Implementing Artificial Intelligence in supply chains


Johnson & Johnson’s Neil Ackerman explains the power of AI and what organizations need to do to be part of the transformation

Artificial intelligence is a technology that fascinates and intimidates so many organizations. Whenever I ask people in supply chain about subjects where they have a knowledge gap, AI is high on that list. Supply chain leaders are now asking themselves – “What does artificial intelligence really mean to me? How can I start using it to create a more efficient supply chain?” The reality is that people want and need to be part of it for their careers as well as their companies. They want to be industry game-changers, implementing and developing the power of AI to make huge transformations, not just sounding smart in meetings. That’s why I’m often asked by different people around the globe to offer insights on AI.

When you think about what Artificial Intelligence really comprises, you start to break down the myths and see it for what it is. AI is a broad term that include robotics, autonomous vehicles, planning, language, machine learning, and virtual agents. AI is not scary and actually the word ‘Artificial’ in Artificial intelligence is a misnomer, because there’s nothing artificial about this. It’s what I’d call real intelligence. AI should be RI.

It takes the real actions of real people and translates them into structured data that you can derive benefits from, whether that means robotics, language, machine learning or something like autonomous vehicles, behind which there is serious thinking and technology.

Whether people like it or not, AI is going to transform the globe over time, and those who win will win big.

 ‘Artificial’ in Artificial intelligence is a misnomer

Rapid growth: dream to reality

Those firms that begin to harness its power will enjoy a step-change in growth and relevance as forthcoming generations continue to be enthralled by the digital revolution. The rapid growth of AI continues to be driven by our technology friends at Amazon, Google, Apple and Facebook, who were at the forefront of an estimated $40 billion commercial investment last year in AI. It’s no surprise that these companies have also grown tremendously in their stock value.

Foundations for successful AI

It is worth considering the qualities the tech companies have that are enabling them to succeed in AI. The highest adopters have SIX key characteristics that make them early winners.

  • digitally mature

  • motivated organizations

  • cultural belief throughout the company

  • utilizing many forms of different technologies overall

  • focused heavily on growth

  • strong senior buy-in (executives understand its potential and encourage development)

Building on digital foundations

AI is dependent on having a digital foundation and access to extensive and unique data sets. The largest investment is currently in the machine learning area, which takes structured data and generates an output that the human mind could not achieve as quickly as AI. Without that base data, you have nothing to build on. These digitally mature organizations not only understand digital, they can also recite cases of where it has been successful - they’re plugged into what’s happening elsewhere. They’re usually large, again because you need a significant quantity of data, as well the people and resources to develop an AI program at scale.

Why culture matters

I was watching the science cartoon Leapfrog with my children, and the song said: “Design it, build it, improve it”.  That’s the most telling explanation I’ve heard of what children do – they’re always iterating, always learning, and that’s how you unlock AI. The culture of innovation requires leadership to understand the improvement part of the equation, and executives must be OK with the idea that at some stage in the process, there are going to be failures.

Getting started on the AI journey

When organizations set out to create an AI pilot, the early steps must include identifying a robust business case. Instead of trying to boil the ocean, identify a small use case that you could communicate in a 30-second elevator pitch. You can then use that to create a learning environment which generates a clear output on what results you want, and build from there.

Other elements for the pilot include ensuring your data is structured appropriately, clear focus on the inputs that will guide you to the output metric, access to proper AI tools and data scientists, building an efficient workflow and ensuring that leadership is excited and focused on success.

A critical point to consider is that an AI pilot is not an experiment if you already know the outcome. The problem is that big companies often want to know the outcome in advance. But in those circumstances, it’s not really an experiment and there’s nothing to improve. This is the mindset that has to change.

Why now? The AI tipping point

Computing power has never been as great as it is now and algorithms have never been as efficient. This is just incredible technology.

Historically, AI has had some mixed results, but what really changed is that the world is finally creating so much data that it is becoming meaningfully predictable. The data shows repeatedly that humans are more alike than we are different. When you put all that data together and begin to track human movements at great scale, you can create such meaningful predictions. As this data is stored and shows this human-like intelligence (which I’ll call real, not artificial), it creates better decision making. Imagine a supply chain where, at scale, you could begin to predict how many people are going to need a product when a specific set of circumstances are present. It’s a very powerful prediction that, based on years and years of data, AI can say that when a specific set of conditions occurs, x more sales occur.

Consider the scenario of a hurricane and bottled water sales. As the hurricane is heading to an area and then arrives, the model predicts people buying more groceries than ever before. But this isn’t the first time a hurricane has come, so the models and AI can tell you with greater clarity where, when and how this is all going to precipitate itself. If we can get better models through AI, we wouldn’t be running out of water bottles before the hurricane comes. These are the things AI can do over time.

Computing power has never been as great as it is now


A new paradigm for forecasting

Anyone can do demand forecasting today, but the problem is that demand forecasting is usually wrong. That’s why they have the mean absolute percentage error (MAPE). Every digit you can improve your MAPE, you will get a better return on your dollar, plus consumers will be more satisfied because they’re getting the product they want, when they want it. AI adoption will continue to grow each year across industries, more data will continue to be stored and therefore machines can begin to show human-like intelligence including language, image recognition, text and decision-making capabilities. But remember that today, AI still has limitations and can be susceptible to bias based on data set sizes.

Fast vs slow: adoption in industry

Along with the technology companies, which you might expect, the highest adopters are the telecomms and financial services sectors, while there is lower adoption in education and healthcare, although recently that has started to change. Commercial considerations, uncertain returns, privacy and regulatory pressures are all rational reasons for slow adoption. Interestingly, the automotive sector is an early adopter due to advanced robotics and self-driving car objectives. Healthcare has been slower, faced with regulatory concerns, privacy and at-scale customer acceptance of the sharing of data.

The 4Ps of AI.

When thinking about a way to contextualize AI, consider the 4Ps, not the celebrated 4Ps of Marketing, but the 4Ps of AI metrics:

Projection – enable demand optimization

Production – lower cost and high quality

Promote – to the right consumer at the right time and place

Personalization – convenient user experiences.

It’s time to start the journey

So now what? I suggest that firms focus on a use case and something they are passionate about solving. Then build those assets and data sets, and watch the magic of AI come alive in a collaborative environment. When you find the true source of value, you will want to do more, and that only helps to build the ecosystem. AI cannot be used everywhere, but can be a tool to help a percentage of activities become autonomous.

If you want to see AI in action, take a trip to Silicon Valley in California. With more than 12,000 start-ups, AI is meaningful. If you’re in China, go to Shenzhen and watch what AI is doing to production of hardware. AI is the next digital frontier and I cannot wait to see how the world will change in the next decade. Go to one of these hotbeds to start your own journey.

Author profile:

Neil Ackerman is the Senior Director, Global Supply Chain Advanced Planning and Innovation, for Johnson & Johnson. He is responsible for accelerating supply chain innovation and enablement of advanced planning processes and technologies worldwide. His team is critical in bringing value-based prototyping to life within Johnson & Johnson and beyond.

Prior to this role, Neil held various positions at, including Strategy Head for Fulfillment By Amazon and the General Manager and Inventor of the Amazon Small and Light global program. Most recently, he was the Global Director of Integrated Supply Chain, eCommerce strategy and technology innovation at Mondelez International. He is an expert on e-commerce and supply chain innovations (including Artificial Intelligence) and has been published and given multiple keynotes globally for his work on the e-commerce flywheel and its impact on supply chain for a long-term competitive advantage.