Luxembourg, with its digitally advanced, globally connected economy,1 is well positioned to benefit from AI as the technology continues to redefine the global economic landscape.
McKinsey estimates suggest that AI could add $17 trillion to $26 trillion to the global economy.2 The European Union accounts for approximately 15.0 percent of global GDP,3 of which Luxembourg makes up 0.5 percent,4 indicating that Luxembourg could be poised to capture $13 billion to $19 billion of that added value.
The paradigm-shifting potential of agentic AI
In a recent report, McKinsey highlighted a “gen AI paradox” in which nearly eight in ten companies report using gen AI while a similar proportion report no significant bottom-line impact.1
AI agents offer a way to break out of this paradox thanks to their potential to automate complex business processes and thereby shift gen AI from being a reactive tool (such as a chatbot) to becoming a goal-driven and proactive virtual collaborator.
An AI agent is a piece of software that has the agency to act on behalf of a system or user to perform a range of tasks. While basic versions of this technology have existed for years, cutting-edge capabilities for natural language processing now enable systems of agents to plan and collaborate, opening up a broad variety of use cases. Agents can now take on routine tasks so humans can focus on higher-value work, but they can also transform processes by accelerating execution, enabling personalization, and increasing adaptability, operational elasticity, and resilience.2
Unlocking the full potential of agentic AI will, however, require reorganizing existing processes significantly. Instead of the scattered initiatives that have characterized many organizations’ experimental approaches to gen AI, a holistic AI transformation approach will be required.3 Organizations will need to reimagine workflows with agents at the core and will need an agentic AI mesh4—a new AI architecture paradigm—for effective governance.
While realizing the potential of agentic AI will require significant investment, these tools could shift paradigms at many organizations. Instead of looking at individual AI models as tools to assist their employees, organizations can now have these employees supervise digital factories of agents (exhibit). The impact on productivity can be huge: AI agents can already double an individual’s productivity, and—based on our internal, illustrative estimates—the use of digital factories could improve productivity 20-fold.
In addition, these already powerful technologies continue to develop extremely quickly. AI agents can already support highly complex and ambiguous use cases across industries and business functions, and in the very near future, AI agents could complete tasks as complex and disparate as loan underwriting, code documentation and modernization, and online marketing campaigns.5
The rapid advance in new techniques such as agentic AI has also made it easier to deploy AI—and capture the value at stake—than was previously the case (see sidebar, “The paradigm-shifting potential of agentic AI”).5 Through agentic AI, a single human employee can supervise a digital factory of AI agents performing tasks on the human’s behalf, significantly increasing productivity. This new type of operating model may be particularly advantageous for Luxembourg given the current labor shortages and skills mismatch in the country.6 Even without agentic AI’s potential, AI more broadly has significant potential to increase labor productivity within the Luxembourgish economy. Depending on how quickly technologies are adopted and the extent to which worker time is redeployed to other activities, gen AI could enable labor productivity growth of 0.1 to 0.6 percent annually in the run-up to 2040. Work automation from combining gen AI with other technologies could add 0.5 to 3.4 percentage points annually to productivity growth.7
Gen AI as a whole could also have a net-positive impact on labor demand; the World Economic Forum has noted that trends in AI are expected to create 11 million jobs, while displacing nine million others.8 This finding reinforces the fact that Luxembourgish companies must continue to work to attract top global talent. Given the likely significant upheaval in the labor market, it is also important to note that a portion of Luxembourgish workers will need support in learning new skills and, in some cases, changing occupations.
There is clear economic opportunity for all sectors in Luxembourg, but our research suggests that there may be particularly high-impact AI investment opportunities in financial services and across public services and social infrastructure, as well as within the information and telecommunications industry. This article lays out priorities and opportunities for these high-potential sectors and provides real-world examples of AI applications that are already unlocking value at scale across banking, healthcare, spacecraft communications, and IT.
