Transforming Businesses with AI: Insights from Rolls-Royce, Mastercard

January 1st, 1983, the day the internet was born. The obscure internet tool forever changed how people interact.

A decade later, the public embraced the internet, spawning giants like Amazons, and Ebay.

But decades before the internet, technology was shaping what machines could do and how they did it. How can AI transform your business? Though slower to grow, AI is already redefining business operations. This article will dig into how Rolls-Royce, Morgan Stanley and Mastercard are leveraging AI in their operations as well as share three things to consider when starting your AI projects.

‍Asset Intelligence: Rolls-Royce: Transforming Asset Maintenance

A prime example of AI’s application is monitoring and maintaining company assets. Traditionally, asset maintenance revolved around fixed schedules or estimates of a machine's lifetime. This approach is, however, fraught with over-maintenance and under-maintenance risks, resulting in unnecessary costs, unexpected breakdowns, and potential disruption to business activities.

Rolls-Royce's Intelligent Engine vision harnesses AI to monitor and maintain aviation assets. This seeks to usher in an era where engines are contextually aware, capable of comprehension, and connected to others in an ever-evolving ecosystem to drive efficiencies for airline operators. 

Artificial Intelligence is at the forefront of Rolls-Royce’s operations in this regard.  As Dr Terence Hung, Chief of Future Intelligence Technologies at Rolls-Royce, puts it:

“Expanding our work with AI while growing our digital ecosystem is a journey we have been on for a long time. The benefits these technologies have delivered to our business have further motivated us to explore new ways to optimize our workstreams. There are huge opportunities to be discovered through digitalization and AI, not just for Rolls-Royce but also for society in general.”

Another way that Rolls-Royce is advancing the field of aircraft maintenance is with its innovative intelligent borescope. Rolls-Royce's AI-powered borescope speeds up engine inspections compared to traditional methods. The infographic below provides a visual overview of how this cutting-edge tool functions and the benefits it brings to aircraft engine maintenance. 

This pen-sized device scans large objects like turbine blades using AI. It conducts inspections by moving through the engine to capture images and scans the pictures through AI technology similar to facial recognition to check for irregularities. Then, an operator finalizes the review and gives necessary recommendations. 

This AI tool has helped Rolls-Royce reduce its client's engine inspection times by about 75%, save up to £100m in inspection costs over five years, and keep aircraft in the air for longer. 

Morgan Stanley: AI-Powered Knowledge Management

Artificial intelligence can be deployed to make innovative use of this information, as Morgan Stanley has been able to showcase with the development of its ‘AI @ Morgan Stanley Assistant’ released in 2023.

This powerful internal chatbot was built to provide answers to Morgan Stanley’s financial advisory team in little time.

Developed in close collaboration with Open AI, the chatbot is a GPT-4-based tool trained upon. It draws its information exclusively from Morgan Stanley’s proprietary content library, including analysts' insights, market research, and commentary.

As such, financial advisors no longer have to rummage through several files on the company’s internal website to search for information. Instead, they can ask the Chatbot, which parses the request and provides the necessary answer with links to the source document. 

“The model effectively unlocks the cumulative knowledge of Morgan Stanley Wealth Management. You essentially have the knowledge of the most knowledgeable person in Wealth Management—instantly.” - ^‍Jeff McMillan, Head of Analytics, Data & Innovation at JP Morgan

He puts the success of the AI tool down to three elements:

• The extraordinary capabilities of GPT4 TO instantly access, process, and synthesize content.
• The intellectual capital of Morgan Stanley has been meticulously developed for almost a century across diverse capital markets, economic regions, and asset classes and bolstered by internal controls.
• Through their expertise and feedback, the incredible people at Morgan Stanley contribute to shaping the tool to help advisors get the insights necessary to serve their clients better.

Morgan Stanley’s co-president Andy Saperstein also notes that the tool is set to bring new efficiencies to advisor practices and ultimately help advisors free up time to serve their clients. This use case is unique in the wealth management industry, with Morgan Stanley being the first among the industry titans to adopt it for this purpose. However, as other players have reportedly worked to develop similar models, one can reasonably suspect that very soon. Advisors at Morgan Stanley won’t be the only ones with an AI sidekick.

Mastercard's Fights Fraud: AI's Role in Financial Security, Fraud Detection and Prevention

“The fraudster's greatest liability is the certainty that the fraud is too clever to be detected," stated Louis J. Freeh, a sentiment strikingly captures the essence of the battle against deceptive practices in finance. 

It is precisely this overconfidence in the in-detectability of their schemes that AI-powered tools, such as Mastercard’s Consumer Fraud Risk system, are designed to challenge and overcome. As fraudsters use more sophisticated tools like generative AI and deepfake technology, companies like Mastercard's Consumer Fraud Risk system need AI protection.

Mastercard’s Consumer Fraud Risk system is another use case of an AI tool deployed to great use in preventing impersonation fraud. Mastercard's system analyzes 5 years of criminal data to detect impersonation fraud.

Using this data, Mastercard, in collaboration with partnered banks, analyzes transactions in real-time to detect fraud by monitoring factors such as account names, payment values, payer and payee history, and the payee’s links to accounts associated with scams.

The effectiveness of this tool has been well-demonstrated in the UK, where nine banks, including Lloyds Bank, Halifax, Bank of Scotland, and TSB, have adopted the solution.

