccuracy (45 percent) (Exhibit 9). Respondents in the global survey ranked risks similarly.
Despite this risk awareness, respondents indicated that many organizations have not yet implemented the kind of practices needed to scale AI responsibly. These include the following:
- Review processes: Clear processes are in place to review and mitigate risk as gen AI solutions are developed.
- Responsible AI: An enterprise-wide body has the authority to make decisions regarding responsible AI. It ensures gen AI models meet ethical standards and comply with regulations and guidelines, and that governance makes clear who is accountable for managing risk.
- Skills: Knowledge of gen AI risks and mitigation controls is mandatory for technical talent.
- Model design: Gen AI models are designed to incorporate audits, check biases, and assess risk through the model’s life cycle.
Value realizers are more likely than others to follow all these practices, according to survey respondents (Exhibit 10). Their apparent higher sensitivity to gen AI risks and their mitigation may well be because they have directly experienced risks, having used gen AI more extensively. Whatever the cause, other organizations might consider taking mitigation action from the outset.
The survey suggests that all organizations in the GCC have scope to begin extracting greater rewards from gen AI. Only ten of the 140 survey respondents said their organizations were generating more than 5 percent of their earnings from the technology, leaving a sizable majority with substantial opportunities for growth. Yet even these organizations have scope for significant improvements. In our global survey, some respondents reported that gen AI accounted for more than 10 percent of their organization’s earnings.
What sets the best performers apart is that value creation has become the name of the game, replacing the tendency to run pilots that primarily focus on proving a concept. “My business has executed hundreds of gen AI use cases over the last few years, claiming gazillions in impact. But I still can’t see that on my P&L,” is how an executive at a UAE utility described the current state of play at his company.
In GCC organizations, maximizing value creation means deploying gen AI more broadly. (Even value realizers apply gen AI in no more than two business functions, according to survey respondents.) But it is not just a case of using the technology to transform specific process points. Rather, maximum value lies in reimagining complete workflows—considering how a family of use cases can bring about an entirely new and more productive end state within a function when integrated alongside other AI applications and new ways of working.
Scaling and integrating gen AI in this way is a much harder undertaking than adopting individual use cases. It likely requires a transformation of the organization’s capabilities—improving technology and data and risk management and strengthening operating models and talent strategies. A few GCC organizations have made a start, according to our survey results. Others have much further to go. And while it is important to bear in mind that realizing full value from gen AI investments won’t be instantaneous, the longer it takes to embark on such a transformation, the longer the rewards are likely to be elusive.