To secure a competitive edge, Luxembourgish stakeholders in the public and private sectors could benefit from acting quickly and decisively as well as from fundamentally shifting their mindsets to treat AI as core to how their organizations create value. This article concludes with a practical blueprint that can enable organizations to rewire their digital transformation and integrate AI across their operations. Those that can build the skills, agility, and resilience to distinguish themselves as AI market leaders can develop a significant source of long-term competitive advantage, accelerating innovation and driving economic growth.
Identifying sector-specific AI priorities and opportunities
As already indicated, the widespread adoption of AI may have particular benefits for Luxembourg given the comparatively small size of the market, ongoing issues with labor shortages,9 and existing strong digital infrastructure.10
Luxembourg’s new AI strategy—“Accelerating digital sovereignty 2030”—recognizes these opportunities and identifies financial services and public administration as high-priority sectors in which AI could bring significant changes for organizations and offer substantial overall economic impact for the country.11
Our research, which considers the size of individual sectors within the Luxembourg economy and the likely impact of AI on those sectors, finds that financial services and public administration should be high-priority sectors for AI investments in Luxembourg. The research also suggests that defense, education, human health, retail, and social work may offer promising short-term AI investment opportunities (Exhibit 1). These sectors might also benefit from the development of specialized AI centers of excellence.12
This analysis also points to a number of additional sectors—including telecommunications—that make up a smaller share of GDP but are areas in which AI could have a substantial impact. Given strategic investments, those sectors could grow to become champions and growth engines for the Luxembourgish economy in the long term.
Testing high-impact AI use cases across priority Luxembourgish sectors
AI is already transforming industries around the world, including within the sectors identified here as high priorities for Luxembourg. This section provides four real-world examples of high-impact AI use cases in relevant sectors. These use cases illustrate how decision-makers in Luxembourg can use existing and developing AI technologies to establish a globally competitive market position and stimulate growth.
IT in banking: Modernizing legacy software development with a digital factory of agents
A large bank needed to modernize a corporate online banking application that was more than 20 years old and consisted of 2.5 million lines of code. The end-to-end modernization effort required an overhaul of everything from the delivery channel and software architecture to front-end and back-end connections.
Over the course of four months, a team of three engineers built a digital factory that contained more than 100 AI agents to modernize the first module of the legacy application. This modernization had three major steps:
- Reverse engineering. To understand why the processes and systems existed as they did, the digital factory documented the code, architecture, and business processes by scanning the legacy code, interviewing business and technical subject matter experts (SMEs), and consulting existing documents.
- Shaping. Under human supervision, the digital factory defined the target architecture to be in line with the bank’s key standards by reading documentation and understanding the legacy app, learning the bank’s requirements, and interviewing architects and other SMEs.
- Modernizing. The digital factory modernized the first module, including developing a fully functional application. Critical activities included reengineering the legacy code based on outputs from previous phases. A detailed understanding of design specifications and the target-state architecture were used to create a user-friendly front end and write the code required to connect to the back end. The bank’s own standards and guidelines were used throughout.
Overall, use of the digital factory reduced the time needed to modernize this first module by 35 percent. A full-scope modernization of the application would result in even larger efficiency savings—estimated at a 70 percent reduction in total hours—because it could reuse what the digital factory of agents had already created. In addition to decreasing overall costs, using agentic AI at scale would enable the bank to shift from vendor dependency in its IT environment to internal capability building.
This use case illustrates the potential of agentic AI to enable change and transformation at scale. This potential is not specific to one industry or back-office function, however; these technologies are already being developed and deployed across manufacturing and operations, customer interactions, and beyond.
Public administration: Accelerating the identification of anomalies in social insurance contributions
A government department responsible for auditing social insurance contributions was juggling scarce resources at a time when a demographic shift in the workforce was reducing the time available for each individual audit.
