AI-Driven Competitive Advantage Isn’t the Future—It’s Now

robots running in a race representing technology advancement

One still hears mainly about the potential of artificial intelligence (AI) technology, but it is currently being applied to practical effect across a range of industries. Further, the pipeline of initiatives already in motion suggests that the use of AI is advancing faster than many think.

These were some of the key insights gained from a comprehensive scan of how more than 90 leading companies in 15 different industries are using AI. The study, completed by A.T. Kearney, includes more than 100 distinct use cases. For each use case, we tracked the type of AI technology employed by the company, how the AI solution was developed, its current stage of development, where in the value chain AI is being applied, and the objective the company is pursuing by using AI.

Roughly 37 per cent of the use cases documented in our scan show AI being implemented at scale and delivering meaningful benefits. On average, these companies report a 15 to 30 per cent return on investment within two years. Another 49 per cent of the use cases we examined were in the prototype and production phases, suggesting that AI applications are poised for substantial near-term growth (see Figure 1).

We found artificial intelligence being applied for four main purposes:

  • Performance improvement focuses on optimizing processes and worker productivity and improving supply and demand forecasting.
  • Customer experience includes health and personal data monitoring to improve consumer decision making, as well as personalization of the customer journey or product offering.
  • Value generation uses AI to help companies identify new streams of company revenue, either through new product identification or new channel offerings.
  • Cost management includes predictive maintenance and identifying outliers that may be causing production disruptions or product defects.

Applying AI in pursuit of performance improvement accounts for nearly half of the use cases across all sectors. Applying AI in service of the customer experience constitutes just under one-quarter of cases and applying AI to value generation constitutes about 20 per cent. While less than 10 per cent of the studied use cases focus on cost management, a number of companies reported that this will become more of a priority for their AI applications going forward.

Overall, we found that the most AI uses already deployed at scale were within customer experience, while relatively more of the use cases in performance improvement and cost management were still in development (see Figure 2).

AI Successes

Our scan of over 90 companies in 15 industries identified numerous successful applications of AI. Here are a few examples:

Manufacturing. The robotics company FANUC as released its FIELD software to harness Internet of Things data from all parts of the shop floor to drive efficiencies. Nike has started installing Grabit robots, which has reduced the time to assemble a shoe from 10 minutes to 50–75 seconds. Using computer vision, Toshiba’s AI-powered image-processing techniques have cut the time required to determine the cause of defects in semiconductor wafers by two-thirds.

Retail. AI has improved the ability of several major retailers to predict demand and improve replenishment. At Germany’s Otto, for example, AI powered by Blue Yonder has reduced delivery times by 80 per cent and helped the company improve turnover: 90 per cent of goods ordered are now sold within 30 days. Tommy Hilfiger and India-based Myntra are using AI to help their designers create the season’s styles. In fact, Myntra’s AI-powered division is now its fastest growing brand.

Energy and Process Industries. GE is beginning to use its proprietary Predix software for the predictive maintenance of truck pumps, and expects to reduce unplanned machinery downtime by about 36 per cent and save almost US$10 million yearly.

Chemicals and Pharmaceuticals. BASF is using AI to model real-world environments (creating “digital twins”) and process historical data to determine where to focus experimentation, reducing the time required to achieve usable results from weeks to days. Similar tests are underway at major pharmaceutical companies such as Pfizer, which is using deep-learning techniques to train AI engines to detect new compounds that could lead to new life-saving drugs.

Banking and Insurance. To connect with millennials, Progressive has capitalized on the prominence of its Flo character to create a chatbot through Facebook. Thanks to the data the chatbot has collected and processed into learning, Flo is now capable of answering 15,000+ queries.

New Frontiers

Nearly three-quarters of the use cases we scanned are for supervised learning where AI algorithms learn from data and human experience to deliver clearly defined target outcomes. The same-day grocery delivery service Instacart, for example, uses personal shopper and customer data to train AI (Python code) to optimize shopper routing within grocery stores.

In unsupervised learning, the AI engine is given wider learning parameters (e.g., cluster analysis) and pursues more open-ended objectives. While unsupervised learning use cases arose less frequently in our scan, we found several intriguing efforts underway. ExxonMobil’s partnership with MIT’s Energy Initiative, for example, has the expansive goal of developing robots capable of “functioning as a scientist” while exploring deep-sea oil seeps. The project is definitely pushing toward new frontiers. In fact, the project requires unsupervised learning techniques that are still a few years from fruition. Nonetheless, other industry players are following suit by investing heavily in the potential of similar robots.

It Takes an Ecosystem

Developing complex AI algorithms in-house is time-consuming, costly, challenging, and capital-intensive. AI capabilities were developed in-house in only 33 per cent of the use cases we examined, primarily among companies that already had large R&D functions, such as J.P. Morgan, GE, Pfizer, BASF, and Amazon.

In most cases, companies have worked in partnership with various outside resources, including AI providers, academic institutions, and consulting firms.

Though highly fragmented, the AI ecosystem offers companies the opportunity to pursue a wide range of partnership models. Start-ups tend to be in the custom solution design space, focusing on industry-specific AI solutions. The major AI platform players tend to be mainstay data companies that work with companies and custom solution designers to implement offerings. Some platform providers (such as Microsoft, Salesforce, IBM, and Amazon) offer configurable application programming interfaces (APIs); however, companies must still provide the data to train the AI engine and customize it to meet specific goals or key performance indicators (KPIs).

Accelerating Momentum

AI is just now hitting its stride. Over the next three to five years, A.T. Kearney expects to see the use of artificial intelligence expand as more use cases that are in production reach scale. As costs are driven down both by increasing the demand and supply of AI-as-a-Service and customized solutions, the potential for optimization and unlocking new value is huge.

Perhaps the most important adjustment that leaders must now make is to shift from fearing AI as a disruptive force to more fully embracing AI’s demonstrated capacity to significantly improve performance, customer experience, value generation, and cost efficiency. The winners will be those who have been more proactive rather than reactive in embracing AI.

Decisive Action

In sum, AI is on pace to be not just an enabling technology, but a transformative one. Companies that fail to decisively pursue AI impact at scale, here and now, risk being left behind. Consider the following guiding principles:

  • Focus on key use cases that have clear benefits and, ideally, are also synergistic; i.e., success in one use case advances learning or amplifies the benefit of adjacent use cases.
  • Adopt a “how to enhance human experience and productivity” mindset, rather than a “how to automate” mindset. View your human workers as “customers” of AI. Ensure that embedded AI solutions are as frictionless and friendly as possible.
  • Do not let enterprise legacy IT be an excuse for holding back. Leverage ecosystems to jump start capabilities and assertively use case deployment.
  • Don’t be slowed by “not invented here” thinking. In many use cases, external cloud-based, pay-as-you-go AI systems can deliver major advances more quickly than traditional in-house enterprise solution development.

Leave a Reply

Please submit respectful comments only, including full name, professional title, and contact information (only name and title will be posted). Required fields are marked *