Artificial Intelligence (AI) has been a buzzword for decades, capturing the imagination of researchers, entrepreneurs, and the general public alike. However, beyond the hype and sensationalism surrounding AI, it is crucial to ground our understanding in actionable insights that can drive real-world applications. By focusing on the practical implications and tangible use cases of AI, we can unlock its true potential and harness its power to solve complex problems, enhance decision-making processes, and revolutionize various industries. An introduction to AI beyond the hype should delve into the fundamental principles, current state-of-the-art techniques, and the ethical considerations that must be addressed to ensure responsible and beneficial deployment of these powerful technologies.
This week, we had the chance to gain valuable perspectives from Gianluca Ruggiero, the Founder and CEO of Massive Data Heights and Product Director of Lisia. Ruggiero resides in the USA, where he established Massive Inc, an AI startup specializing in market data analysis. His experiences at internationally renowned advertising agencies, holding pivotal strategic roles, have been instrumental.
AIM Research: Could you briefly explain what Massive does to set the context for our discussion on the hype around generative AI? How can we move from hype to actionable insights that will benefit our audience?
“Massive is a platform with a very simple task: it informs CPG companies about which products they need to kill, build, and improve in their existing assortment.”
Gianluca Ruggiero: I was born and raised in southern Italy and started my career at Unilever, where I was responsible for global brands, innovation and market strategy. I was in marketing at the time and later moved to consultancy. I led the Italy office of McCann Worldgroup, one of the big five advertising agencies. Then I moved to the States to lead the strategy globally for J. Walter Thompson, another big five advertising company.
These experiences were crucial for the creation of Massive because, while working with Fortune 500 companies globally, I noticed that it was increasingly difficult for companies, as well as for me as a consultant, to cope with the speed and the complexity of modern markets, specifically, but not exclusively, in the consumer product sector. The idea came that data were available to improve strategy, but they did not make sense of. They were too big for traditional methodologies. This is where the idea of creating Massive came from.
Massive was launched with a soft launch in 2019. We were lucky to start working with the best and brightest in the industry. Our first customer was Procter & Gamble; they’ve been a customer of ours for five years and now six. We’re just signing the six-year contract. Then came Nestlé and many other big names in the CPG sector. We spent our first years developing the technology with these big names. Basically, Massive is a platform with a very simple task: it informs CPG companies about which products they need to kill, build, and improve in their existing assortment. It’s a simple task, but with very complex functionality, where we are basically helping our customers spend better tens, if not hundreds, of millions of dollars with the right market strategy and product strategy and so forth. So, Massive is of course making use of AI. Our team has more than 100 papers published on the topic, so we know quite a few things about artificial intelligence. And so, this is in a nutshell, me and Massive.
AIM Research: The hype around Gen AI started not long ago, and now everyone is experiencing fatigue. Given that Gen AI requires significant resources, time, and computational power, how are you seeing these issues? Are your customers getting fatigued with the promise of Gen AI? What conversations have you had, and how have you and your company maneuvered around these topics to deliver genuine and actionable insights to your clients?
“The problem with AI starts with the term “artificial intelligence””
Gianluca Ruggiero: Because of my background in a marketing corporate culture, I can tell you that this overhype of new concepts, especially in digital, is not new. We’ve seen it in all sorts and forms in the past, and it’s not good for the industry. It’s good for quick wins for the big players and media to jump on this bandwagon, but for the industry itself, it’s not beneficial. After the initial hype, fatigue sets in, and people get tired. The saying “under promise and over deliver” is good for business, but “over promise and under deliver” is very bad for business. This happens all the time when we overhype things, and AI is just one of these cases.
When I was in marketing at big agencies, I remember the hype around the digital thing. One day, someone came to my office and said, “There’s this new thing we have to jump on. It’s the QR code. It’s super important. This is the next big thing.” I was like, “Guys, this is just a link.” They didn’t even know the story of the QR code, which was created for logistics management in Japan. It was a brilliant idea and execution. Everyone started putting QR codes on posters and prints, but it doesn’t work like that. It led to a lot of fatigue. Years later, we see QR codes properly used everywhere. Unfortunately, we have this cycle of hype, high expectations, the valley of despair, and then the serious people survive and continue the work, using the technology for the greater good and business benefit.