The fraud prevention tool has proven successful, especially with other tools that give insights into customers and their behavior. At Lloyds Bank, for example, AI pattern-recognition tools are used for behavioral analysis to build “a detailed profile of how a customer usually acts,” including how long it takes them to type and how they navigate their screens. Fraud prevention director Liz Ziegler says this helps the tool freeze payments when it spots unusual activity.

In TSB in particular, the Mastercard’s CFR system helped the bank dramatically increase its fraud detection rate in just four months. If TSBs results are replicated by all banks, this would equate to almost £100m saved across the U.K. 

In a Bloomberg interview, Paul Davis emphasized that the system has demonstrated remarkable proficiency in identifying purchase scams, constituting almost 50% of all authorized push payments frauds. 

In a Bloomberg interview, Paul Davis emphasized that the system has demonstrated remarkable proficiency in identifying purchase scams, constituting almost 50% of all authorized push payment frauds. 

Looking to Launch Your AI Project? Here Are 3 Things to Consider 

Business realities are different across companies and industries. Nonetheless, there are important relationships that every business must consider when implementing AI into their workplace. These are AI and:

The Problem It Solves

In deciding how to incorporate AI into a business, the relevant stakeholders must determine use cases where it is practicable. You can do this by examining pain points that need to be addressed and areas where gains can be made.

Second, you’ve got to determine how the system helps you achieve this goal and whether the projected gains justify its cost. Endeavor to develop criteria that will help you evaluate how the tool meets your needs and monitor its development.

You should also consider the integration process of the AI tool and the best means to deploy it without significantly affecting business operations.

The People Who Use It

Organizations must also consider how users interact with the AI systems in place. Comprehensive training on using the tool is non-negotiable, and employees should be more than adept at using it.

This was well demonstrated at Morgan Stanley, where learning to talk to the AI machine is important for getting a good response. McMillan, while discussing the Chatbot, explains why:

“For example, in the old world, you might type ‘IRA rollover’ into the search bar, and then sift through all the links in the return until you find what you’re looking for. In the new world, you would say, ‘I’m trying to do an IRA rollover for a 70-year-old client who is no longer employed.’ GPT-4’s response would include instructions completing the rollover, justifications for doing so, a counterpoint if appropriate and links to important source documents.”

You should also establish clear parameters that determine when a human should step in to make decisions that require nuanced decision-making or empathy—areas where AI might need more hand-holding. For instance, in customer service, identifying when a complaint should be escalated from an AI chatbot to a human agent ensures a seamless transition and addresses situations beyond the AI's capabilities.

It’s beneficial to make these escalation points become refinement opportunities to learn and become better. This approach is used, for example, by Aida, a SEB bank AI chatbot—or permanent employee as her employers describe her–- that interacts with millions of customers.  

Aida excels at chatting in natural language, using her wealth of data to handle complex customer queries and develop thoughtful follow-up questions. However, she smoothly hands the conversation to a human customer service representative in about 30% of cases where challenges can't be resolved. Aida then keeps a close eye on that interaction, learning from it to tackle similar problems better.

The Datasets It is Trained On

Thanks to large language models such as GPT-4, Llama, Claude, and many others, there’s no telling what limits companies can achieve with their datasets. However, any AI system is only as good—in a functional and ethical sense as the data that trains it. Hence, businesses must be extra meticulous with the data used as feedstock in building their AI systems. 

This care must extend to the quality of the data. The best AI tools rely on comprehensive and diverse datasets to ensure their proficiency in handling different scenarios. And things can go awfully wrong when the data is just not good enough.

This means businesses are responsible for ensuring their datasets are well-curated and encompass many instances.

You should also watch out for biases and inaccuracies embedded in the datasets. As one Harvard research suggests, some of the best ways include conducting third-party audits and utilizing technical tools such as IBM’s Fairness 360.

Security considerations are also relevant here. Both from a customer’s perspective and a business one, particularly when the AI has been trained or has access to confidential information. 

Can the user of your company’s AI system fool it into exposing non-public business information? Can an unauthorized personnel access a customer’s personal information by entering certain prompts?

This means privacy measures should be baked into the system immediately, not as an afterthought. The principle of least privilege (PLoP) is one of the measures you should consider implementing to handle this challenge.

Put simply, the least privilege entails users accessing only the information required to perform their tasks. Thus, before any information is revealed to them by the AI model, a layer of security must be in place to verify that they are qualified for access. 

Navigating the AI Journey: Balancing Innovation with Prudence

As organizations continue to seek out areas to improve efficiencies, AI development is bound to continue to grow. It is perhaps inevitable then that there will be a few misses along the way. However, that is not necessarily bad; after all, artificial intelligence is supposed to mirror that of humans: we learn from the information we receive and from our own experiences. This underscores why AI projects must remain the same; companies should prioritize extensive testing and addressing feedback before rolling out the tool. 

Also, businesses should be mindful of the limitations and maturity of their AI systems, as such protocols and human oversight must be put in place to produce desired outcomes. For example, in situations like Rolls-Royce's aircraft engine inspections, human operators bring their experience, intuition, and contextual understanding to complement AI’s capabilities. Internal controls are embedded within Morgan Stanley’s AI chatbot to ensure that advisors can confirm the accuracy and reliability of the information provided. 

By designing AI systems in line with their functions, users, good training dataset, and industry guidelines, it is only a matter of time before AI begins taking on even more significant roles as part of the industries' workforce, and we can’t wait to be a part of that.