Artificial intelligence: how leading companies define use cases, scale-up utilization, and realize value

One of the largest Airline Carriers worldwide took a deliberate approach to reap AI benefits. Airline operations are complex and expensive, often consisting of more than 150 aircraft to coordinate, over 700 departures to organize per day, and more than 30 million passengers to transport per annum. A companies’ success largely depends on creating and synchronizing plane and crew schedules, maintaining aircrafts, and reacting to disruptions such as unforeseen extreme weather, which may cause passenger delays and require flight schedule and booking changes [11]. Costs increase with the number of flights that have to be rescheduled or canceled due to unexpected disturbances. Hence, the carrier investigated how AI can help reduce high operations complexity and costs.

Step one for the airline carrier was to identify the most relevant pain points for the business. Consequently, leadership teams of each business area involved were interviewed. Step two was to assess the expected value AI may unlock and the time required to implement the AI solution per pain point. For instance, the value potential and needed time to optimize the schedule adaptation with AI were assessed. The implementation time is determined by the availability and quality of necessary data and skills, technological maturity, and ease of implementation.

After step two, three priority projects were identified and endorsed by top management: (i) optimizing flight scheduling dynamically, (ii) crew scheduling optimization to deal with disruptions, and (iii) dynamically handling day-to-day disruptions on the day of operations. Cross-functional teams then developed and implemented solution concepts leveraging AI for each priority topic.

As a result, the airline could reduce its non-performance costs, i.e., costs accruing from service delivery failures, by more than 25% compared to before. Moreover, unexpected delays impacted approximately 30% fewer passengers due to intelligent handling of disruptions. These results were obtained by ensuring that frontline professionals used the new AI solution as part of their daily working routines and provided ongoing feedback on new releases and features. Additionally, based on its own experience, the carrier identified a growing demand for optimizing flight scheduling and operations in the industry. Since there were no previous offerings available in the market, the airline expanded its service portfolio by launching and marketing a new business line for improving other airlines’ operations.

Employees need an enriched skill set to work in organizations enabled or supported by AI. As AI is likely to overtake standardized automatable tasks, humans need to contribute “people skills” like judgment, creativity, or communication ability [2]. Due to the specificity and complexity of AI, Beck et al. [3] emphasize the importance of dedicated AI-related roles to exploit its potential. The authors provide examples such as an “AI engineer” who develops AI products, is savvy in enterprise architecture, and integrates AI into existing systems. Another example relates to complementary data capabilities that “fuel” AI systems. The “Data Czar” offers sophisticated business expertise to source, clean, and transform the data needed to run the system and overcome the particular business problems involved.

As research shows, these capabilities are often not available. According to Brock and Wangenheim [1], more than 3500 answers across over 1200 companies have described three key challenges in implementing digital technologies and AI. “Lack of skilled staff and knowledge in digital technologies” was mentioned in 54% of cases, which is twice as frequent as the following challenges “lack of organizational agility” (27%), “lack of sufficient funding” (26%), or less relevant technology inhibitors.

According to our study, organizations differ enormously concerning the digital capabilities of their workforce. Almost twice as many digital champions as laggards (44% vs. 23%) already dedicate a significant proportion (over 15%) of their full-time equivalents (FTE) to digital roles and responsibilities. When asked how many people in the entire workforce shall undergo re- or upskilling in the coming year, 51% of the champions reported “over 15%.” In comparison, only 28% of the laggards provided the same response (Fig. 3). Besides, firms across Asia report far higher ambitions for hiring and upskilling digital talent than comparable organizations in Europe and the US.

Next to the “what,” i.e., the available workforce to deliver digital projects, the “how” can be considered equally important. Successful companies employ agile ways of effectively developing minimum viable products jointly in cross-functional business and IT teams and converting them into operative solutions. As indicated in Fig. 3, 35% of the champions state that more than 25% of joint business-IT projects apply agile techniques of working, a factor of almost two compared to laggards who apply agile techniques in 18% of all cases on the respective level.

To conclude, people dedication and hiring, up- and reskilling the entire workforce, and working in agile ways are key success factors to scale-up AI to achieve better outcomes and realize value benefits.

Technology is the second enabler for scaling the utilization of AI successfully. It includes an advanced technology function, state-of-the-art cybersecurity capabilities, and a digital and data platform⁴ as the data-centric organization’s foundation. The latter is described in more detail later in this section.

