The Key Types of AI and Their Roles in Supply Chain Risk Management

By almost any available metric, the growth of artificial intelligence—and in particular its latest incarnation, generative AI—over the last few years has been staggering. According to research conducted by Goldman Sachs, in 2020 private investments in AI in the U.S. totalled around $28 billion. The investment banking firm projects that figure to increase by almost 250% in 2024, rocketing to $68 billion. A recent survey carried out by McKinsey & Co. offers further evidence for the aggressive adoption of the technology within the corporate world. The consultancy found that the number of businesses that are regularly using generative AI nearly doubled over the course of a single year. As of May 2024, 65% of companies surveyed—which included businesses in a broad cross-section of regions and industries—were utilizing artificial intelligence for functions like marketing, sales, and product development.

For businesses and professionals operating in the field of supply chain management, the capabilities offered by AI can sound tantalizing. The tools can perform predictive analytics, demand forecasting, and risk assessments, among other critical applications with immediate value. But the past 18 months have also seen many companies beguiled by the extravagant possibilities of artificial intelligence without analyzing whether these platforms can provide real benefits in the context of their specific business. With that in mind, it’s important for supply chain management teams intrigued by the potentialities of AI to understand the recent history of this technology, explore the different varieties of artificial intelligence currently available, and establish clear objectives for how they want the software to add value to their operations. In an era where corporations are throwing money at AI startups in a kind of rampant, unreflective frenzy, carrying out a sober examination of the technology's capabilities can be a pivotal due diligence measure. 

The Key Types of AI for Businesses to Know

The past decade-and-a-half has been a period of burgeoning growth for artificial intelligence. Various iterations of the technology have emerged at different points dating back to at least 2011, when Apple first introduced Siri with the fourth generation iPhone. The world’s first “virtual assistant” to be widely available on a smartphone, Siri drew on machine learning and “weak AI” to answer questions, retrieve information, and perform certain tasks. In the years since, the capabilities of Siri and other virtual assistants have been roundly surpassed by revolutionary platforms like ChatGPT, Dall-E, Claude, and Gemini. 

There are several different technologies within the broader AI umbrella. Each offers a discrete set of strengths and functionalities that will work for some companies and sectors better than others.

Machine Learning 

Machine learning (ML) is a type of artificial intelligence that uses large volumes of data to recognize patterns, make predictions, “learn” over time, and guide people toward making more precise, informed decisions. Instead of traditional computer programming, in which developers issue highly specific instructions for a computer to follow, machine learning is a more open-ended process that allows the computer to develop knowledge and expertise by poring over text, figures, photos, reports, and other types of data. While doing trailblazing work on AI in the 1950s, computer scientist Arthur Samuel defined machine learning as “the field of study that gives computers the ability to learn without explicitly being programmed.” 

Though other versions of artificial intelligence may be garnering more attention and investment dollars at the moment, machine learning remains a highly versatile and unusually proven version of the technology. MIT computer scientist and ML expert Aleksander Madry went so far as to assert to the MIT Sloan Management Review that “Machine learning is changing, or will change, every industry, and leaders need to understand the basic principles, the potential, and the limitations.” 

Machine Learning Applications

• Identifying business trends
• Developing recommendation algorithms
• Spotting supply chain slowdowns and bottlenecks
• Making medical diagnoses
• Providing customer service via online chatbots

Predictive AI

Within the increasingly expansive artificial intelligence landscape, there are a welter of tools, terms, and concepts that often have significant overlap with one another. Predictive AI, for example, actually uses machine learning—along with statistical analysis—to examine vast quantities of historical data and render effective, actionable predictions about patterns, behaviors, and even future events. 

Predictive AI is frequently seen as a more technologically advanced successor to predictive analytics, which requires more human engagement to utilize. Predictive AI, in contrast, can learn and develop completely autonomously. 

Predictive AI Applications 

• Predict consumer behavior
• Carry out financial forecasting
• Conduct demand forecasting
• Anticipate equipment failure
• Manage inventory 

Large Language Models (LLMs)

Large language models are a category of AI that analyzes exhaustive quantities of text and data in order to understand and produce language and fluently converse with human users. Today, LLMs are considered “foundation models.” Coined by Stanford University researchers in 2021, the term foundation model refers to large-scale AI platforms trained on enough data to make them proficient in a range of tasks and functionalities. As Georgetown’s Center for Security and Emerging Technology explains, rather than being developed for a narrowly defined purpose, “the original model provides a base (hence “foundation”) on which other things can be built.”

LLMs have exploded in popularity since the auspicious debut of Chat-GPT in late 2022. These foundation models are now used in innumerable ways—emails, reports, school papers—by everyone from students and teachers to researchers, content creators, and business professionals. Well-known alternatives to OpenAI’s platform include Google Gemini, Anthropic’s Claude, and Meta’s Llama.

