At the height of the coronavirus pandemic, most of the world’s attention was on the race to develop a vaccine. But even after the mRNA vaccine was developed—processes that typically take several years were condensed into a few months—countries faced another significant obstacle, one that has long plagued immunization and public health efforts: How do we get these life-saving vaccines to the people who need them the most, as quickly and efficiently as possible?
The slogan “No product, no program” is used by many vaccine experts as a reminder that, without a consistent and reliable supply chain—the network of staff, equipment, vehicles, and data that carry a vaccine from the manufacturer to a patient—vaccine programs will fail. This is not merely theoretical; we have seen the success of efforts to contain diseases such as Ebola, polio, and smallpox hinge on the reliability of supply chains that can safely and reliably manage, store, transport, and deliver vaccines to people.
During the pandemic, however, low- and middle-income countries (LMICs), particularly those in the Global South, faced a unique set of challenges. Unlike their wealthier counterparts, they couldn’t rely on vast resources to buffer against inefficiencies. Their primary hurdle was the challenge of predicting demand and matching it with their often-limited supply, straining already overburdened health-care infrastructures and leaving many people without access to vaccines. The disparities were stark: While today high-income countries boast vaccination rates of nearly 80 percent, only about 33 percent of people in low-income countries received a single dose.
While the pandemic highlighted the vulnerabilities of LMICs, it also revealed a curious trend among nations with the potential for innovation: doubling down on traditional methods. The Global Vaccine Action Plan, introduced in 2012 as a synchronized public health response to the aftermath of previous health crises, surprisingly anchors itself to a logistics framework conceived in the 1970s. India’s decision in 2021 to earmark a staggering $1.4 trillion in projects to bolster its supply chain infrastructure over the next five years serves as another case in point. Such capital-intensive solutions, while foundational for growth, are increasingly being viewed with skepticism in an era when emerging technologies offer potentially more agile and cost-effective alternatives.
A New Solution
Traditional supply chains operate on a fixed, predictable model with predetermined routes. Unforeseen disruptions can generally lead to significant delays. In this model, the entire system is designed around a central hub, with spokes extending to various destinations. This hub-and-spoke design, while efficient in a stable environment, struggles to adapt quickly to changes or disruptions.
But the AI-driven supply chain, underpinned by more adaptive and responsive systems, continuously analyzes real-time data to make dynamic routing decisions. If there’s a sudden change in demand in one location or an unexpected obstacle on a particular route, the AI-driven system can adjust, redirecting resources or altering routes as needed.
“AI can predict resource requirements and optimize inventory placement,” says Derek Szopa, CEO at CloudSort, an AI supply chain start-up. “It can also help identify inefficiencies and automate repetitive tasks, freeing up time for more strategic decision-making.”
In practical terms, if there’s a sudden surge in vaccine demand in one city, AI can anticipate this need using predictive analytics and redirect supplies accordingly, all while accounting for factors like storage conditions, transportation times, and local infrastructure.
AI and machine learning are poised to be the technology that moves current vaccine distribution and supply chain systems into the future. According to an IBM report, 46 percent of surveyed supply chain executives anticipate that their greatest areas of investment in digital operations over the next three years will be in AI, cognitive computing, and cloud applications.
In Tanzania, the vast terrain and remote populations complicated the task of ensuring all individuals received their dose of the COVID-19 vaccine. Traditional methods of tracking vaccines, often based on outdated data, were proving inadequate. Pendulum, an AI-focused company, had already begun exploring innovative solutions for vaccine distribution even before the pandemic. In 2018, the company joined forces with the Ministry of Health in Tanzania to sift through public data, government records, and satellite imagery to predict vaccine utilization using the Connected Health AI Network (CHAIN). The outcome was significant: Three areas in Tanzania that implemented CHAIN saw vaccine wastage drop by more than 96 percent, showcasing the system’s heightened accuracy in predicting demand.
“With vaccines, one of the major elements is being able to not only forecast demand but also properly allocate that supply to places that are equipped to receive it,” says Brittany Hume Charm, head of growth, global health at Pendulum.
Uganda faced a similar set of hurdles. The reliance on paper-based forms by health-care workers meant that there was no centralized digital overview of the vaccine supply. This lack of clarity often led to last-minute vaccine replenishments, risking the efficacy of the doses. Logistimo, an AI-driven logistics solutions company, stepped in with a solution that seemed deceptively simple: a vehicle-routing product. This tool, by using machine learning to identify optimal delivery routes and adjust deliveries based on real-time needs, improved the efficiency of the distribution process. As a result, warehouse managers could consistently meet the vaccine demands, facilitating timely and effective distribution.
“Health workers are deeply occupied on the clinical side of things. Paperwork being a cumbersome task, reporting was often irregular, incomplete, or cooked—and planners upstream relied on poor data on which to base decisions,” says Anup Akkihal, CEO of Logistimo.