The department implemented AI models to assist auditors by scoring cases and adding hints for audits. The model performed three important tasks:
- Prioritizing audits. A sophisticated AI model was built and trained to produce a score for each case, which gave insights into the likelihood that the case might contain an anomaly. Based on this likelihood, each case was given an overall score. This score was integrated directly into the auditing systems and used to prioritize the cases.
- Guiding auditors with case-specific hints. While the audit was still entirely in the hands of the auditors, the AI model provided them with informative hints on where anomalies might be expected.
- Continuous model improvement through integrated feedback. The model allowed auditors to offer feedback on its accuracy and usefulness for each individual case. This feedback was used to continually improve the model.
The model is expected to mitigate the staff shortage over the coming years, supporting auditors in their daily work. This use case illustrates the potential for AI to improve efficiency and service levels in the public sector.
Banking: Hyperpersonalization through at-scale gen AI delivery
Another large bank was looking to modernize its systems as a way to stand out in a competitive market and keep up within an increasingly complex regulatory environment. Recognizing the potential of gen AI to meet those goals, bank leaders chose marketing hyperpersonalization as one of five strategic domains in which to roll out gen AI programs.
Over the course of 12 weeks—bringing in external help when needed—the bank built a gen AI copilot for marketeers and created and ran five campaigns across two markets. This copilot provided the following key features and characteristics:
- An end-to-end solution, executed by marketeers. With the gen AI copilot, the bank’s marketeers can add customers to subaudiences, generate personalized content for each subaudience, and send the content to the customers.
- Hyperpersonalization by channel, timing, pricing, content, and language. The copilot helped ensure that customers could be targeted at the right time and in the right way, without being overloaded:
- • Models enable the bank to identify the customers most likely to be interested in a product as well as the events that might make drawing attention to the product timely. This logic decreases both campaign costs and unsubscribe rates.
- Pricing models provide a personalized commercial offer while ensuring balance between conversion rates and margin.
- • A hyperpersonalization engine customizes messaging for each customer to maximize conversion by explaining how the product solves that specific customer’s needs. Tone of voice and sentence length are also adjusted to appeal to different audiences.
- Personalized messaging can be disseminated through emails, in-app banners, and the bank’s landing page.
- Easy integration with existing systems and scalability. The gen AI copilot was designed to fit into existing campaign processes, including integrating with local marketing stacks. Initial campaigns ran in two markets, but the solution architecture was built to be scalable, enabling rapid future deployments across other markets.
After just 12 weeks, more than 2.2 million customers had been reached with messages created by gen AI, with an uplift on click-through from the app to the start of the sales funnel of up to 250 percent. The campaigns led to an increase in conversion rates of up to 20 percent overall, with significantly higher rates (more than 85 percent) among Gen Z audiences.
These results illustrate the potential for gen AI to support hyperpersonalization in customer service operations across industries, enabling frontline teams to serve customers better, simplifying internal processes, and reducing costs.
Telecommunications: Revolutionizing spacecraft communications through AI-driven solutions
A top spacecraft communications company operated ground stations that communicated with spacecraft as they passed overhead, sending telecommands and receiving housekeeping and scientific telemetry. The ground stations experienced occasional communication failures that disrupted data flows, slowing down the company’s operations process. Reasons for failures often required manual intervention—including searching for information buried in computer logs—which made taking action to prevent or rectify these issues highly challenging.
The company used AI-driven solutions to understand the root causes of these failures, facilitate solutions, and decrease their impact. The process encompassed three steps:
- Improve data accuracy. The company used advanced machine learning models to identify issues in the station’s data sets. It then used applied statistical modeling to address the issues (such as getting rid of data noise) and produce a higher-quality, labeled data set.
- Predict risk. Using the labels from the station logs, combined with weather and telemetry data, a machine learning model determined whether failure was likely to occur during the time that any individual satellite spent in the range of a ground station as it passed overhead.
- Identify root causes. AI models were used to determine where the failures occurred and identify their root causes, with results presented in a digital app. This information was used to address issues quickly and—where possible—prevent them from occurring in the future.