The problem with AI starts with the term “artificial intelligence”. It’s a nice expression, but we don’t even know what intelligence is. Many brilliant friends in the academic world, working on computational biology and serious topics related to consciousness and intelligence, cannot define what intelligence is. Trying to recreate something we don’t fully understand is flawed from the beginning. A lot of hype comes from this, as we try to imitate human beings. Artificial intelligence is an amazing tool that empowers human intelligence to do things otherwise impossible, especially in seeing patterns through chaos when things become too complex, simplify and help focusing on the right problems.
Regarding AI fatigue, we see that a lot. The hype around generative AI, like ChatGPT or Dall-E, is an example. These are the models behind those applications, which are UX and user interface which are amazing and powerful, but if we start focusing less on the hype and more on practical applications, we can achieve better outcomes. In the short run, this leaves many people unsatisfied. Let’s go back and focus on the foundation of AI and solve problems.
AIM Research: You mentioned the “valley of despair” and then moving towards a more modern approach. The problem here is more nuanced, given the significant investments involved. How do you see us as an industry helping large enterprises navigate towards the reality of producing actionable insights, rather than just the hype, as soon as possible?
“Most of the initial and current investments in artificial intelligence are behind the consultative model from a business perspective.”
Gianluca Ruggiero: I think the industry is still struggling to find where the value is. Everyone is looking for the golden part, but because of the hype, we will reach the end of the rainbow a bit later than expected. So, where is the golden part in my opinion? Most of the initial and current investments in artificial intelligence are behind the consultative model from a business perspective. In other words, I go to a customer, the customer has a problem, and they typically say, “I have a lot of data about —supply chain data, sales data, data from different sources—and I can’t make sense of it. Can you help me predict my future? Can you help me solve a business problem?”
One of the first applications is a customized application based on enterprise data from the customer. In that fashion, AI experts work as consultants, coming into the business, scouting internal data, interviewing stakeholders, understanding where the value really is, reorganizing things, applying the right AI engines and technologies, and coming up with a heavily customized solution in terms of processes and data etc. This model, which was initiated by IBM (a former client when I was at WPP handling their IBM Watson campaigns), is very labor-intensive and process-heavy.
On the other end of the spectrum, we have simplified applications like ChatGPT. These are nice from an interface standpoint and can solve specific problems, like reorganizing an Excel file, improving an email, or providing information faster than a search engine although biased. However, these applications deal with unstructured online data, and getting actionable insights from them is almost impossible. When you need actionable insights, you’re looking for something you can act upon and explain. You can’t tell your board to invest $10 million because “ChatGPT told me so.” You need a thorough, explainable reasoning backed by statistical analysis, which ChatGPT and similar tools are not designed for.
A lot of the hype and subsequent fatigue comes from these quick-win applications that resemble a human consultant but can’t build a business or provide actionable insights. The solution, as I see it, lies in the middle. You want to build something scalable and actionable, focusing on the data problem within a specific vertical. Identify the sources of data, understand their biases, fill in the gaps, and create an AI application tailored to that vertical, transforming or enhancing raw data into insightful information. This approach moves from data points to actionable insights without relying on heavy consultative processes.
The consulting businesses I mentioned before are doing this, but in a way that is very customer-specific and hard to scale. The goal should be to build the best AI possible for each vertical and then replicate that success. Another issue with Generative AI is that it is similar to general AI. Looking and fighting for general AI is very dangerous. I don’t believe we’ll ever reach general AI as people envision it. I think general AI will be the sum of vertical AIs, not a single system that understands every task at hand. It’s not impossible for an AI to analyze even millions of variables, but for the designer of the AI understanding what are the implications of all these variables considering all the inner and biases about each one of these is impossible. And the belief that this AI is a black box that will figure it out anyway, because it’s smarter than us, which is also very dangerous as a society because it gives us this false idea that there is someone or something that will figure it out for us, which is not true at all.
