Practical AI Implementations: Real-World Use Cases and Success Stories
.jpg?width=50&height=50&name=Kuba%201%20k%20(1).jpg)
What was preventing businesses from implementing AI after investing in it?
In my opinion, the primary barrier was the lack of preparedness and knowledge, which often leads to a fear of failure.
However, since ChatGPT dropped, we’ve all observed a significant shift. AI has evolved from merely driving innovation to becoming an integral part of today’s business.
The global AI market is expected to reach $390.9 billion by 2025, growing at a CAGR of 46.2%. AI’s potential to enhance customer experiences and streamline operations is immense.
While some companies have successfully implemented AI technology…
others have struggled.
I want to delve into some real-life examples which I featured in my newsletter – I launched AI’m Informed in June 2023 – and explore what has worked, as well as what hasn’t.
The ones that worked
Target's GenAI-powered store companion
The Target Store Companion app, powered by generative AI is a great example of AI implementation.
The AI-powered chatbot was created to help employees access information more efficiently, thereby enhancing their ability to serve customers effectively. By integrating AI, Target wanted to significantly boost employee productivity and improve guest experiences across its nearly 2,000 stores.
The chatbot is installed on company devices so employees can get instant answers to questions about a range of topics, from the Target Circle rewards program to how to restart a cash register after a power outage.
This initiative not only allowed them to streamline their operations but also helped to foster a more knowledgeable and responsive workforce. Additionally, the application provided personalized training modules, enabling staff to upskill continuously. This approach ensured immediate problem-solving capabilities and contributed to long-term employee development.
Michael Kors’ AI-enhanced customer experience
Another fantastic mention for the retail sector can be found with Michael Kors. The companypioneered the integration of Mastercard's generative AI assistant into its website. The AI tool, known as Shopping Muse, translates customers' casual language into personalized product recommendations, leading to a 15-20% higher conversion rate compared to traditional search queries during initial tests.
Michael Kors' Shopping Muse mimics the in-store experience by converting consumers' everyday language into personalized product recommendations. This allows fashion-savvy shoppers to swiftly discover the perfect look that aligns with both their inquiries and their shown behaviors and preferences
The implementation of AI not only improved customer satisfaction but also drove higher sales conversions. This dynamic interaction helped build a more engaging shopping environment, increasing the likelihood of repeat purchases and customer loyalty.
Trustly's AI-powered recurring payments
When it comes to finance, I believe Trustly is undoubtedly a leader in open banking payments.
The team launched an AI-powered solution designed to automate repeat transactions. This initiative aimed to tackle the 50% abandonment rate at checkout due to cumbersome payment processes.
Trustly Recurring Payments allows businesses to accept recurring transactions directly from customers' bank accounts. It combines Direct Debit, pay-by-bank, and Trustly Azura, a data engine that improves signup, security, and fund collection. The solution predicts the best times for successful payments and ensures security with biometric authorization and verified payment credentials.
Trustly streamlined these transactions to improve the customer experience and reduce revenue loss. The AI system ensured a seamless process for users and enhanced the overall efficiency of payment operations.
The implementation addressed specific pain points in financial services, ultimately leading to higher customer retention and satisfaction.
ABN Amro’s AI in trade finance
Another great example can be found with ABN Amro. The Dutch bank implemented AI to automate manual processes in trade finance, particularly in handling letters of credit and documentary collections.
The solution leverages various AI techniques, such as large language models (LLMs) to parse document information and increase the level of digitisation possible, allowing it to extract key data and check documentation for compliance.
This application allowed the company to streamline complex, time-consuming tasks, reducing errors and increasing operational efficiency. The AI system can also cross-verify documents in multiple languages and formats, reducing the risk of fraud and ensuring compliance with international trade regulations.
The ones that didn’t
McDonald's drive-thru AI project
Despite the potential of AI, not all implementations succeed.
