Business transformation in the age of AI

Let’s start by understanding what’s changing. One way of framing AI is that it is the next paradigm shift in business operations, following others that have come before. These shifts transformed the way that businesses worked, whether that was in how globalization changed the way back-office functions were delivered, to how digital transformation changed organizations’ engagement with customers and markets.

The recent excitement about AI has been centered around LLMs (Large Language Models) and GPTs (Generative Pre-trained Transformers). However, industry veterans have had two reactions: first, that AI has been around for decades – including neural networks, the precursor to LLMs - and second, that LLMs in their current state can only do so much. The reason that we have become so excited about LLMs since GPT 3.5 is because we have all been able to experience the technology easily, have been impressed by its capabilities, and consequently can imagine their transformative potential. This is exactly the same “capturing of the collective imagination” that led to the hype around the Internet in the 1990s and smartphones in the 2000s.

As a result, a lot of companies have spent an inordinate amount of time, money and energy pursuing “technology forward” experimentation centered on LLMs. We believe it’s not unfair to say that the results in terms of business impact have not yet lived up to the hype. A recent study by S&P global states that, in the past year, the percentage of companies abandoning the majority of their AI initiatives has climbed from 17% to 42% (*1). Why is that?

First, LLMs are only one fairly narrow part of a broader set of transformational AI capabilities. The broader field encompasses a wide range of modelling techniques, such as machine learning and reinforcement learning, and processing capabilities, such as computer vision and speech to text. AI then sits alongside a suite of other established and emerging technologies, from basic UI / UX based transaction processing to IoT and robotics.

It is relatively unlikely that harnessing LLMs alone is going to drive disruptive business impact, in the same way that IoT sensors, or computer vision would also fail to do so on their own. To drive meaningful impact requires assembling the right combination of technology capabilities to resolve an end-to-end use case. As the use case or ‘impact theme’ becomes more complex, so too does the required set of solutions to address it.

For example, business transaction processing can be achieved with relatively simple UI / UX based architectures and is enhanced when bringing RPA (Robotic Process Automation) or Mobility into the picture. However, if you want to automate a complex physical process, such as enabling self-driving cars, you will need to combine, UI / UX, Mobility, Machine Learning, Computer Vision, IoT and Robotics. The trade-off is that with this increased complexity also comes the opportunity to solve extremely high-impact business challenges.

The heatmap below provides a conceptual approach to think about how technologies combine to resolve increasingly complex business opportunities.

Let’s look at some these principles in action. Fujitsu’s portfolio company GK Software is a great example of how we can bring together a broad ecosystem of technologies to solve real business challenges, in this case, for the retail sector. At its core, GK Software delivers SaaS-based retail technologies including checkout, consumer engagement and store operations to global retailers. It has been the market leader for over 35 years.

GK has developed from its roots in POS solutions to build a comprehensive suite of advanced software capabilities for the retail sector that bring together AI, business workflows, sensors, and IoT to drive transformative impact in Retail Operations.

One example is how GK has given its retail customers the choice to add certain AI technologies to their Self-Checkout solutions to reduce costs and improve convenience at a feasible price point. A great challenge for leading retailers is the prevention of unintentional or intentional fraud at the self-checkout by customers failing to scan all their items. To effectively tackle this challenge, retailers have faced enormous costs of exchanging their entire hardware landscape or hiring additional staff to oversee customers. GK’s “Vision” technology, effectively allows retailers now to integrate AI in their Self-Checkout retroactively and leverage computer vision technology to assist customers and staff in ensuring all items have been scanned correctly. This retrofit approach has opened the door for retailers to apply new technology at an effective price point and at scale, creating a product that is highly applicable in today’s scenarios and adds real value. The technology also creates a more convenient experience for customers by guessing their age (thereby avoiding a wait for an assistant to check) and recognizing items that need a manual lookup like fruit and vegetables.

Thinking back to the heat map above, here UI, CV, OCR and ML are integrated together to deliver the impact theme – in this case fraud and loss prevention. Importantly, we also thought through and addressed the significant operational challenges of rolling this out into a live environment.
A second example is GK’s Price Optimization solution. A major cost to the retailers comes from fluctuations in supply and demand – meeting demand from customers with the right levels of stock. Overstocking leads to holding costs, which might be 30% of the inventory’s annual value. Conversely, not holding enough stock makes the retailer vulnerable to sudden changes in demand. Consider, for example how an unexpected spell of good weather changes what consumers might buy from a supermarket. Running out of stock is a cost not only from lost revenue but from customer goodwill.

GK Software delivers solutions to dynamically optimize pricing – enabling a smoothing out of supply and demand and reducing the cost exposure of the retailer to surpluses or shortfalls in stock. The technology determines the optimum price for each item and changes it in real time on the shelves. Again, the solution combines component technologies, in this case it’s the stock management system, Machine Learning, and intelligent labels in the physical store that are integrated together to deliver the impact theme – optimized pricing.

The solutions are typically handling daily price calculations for over half a million items of stock and a 3-5% increase in revenue is typical.

These capabilities span across from the customer, the physical front-end, the in-store experience through to back-end systems, logistics and into the retailer’s supply chain. Fujitsu complements GK’s software expertise with technology, spanning AI enablement, cyber security, managed services, platform configuration, and ongoing innovation.

While AI may be the disruptive and game-changing technology in each case, it is just one component of a complex system. These examples show that the power of AI comes not through pilots of a hundred different use cases, but rather through integrating it into core processes that deliver big ticket gains for the business.

Looking ahead, we’re continuing to expand this foundation, delivering AI-powered workflow management, content optimization, and LLM-based training services. As a combined force, we’re building the end-to-end, intelligent digital platform for modern retail.

Having gained an appreciation of the technology solution set and what it will take to harness it to drive impact, the business is in a much better position to make choices about how to identify and prioritize the opportunities that arise. Even with a clear vision, making the right choices can still be challenging. To help guide better decisions and smarter investments, here are three key considerations to keep in mind. Is it relevant? Is it realistic? Is it practical?

Is it relevant – does it address a significant business issue?

Your innovation processes need to stop thinking technology forward and start thinking business back. Spend time exploring how business operations could work in the future – envision and debate. How will the era of AI upend the cost basis, product offering and customer value proposition of your industry.

Find big problems to solve. Any solution that will drive significant impact is going to require a huge amount of effort to solve, so focus on those that are worth the investment. Don’t “let a thousand flowers bloom”, because most likely none of them will. Instead, identify 5-10 game changing solutions and prioritize. Sure, you can start with smaller POCs and pilots, but make sure you have anchored on a big outcome first.

Is it realistic – can you economically deliver the solution? 

As you ideate on potential opportunities, avoid theoretical discussions. We have seen too many clients trying to recover from misguided efforts to explore the “art of the possible”. Focus on the “art of the viable”. As you explore options, couple business expertise with technical expertise grounded in the reality of what it will take to actually deliver a solution.

Prioritizing then requires anticipating the requirements to deliver a market functional solution. Are edge cases or anomalies likely to undermine the solution? Can the timeline and cost structure deliver acceptable ROI? And plan for maintenance. What is the post-implementation upkeep and evolution? These are critical to long-term success. In fact, a major part of doing this successfully is creating a cultural DNA that is setup to enable continuous innovation of the new processes and systems.