Overall, the engineering team’s efficiency increased by about 30 percent from the baseline estimate.
While this use case highlights the significant impact that AI models can have in proactively troubleshooting issues in advanced industries, agentic AI could further propel efficiencies in these and similar situations.
Embedding a rewired enterprise mindset
These use cases illustrate the potential transformative power of AI across sectors for Luxembourgish businesses and the country’s economy as a whole. However, capturing this value requires more than simply implementing a system or acquiring a technology. Instead, success is likely to require hundreds of technology-driven solutions that work together—as well as the systems, skills, capabilities, and mindsets to deploy them effectively—to fundamentally rewire how a company operates.
Based on McKinsey research and experience, six key enterprise capabilities can rewire organizations to potentially outperform in the dawning AI age (Exhibit 2).13
While these six capabilities touch every part of an organization, our experience suggests that no digital and AI transformation can be successful without building at least a baseline of competence across all six areas.14 While all capabilities will require careful attention, it is worth highlighting a number of specificities related to the context in Luxembourg.
First—as mentioned above—talent is a particular issue in Luxembourg, stemming from overall labor shortages, including in priority areas.15 Undertaking a holistic AI transformation thus is likely to present both a challenge and an opportunity for Luxembourgish companies. On the one hand, labor shortages may make it particularly challenging for companies to ensure that they have the AI-related skills and capabilities they need to innovate and execute. Attracting and retaining talent should therefore be a top priority.16 On the other hand, a successful transformation can substantially increase labor productivity, as we have seen, which may help to alleviate labor shortages in the medium to long term.
The details of the required technology transformation may also look different for many Luxembourgish businesses compared with other markets. Luxembourgish businesses may be more likely to be “takers” (adopting ready-made AI solutions to quickly capitalize on proven technologies with minimum customization) or “shapers” (enriching pretrained models with proprietary data sources), rather than “makers” (developing entirely new AI solutions).17 Regardless of the solution chosen, having the right operating model in place will be vital for success.
The updates required to build an operating model that will enable companies to successfully deploy AI may be particularly beneficial in the Luxembourg context. A sticking point for many transformations is that initial AI pilots may be too small and therefore do not deliver the impact that might create the momentum needed to get to scale. While this issue will exist in Luxembourg, as it does elsewhere, the value at stake in reaching scale may be particularly substantial. Companies that can get this right and roll out high-impact, scalable AI solutions may be well on their way to solving the scale-related issues that have sometimes limited the growth of businesses operating in the comparatively small local market.
Throughout the transformation, appropriate investments in change management will be crucial. Our experience suggests that when managing gen AI costs, companies will need to spend about $3 on change management for every $1 spent on developing a model.18
While companies need to be aware of these Luxembourg-specific context factors, they should not lose sight of the fact that the six enterprise capabilities are also interconnected, and they must be treated as such. This interconnectivity is why companies need to think in terms of a holistic AI transformation—or rewiring—rather than focusing on a few narrower initiatives. As such, a successful AI transformation will require significant time and effort, and an unparalleled level of collaboration across the C-suite.
The value at stake is significant. Our research shows that digital leaders—those that have succeeded in fully rewiring their organizations—were able to improve their return on tangible equity and TSR significantly more than those who lagged behind on digitalization.19 In effect, digital excellence can translate into financial outperformance.
Luxembourg maintains real strengths across several industries, including financial services, communications, and logistics. However, the small local market makes it harder to compete through scale effects, and difficulties in reaching scale are further compounded by ongoing talent shortages, including in critical sectors.20 To ensure that the country continues to outperform economically, leaders across the public and private sectors should consider fully integrating AI across key processes—and quickly. By doing so, Luxembourgish organizations will be able to retain, enhance, and cement existing competitive advantages, and grow the economy as a whole.
Related Articles
The next big arenas of competition