Companies classified as digital champions invest a significant share of their operating expenses (OPEX) into digital initiatives. More than half of the firms in this category (55%) state they invest more than 15% into such initiatives compared to only 29% of digital laggards (Fig. 3). Of this spending ratio, champions spend on average 50% of the digital budget on technology and data, compared to only 38% of the laggards. Thus, champions invest more into digital initiatives, mainly digital technology and data.

Higher investments provide the basis for faster innovation in technology and data. Nearly twice as many digital champions report that 25% or more of their applications are application programming interface (API)-ready in comparison to the laggards (40% vs. 22%). Similarly, almost twice as many champions as laggards have been able to organize at least 25% of their data under the architecture of a unified data model, which increases consistency, improves reusability, and speeds up access to all data in one place. Such a data model unifies data from different sources and platforms in one location to be considered when conducting analyses and making decisions. Champions seem to be striving more intensively towards becoming a data-centric organization.

To understand what differentiates AI leaders (stage 4) from AI starters (stage 1), we further analyzed the group of organizations that show the highest level of maturity in AI and that have been able to generate substantial value from AI, measured by the dimension “Artificial intelligence” in the DAI framework (See Appendix, item “DAI dimension Artificial intelligence”). AI leaders agreed to the statement: “AI is successfully adopted in core offerings and processes, providing competitive advantage. AI is rolled out across the whole value chain and shapes new business models. AI is key to decision-making in the organization.” In comparison, AI starters chose the lowest maturity stage: “Basic understanding of AI in parts of the organization but neither larger adoption nor use case prioritization. No AI job profiles exist, and the common perception is that existing analytics experts like statisticians could cover AI.”

AI leaders have successfully set up a data-centric organization and use data analytics to rationalize and improve decisions. In line with the DAI framework, three dimensions compose a data-centric organization: data strategy, data governance, and digital and data platforms. Between 77% and 79% of AI leaders are already in the two highest maturity stages for the three respective dimensions. In comparison, and to take an extreme view, of those organizations that chose either of the lower two stages in the question to measure their AI maturity, i.e., digital starter or digital literate, only 48% reached the highest maturity in data strategy and only 39% in data governance. In particular, the infrastructure, a digital and data platform, is a challenge for these organizations as only 6% have reported that they have reached the highest maturity stage here.

So how do AI leaders differ from AI starters? First, AI leaders pursue a formulated data strategy. They recognize data as a critical corporate asset and have various automated, scaled, and commercialized analytical solutions. They regularly measure how to exploit value from the existing data assets and have defined performance targets to reach from commercializing data. A transparent operating model covers all data-related issues [7].

Second, AI leaders have established data governance. They have assigned essential data and analytics roles, implemented data quality policies, perform master data management, and provide methods and tools to harmonize and link data from distinct sources. Everything operates in a hub-and-spoke model [14]), fueling the entire organization with the help of data warehouses and data lakes.

Third, AI leaders have also built-up the necessary technology infrastructure to enable data integration and support focused development of analytics upon required data [13]. In the DAI methodology, this capability is named “digital and data platform,” and AI leaders have implemented such a fully optimized platform based on best-in-class technologies, e.g., a Hadoop ecosystem [12] and using one or more data lakes in full production. An overview of these three differentiators is provided in Fig. 4.

Pioneering companies aim for AI systems that are fully integrated into day-to-day operations, i.e., in the production environment, instead of small pilots. For instance, a leading supplier for gold mining operators faced saturated customer demand in a consolidated market. The business’s core was to manufacture and provide cyanide, a special chemical in a market segment with few growth opportunities. Therefore, the company’s ambition was to build a new business model enabled by AI to support clients beyond a chemical substance’s mere offering.

The companies’ strategy was to look for new opportunities that help mining operators improve in gold recovery. First, the company started by gathering data from historic production cycles for relevant process parameters (e.g., oxygen in the tank, throughput per hour, leach feed grade, and cyanide concentration) and the resulting gold recovery. Data scientists then analyzed the data to identify potential gold recovery improvement levers. Various machine learning algorithms were deployed and tested. However, despite sophisticated algorithms and the breadth of data, no significant improvement levers were identified initially. Hence, the chemicals supplier started to involve metallurgical experts to refine the analyses and account for the gold recovery process’s key steps and corresponding variables. Combining chemical and metallurgical process know-how with collected data allowed the company to build a “digital twin,” meaning a virtual model, of the gold recovery process. This new model used the data for the process parameters to explain, predict, and control the resulting gold recovery. In fact, it explained 60% of the overall variance of the selected input parameters.