Large Language Model Applications 

• Customer support
• Summarization
• Content creation
• Code generation
• Market research 

Generative AI 

As with machine learning and predictive AI, there’s significant overlap between LLMs and generative AI. The latter is a term that’s used to describe any artificial intelligence tool or platform that is primarily used for generating content. So while all LLMs are a form of generative AI, the reverse is not the case. In addition to generating text and engaging in natural conversation, generative AI models can create high-quality images, computer code, and even music and audio. Examples of generative AI tools that are not LLMs include DALL-E, Midjourney, AIVA, and IBM’s watsonx Code Assistant. 

Though generative AI is a relatively young form of artificial intelligence, it’s swiftly positioning itself as a transformative technology businesses are clamoring to effectively deploy. Market research firm IDC projects that organizations’ total global spending on generative AI will approach $40 billion by the end of 2024. 

Generative AI Applications 

• Marketing and advertising
• Content creation
• Computer programming
• Drug discovery
• Fashion design
• Product development 

AI Use Cases in Supply Chain Risk Management 

For organizations and teams interested in implementing AI in risk management, understanding the unique applications of each AI tool can be critical to effective decision-making. That’s because while generative AI and large language models may be currently attracting the most media buzz and corporate spending, they’re not necessarily the best fits for supply chain risk management (SCRM) needs. Instead, strategic sourcing experts and other SCRM professionals should consider the myriad functionalities offered by all the AI tools currently available to businesses, including predictive AI and machine learning.

A longstanding challenge in the field, comprehensive supply chain mapping can help manufacturers gain visibility into their suppliers, better understand the scope and nature of their sub-tiers, and identify risks before they materialize into costly disruptions. AI platforms can be an essential asset to this complex, arduous process. These tools are capable of locating product orders, freight bookings, transportation records, and other chain-of-custody documentation and synthesizing them to create detailed supply chain maps that visualize multiple tiers. By utilizing artificial intelligence models that draw on machine learning, professionals can enhance their supply chain visibility and strengthen their overall resilience. 

In addition to augmenting businesses’ mapping capabilities, AI can also be leveraged to anticipate equipment failures before they happen. Predictive analytics has been utilized for this exact purpose for years, and predictive AI represents a meaningful advancement of this existing technology. This form of artificial intelligence uses sensors that monitor a company’s equipment and infrastructure to extract large volumes of data on metrics like temperature, acoustics, humidity, pressure, and lubrication. 

After analyzing enough of this behavioral information, these platforms are able to recognize the equipment’s ideal operating conditions and alert maintenance professionals when those conditions start to deviate, deteriorate, or otherwise betray signs of substandard performance. This predictive maintenance functionality allows team members to identify precursors to equipment breakdowns or failures and respond proactively—a capability that enables manufacturers to avert the productivity sink of lengthy equipment downtime. 

Strategic sourcing professionals can also employ AI in risk management to rapidly assess the fallout from a supply chain disruption. Let’s say a direct supplier for an automotive manufacturer is forced to shut down several factories due to an extreme weather event. In response, the automaker can feed relevant historical data and current conditions into an artificial intelligence tool and receive an assessment of how the natural disaster is going to affect production and manufacturing timelines, including the length of delays. 

It’s important to underscore that this type of disruption forecasting is undoubtedly a raw, unrefined technology—an example of how AI in risk management remains in its infancy. Nevertheless, current artificial intelligence software can provide firms with objective, actionable estimates that can help leaders shape strategic responses. Possessing this type of intelligence, however imperfect, aids manufacturers in deciding whether to hold steady with the impacted supplier, reach out to an alternative supply chain partner, or implement other appropriate contingency measures. 

The Bold Future of AI in SCRM 

Indefatigable hype and stock market exuberance notwithstanding, LLMs, generative AI, and other bleeding-edge artificial intelligence tools are still in their earliest stages of development. While the unproven nature of these technologies certainly justifies caution, it doesn’t mean companies have to glue themselves to the sidelines while more brazen competitors bet big on their promise. Rather, supply chain risk management professionals should be diligently researching the potential impacts of different forms of AI on their business. Such investigational efforts include conducting a cost-benefit analysis, carefully studying applicable use cases, and even utilizing free demo periods from the leading AI firms. These due diligence measures can guide businesses to informed, deliberative company spending—and they start with understanding the different tools that comprise the rapidly evolving AI landscape.

Companies intrigued by the potential of AI in risk management may want to consider the predictive capabilities of SCRM platform Z2Data. The industry-leading software offers multiple forecasting functionalities, including end-of-life projections for electronic components with a historical accuracy of 90% and a sanctions watchlist that uses congressional reports, NGO investigations, and other high-credibility documents to highlight global businesses at heightened risk of becoming targeted by sanctions. Taken together, these predictive tools can help manufacturers become more risk-resilient and build an agile, proactive threat management strategy. 

To learn more about Z2Data and the forecasting features it provides its customers, schedule a free demo with one of our product experts.