McDonald's has decided to end its two-year test of drive-thru automated order-taking (AOT) technology, developed in collaboration with IBM. The project did not meet the company's expectations, though McDonald's remains optimistic about AI's future in drive-thru services and is looking for new partners to realize this vision.
The AI system, which uses voice recognition software to process orders, faced challenges in accurately taking orders and integrating smoothly into the fast-paced environment of drive-thru operations.
The way I see it, their setback demonstrates the importance of continuous iteration and improvement when implementing AI.
One of the biggest problems was that the AI struggled with understanding diverse accents and slang, leading to frequent errors in order processing. This demonstrated the clear importance of extensive training and localization of AI systems to cater to varied customer bases.
Amazon's AI Recruiting Tool
Amazon developed an AI-powered recruiting tooldesigned to streamline the hiring process by evaluating job applicants' resumes.
The hiring tool used AI to give job candidates scores ranging from one to five stars - much like rating products on Amazon.
However, the project was ultimately abandoned after it was discovered that the AI system was biased against women. The AI had been trained on resumes submitted to the company over a ten-year period, which were predominantly from men. This led to the system downgrading resumes that included the word "women's" or that were from all-women colleges.
What we see here is the critical importance of ensuring diversity and fairness in AI training data to avoid perpetuating existing biases. When implementing, in order to avoid mistakes, its always best to provide a correct data set and iterate.
IBM Watson in Healthcare
Looking into the healthcare industry, IBM's Watson was the first failing example I can think of. This tool was implemented to assist doctors in providing treatment recommendations for cancer patients.
IBM Watson's cognitive and analytical abilities allowed it to respond to human speech, analyze large amounts of data, and provide answers to previously unsolvable questions. As it processed new data, Watson used machine learning to expand its knowledge and enhance the insights it delivers.
However, it encountered major problems by often suggesting incorrect and unsafe treatments. This was because it was trained on hypothetical scenarios instead of real patient data, resulting in unreliable outputs and distrust from healthcare professionals.
This example highlights the challenges and risks of deploying AI when training data does not accurately reflect real-world conditions.
What did we learn?
In my opinion, successful AI implementations often share common traits: clear objectives, strong data infrastructure, and continuous monitoring and adjustment.
Companies like Target and Michael Kors demonstrate the benefits of aligning AI initiatives with business goals and investing in the necessary technology and training. For example, Target's Store Companion app is directly linked to improving employee efficiency and customer service, while Michael Kors' AI assistant enhances the online shopping experience.
These examples show the importance of a strategic approach to AI, where the technology is deployed to meet specific business needs and is supported by adequate resources.
Successful implementations often involve a phased rollout, allowing for iterative improvements based on user feedback and real-world performance.
On the other hand, failures like McDonald's drive-thru project demonstrate the importance of realistic expectations and thorough testing.
Misalignment between AI capabilities and business needs, as well as inadequate preparation for deployment, can lead to disappointing outcomes. The challenges faced by McDonald's show the need for rigorous testing and validation of AI systems before widespread implementation.
Another point is that if you want to implement AI tools, you need accurate data. We saw in both the Amazon and the IBM case that the lack of accurate data backfired their efforts.
And where are we headed?
Success with AI requires a strategic approach, ongoing adaptation, and a willingness to learn from both successes and failures. By understanding the factors that contribute to successful AI implementations and recognizing common pitfalls, you can better prepare for AI adoption.
As I look ahead, the integration of AI with other emerging technologies like the Internet of Things (IoT), blockchain, and quantum computing could further amplify its impact.
For instance, AI-driven IoT devices can provide real-time data and analytics, leading to more informed decision-making in industries ranging from healthcare to manufacturing. Meanwhile, blockchain can enhance the security and transparency of AI algorithms, addressing concerns around data privacy and trust.
Businesses that not only invest in AI technologies, but also in understanding and mastering them, will be the ones that are better positioned to lead in their respective industries, drive innovation and create value for their stakeholders.
.jpg?width=240&height=240&name=Kuba%201%20k%20(1).jpg)