AIM Research: When discussing AI, especially in contexts where executive leadership or any human decision-maker relies heavily on it, such as in financial investments or strategic planning for business problems like optimizing production lines in the CPG industry, how can we ensure decisions are not blindly made? Can you provide an example where AI is effectively used to guide decisions in the CPG industry, illustrating how it translates into real-world solutions and helps visualize its impact?
“But if you pretend that general AI can go about this very complex problem and sort it out and figure it out, that’s a problem.”
Gianluca Ruggiero: I can speak very precisely about the vertical task. I’ve worked for a long time, for 10 years, in this field. The narrow versus general AI issue is crucial here. I am a big advocate of narrow AI. If we want to understand what’s really going on, we have to focus on narrow applications. Generative AI is often referred to in the same breath as general AI. This is because it is designed by some, for business purposes, to promote a certain type of AI. They say, ‘Generative AI is general,’ which is clearly not the case. You also mentioned two things: answers and content. It’s great for building content, but not great for providing answers, and there’s a big difference between the two.
To your point, how can we apply this to businesses like CPG companies? First, there are two types of data: internal and external. Let’s focus on CPG companies. Even small companies selling products, whether it’s a phone or a jar of maple syrup, need to understand two sets of data. Internal data includes performance metrics such as sales, production, and distribution. The second set is competitive data, which answers how everyone else is doing, not just you. You might wish for a crystal ball to see not only your performance, like Procter & Gamble’s, but also your competitor L’Oréal’s performance. You need to know because performance is a relative concept. You can’t say if you’re doing well without knowing how the market is performing.
For example, some markets have grown triple digits for years. When you ask a commerce manager, ‘How are you doing?’ and they say, ‘We’re growing double digits,’ everyone thinks, ‘Wow, that’s impressive.’ But what if the market is growing at triple digits? Are you doing well or poorly? If your market is growing at an average of 50% and you’re growing at 23%, you’re growing ten times more than your brick-and-mortar business, but you’re still losing market share. So you need to understand that there are two sets of data: internal and external. AI can easily make sense of internal data, sometimes without even needing AI. But what about competitive data in markets with tens of thousands of competitors? When we started working with P&G, we were asked to work with their global leader brand, Olay, a huge global skincare brand. In 2019, the number of competitors in brick-and-mortar stores could be counted in the hundreds. When we showed them the data from the web, they were shocked to see 3,000 competitors, not just a few hundred. Today, the market has tens of thousands of competitors in just facial skincare in the US. This is a scale change from hundreds to tens of thousands in just half a decade.
AI helps Procter & Gamble see the entire market, which they can’t do with non-AI tools. AI fearlessly gathers all the data and makes sense of it, identifying patterns in the chaos. For example, they came to us needing an innovation idea, looking for white space and unmet needs. With our data, in just a few clicks, AI analyzed the entire market and identified a clear unmet need in sun protection factor (SPF) products. Millions of reviews showed that many people are unhappy with typical SPF products because they leave the skin looking white and clownish. Applying makeup over SPF is difficult because it disrupts the application process. Three brands tried to solve this issue, but they failed because their formulations contained too many allergens. There’s a big opportunity to launch an SPF product that makes the skin glow without looking dull and without allergens. If you do things right, you could price this product at $55 a jar, or even more. Normally, this kind of analysis would require months of expensive, biased market research. AI, designed specifically for this task, can provide these insights in hours instead of months.
And AI, if properly designed with this specific task at hand, which in our case is pretty not so narrow. Tell CPG companies what products they have to kill, build, and to improve in general looking at the whole competitive landscape that they have, optimizing their resources and their whole marketing mix to maximize their return on investment. That’s basically the narrow application of Massive. If you build an AI with this in mind and only with this in mind, you can provide those answers in days instead of not even in days, in hours of analysis instead of months, years, and so forth. But if you pretend that general AI can go about this very complex problem and sort it out and figure it out, that’s a problem.