With the help of the “digital twin,” the company was able to identify four key improvement levers. Gold mining operators can improve their operations by optimizing (i) dissolved oxygen, (ii) throughput levels, (iii) feed blend approach, and (iv) cyanide concentration. As a result, miners’ earnings before interests, taxes, depreciation and amortization (EBITDA) has been able to increase by 10–15%. Of this effect, 80% accrued from a higher recovery rate and translated into top-line growth, whereas 20% came from cost savings. This value creation evidence was obtained in less than one month, including developing the “digital twin,” simulating the process variables, and deriving the business insights. Afterward, the chemicals provider prepared a comprehensive business model to actively market the value proposition of AI technology for gold recovery processes, including the required operating model. In other words, the developed capabilities and agile collaboration between chemical, metallurgic, and data experts empowered the Chemical Firm to transform from a chemical commodity provider into a digitally enabled operator of highly specific processes.

Organizations that take on a business lens, build fundamental enablers, and measure AI, achieve tangible impact, which is notably higher than those that do not. We included indicators to analyze how digital champions perform against laggards (Fig. 5). Some measures report shareholder value from external sources⁵, e.g., market capitalization, total enterprise value. Others offer self-reported indicators, e.g., earnings before interests and taxes (EBIT) from digital, the magnitude of scale (Appendix, item “Scale”). To assess a firm’s competitiveness, the respondent evaluated the perception of where the organization stands compared to industry peers over the past 3 years (Appendix, item “Competitiveness”). It is not surprising that both digital champions and AI leaders outperform other organizations in our sample, but the magnitude strikes. About twice as many champions as laggards (64% vs. 34%) consider their firm to be more competitive than industry peers, i.e., better in cost-efficiency, product quality, time-to-market, and customer satisfaction. Digital champions are growing stronger in total enterprise value every year and experience higher EBIT impact from digital investments. Furthermore, they are better at scaling-up their digital pilot projects into operating solutions that deliver the full potential of the technologies behind them.

Moreover, the Covid-19 crisis has accelerated many companies’ digital transformation as the value of digital capabilities in responding to and recovering from the crisis became evident. A closer look into the total shareholder return of champions vs. laggards from February 2020 to January 2021 shows that digital drives resilience. Champions’ market capitalization is not only back to its pre-crisis level but has grown by 23%. In contrast, laggards stood at plus 7%, so 16 percentage points lower (Fig. 6).

Companies invest in AI to generate tangible value. An example is a European energy firm with business activities in natural gas supply, petroleum products exploration, combined heat and power (CHP), and renewable energies. The company aimed at reaping the potential by moving from efficiency-centric operations to innovation-centric operations enabled by AI. The ambition was to deliver on the value promise of AI by building use cases, changing the firm’s operating model, and setting up new forms of data governance and comprehensive digital and data platforms. For instance, digital collaboration tools and agile collaboration processes were introduced to foster upstream coordination and joint innovation with suppliers. Energy production was advanced by more sophisticated ways of asset management for improving safety and field productivity, as well as advanced practices for energy trading. Downstream activities were innovated through personalized sales approaches, non-fuel sales improvements, renewables production forecasting, cross-selling identification, and digitally enabled energy services.

The company identified and prioritized potential AI use cases by expected value and speed. For selected use cases such as predictive maintenance, safety, and turbine productivity, cross-functional teams were formed combining business experts, data scientists, and IT architects. These teams worked agile and started to collect the required data from existing and newly deployed Internet-of-Things devices and sensors (e.g., process data on turbine lube oil temperatures and performance). To implement the use cases, the company followed a four-step approach.

First, machine learning algorithms were trained with existing data to develop “local” models of lube oil temperature for each turbine. Second, differences between modeled and actual lube oil temperatures were identified, indicating potential anomalies, i.e., potential upcoming problems. Third, early maintenance was performed on turbines with abnormalities to preempt possible shutdowns and system downtimes. Last, the maintenance findings and root causes, such as faulty compressor units, were fed back to the machine learning prediction model to further refine the identification of abnormality causes. To accelerate the scaling of this value-generating use case, the company pushed the data from legacy systems into a cloud-based data lake, implemented a shared, quality-assured data repository, and defined clear data governance roles.

As a result, the energy utility company successfully set up a new Integrated Operations Center, which fully enabled AI activities with over 100 employees. These employees combine their expertise and leverage AI tools in their day-to-day work. Production volumes have increased through higher capacity utilization by 3–4% based on innovation and redesign of upstream activities, energy production, and downstream activities. AI has enabled the energy utility to preempt breakdowns before their occurrence, increase daily production, and proactively reduce bottlenecks regarding installations.