This article was originally posted on Forbes.com.
Artificial intelligence (AI) is here. It's in the driver-assisted features of our cars, it's in the virtual scheduling assistants we use, and it's even embedded in all of our digital shopping experiences.
In the manufacturing and packaging industries, the question isn't whether we will use AI or even how we will use it. We need to start with why we will use it. In which business opportunities will we drive meaningful impact? How will we move faster, become more efficient, automate or solve a real challenge our organization faces?
We have to start with why because data is the foundation for creating meaningful AI insights. AI systems can't generate quality results without accurate data-feeding training systems. If you've played with ChatGPT at all, you've likely seen the results when the system lacks data. ChatGPT will generate plausible-sounding—but completely untrue—responses to inquiries. While that is fine when you are creating a playful poem, it most certainly is not OK when it comes to operating your business.
Unfortunately for most organizations, pockets of siloed, inconsistent data still plague the operation. Understanding which business opportunities you want to tackle first will point you toward the right dataset to collect and organize.
AI Is A Team Sport
There is no doubt that in the packaging industry, AI and machine learning (ML) are important inventions. However, it's not just about new technology or inventions; it's how we use this technology to rethink work processes and business strategies. Our work is intimately integrated with our supply chain and requires a network approach that innovates new ways of applying AI to our work.
Any one company can't just look at its own data. It must have visibility into its suppliers' data. By taking a collaborative approach, AI has a better, more comprehensive foundation to build off. Without this full visibility into data, AI can't run optimally. This technology is only as strong as the dataset/foundation it is based on.
Focusing On The Right Data, Not Just Any Data
This is particularly true of sustainability initiatives. It's not enough to know our suppliers' data. We need the data of our suppliers' suppliers. In fact, up to 90% of an organization's carbon footprint can come from Scope 3 networks. Getting that data is hard.
In 2022, we published our State of Sustainability report, which was developed from the responses of approximately 500 supply chain professionals. By far, the biggest challenge for companies (chosen by 41% of participants) is aggregating data. The next closest response (17%) was tracking progress/reporting. Even if you leverage AI/ML technology, recommendations and analysis won't be based on the most accurate data available if you are not gathering the right data.
Individually, an organization can't get enough visibility to feed AI models efficiently. Technology needs to aggregate the data for economies of scale, as it did for certification of insurance/authentication (COI/COA).
It used to be that no standardization existed for COI/COAs. Different parts of the business would use their own naming conventions. It got even messier when working with suppliers. Today, technology can bring together enough COI/COAs to make an AI engine intelligent. It can automatically recognize patterns and automate workflow—saving time and reducing errors.
Imagine what is possible when we create truly collaborative digital workspaces across the entire network and all related activities.
AI Can Be Risky Business
If we train a system on bad data, we will get poor results. Before we can imagine future models, we have to get access to the data. We need to break down data silos that often exist for supplier portals, product catalogs, stale websites and more.
For example, if your model is based on bad data about a supplier, your predictions about them are fundamentally flawed. You'll be worse off than educated guesswork, and you could be directed to pick a less-than-optimal supplier.
Demanding Trust
We don't have infinite tries to get AI right. Humans are using the recommendations AI engines provide. It only takes one costly mistake for us to lose our faith in the system. Without confidence, technology provides little gain for the team.
Regulations and the greater good are important motivators, but there is also a more practical driver of information transparency. Suppliers who make this easier will grow faster because this is becoming a required standard.
We saw this in the past. The open-source movement proved that it's not the code that drives innovation and intellectual property; it's how the code is applied. Uber's strength is not in its map or its GPS. Rather, it's a new business model of how the technology is used.
The same evolution is happening to the packaging industry.
Conclusion
AI unlocks scenarios that all of us deemed impossible in the past. We're dipping our toe in the water now for a bright future of predictive analytics, but we can't leap ahead. First, we have to collect the data and share it. There are some industries where really complex analysis needs to happen to make advances.
Packaging is actually not quite there yet. It's really in its infancy stage. Most companies are still trying to collect the right packaging data, but we now have better incentives to do that through regulator pressures and tax implications.
It's about collecting accurate data in one place. Then you can leverage AI and ML technology to learn, optimize and innovate much faster.