The vehicle-routing solution by Logistimo represents a paradigm shift from the colossal, capital-intensive infrastructures of the past. Gone are the days when supply chain improvements necessitated the construction of expansive railroads or sprawling shipping ports. Instead, the modern approach, empowered by AI, is about agility and precision. It’s about making the most of existing resources, optimizing routes, and executing every delivery, no matter how remote the destination, with precision.
Navigating the Obstacles
Despite the promising strides made by AI in revolutionizing vaccine supply chains in Africa, a number of hurdles stand in the way of the AI-driven supply chain successfully scaling across LMICs, particularly in the Global South.
The first major challenge is data. In health care, data is the lifeblood of AI. However, the difference between LMICs and higher-income countries in data availability is extreme. Dykki Settle, chief digital officer of PATH, a global health nonprofit, draws attention to this by contrasting the “data jungle” of high-income countries with the “data desert” scenario more prevalent in many LMICs. Without the right kind of data, AI systems risk perpetuating or even introducing biases. The World Health Organization (WHO) underscores this concern, noting the frequent exclusion of marginalized groups from AI training datasets. This data gap is further widened by the digital divide, leaving millions, especially women, disconnected.
Emerging from the data challenge, the next hurdle is system fragmentation. In many LMICs, digital health initiatives operate in silos, making data integration akin to “a form of investigative journalism,” says Hume Charm. For AI to truly flourish, these isolated systems must communicate seamlessly.
Solutions like Kenya’s M-TIBA platform hint at the potential of interconnected systems, bridging the gaps and fostering collaboration. By connecting patients, health-care providers, and payers through a single platform, M-TIBA reduces the chances of data redundancy and provides health-care providers with a holistic view of a patient’s history, leading to better-informed decisions. So far, the platform has helped to connect more than 4 million users and 1,200 health-care providers in the country, according to a McKinsey & Company report.
Finally, there’s the problem of “technological solutionism,” a perspective in which technologies like AI are seen as a magic wand to fix issues in LMICs in the Global South. Most developments in AI and machine learning are provided to LMICs by companies from higher-income countries in the Global North. But many public health organizations like PATH are wary of external vendors who come in, provide new technological solutions, and then leave the scaling and management to the local governments and health workers who may be overworked and underprepared for the job—culminating in the failure of that solution. Settle, from PATH, emphasizes that to avoid an “overreliance on external vendors,” supporting local innovators and helping governments become self-reliant in the creation and use of digital systems should be a priority when implementing AI-backed health systems in LMICs. A deeper understanding of the cultural, structural, and institutional barriers is necessary for those systems to successfully integrate into the country’s health-care infrastructure—so as to avoid becoming part of what Dr. John Bawa, team lead for vaccine implementation in Africa at PATH, calls “closets full of dumped technologies.”
To handle these challenges effectively, some countries have adopted a collaborative approach. For instance, in some regions, local health-care professionals have been paired with international tech developers to co-design tools that are both technologically advanced and culturally apt. Additionally, pilot testing and iterative feedback loops have been used extensively to align technologies with the needs and values of the community. Such strategies support not only better integration but also the longevity and relevance of the technologies introduced.
For instance, the Mobile Alliance for Maternal Action used mobile technologies to deliver timely health information to new and expectant mothers in countries such as Bangladesh, South Africa, and Nigeria. These initiatives positioned local medical professionals and mothers to play a vital role in tailoring the content to the cultural and infrastructural realities of each region. This way, the technology became integrated into the fabric of the community, rather than just becoming another “dumped” solution.
The future of AI for health care and vaccine distribution among LMICs in the Global South depends on the ability of innovators to create solutions that are as lean and creative as possible and work offline, reduce biases, and take advantage of local expertise—but their breakthroughs have the potential to affect the rest of the world as well.
“All of the tools that we designed to work offline in rural Liberia are applicable anywhere in the world,” says Peter Lubell-Doughtie, co-founder and CTO of Ona, a company that provides data solutions to humanitarian relief and first response organizations. “Designing for that hardest use case helps us bring best practices everywhere else.”
A Way Ahead
The coronavirus pandemic exposed several deep cracks in the health systems of most countries. But it wasn’t the first pandemic, and it certainly won’t be the last: Climate change may cause future pandemics because of cross-species infections, and much faster than expected. However, public health experts who’ve worked in the field note that apart from periods of crisis, supply chain optimization is not always a priority.
“The whole world forgot about Ebola because we thought it was limited to parts of Western Africa. But it became a global epidemic by snowballing in a similar way that COVID-19 did,” says Dr. Bawa. “If we had learned lessons from it, we could have been more proactive and more resilient and would have prevented the significant loss of life that we saw across the globe.”
During the pandemic, Pendulum’s collaboration with the state of California and the government of Sierra Leone showcased the potential of AI in addressing supply chain challenges. Confronted with outdated or incomplete data on cold-chain infrastructure, Pendulum bypassed traditional, time-consuming methods of data collection. Instead, it harnessed an enhanced web-scraping tool, analyzing satellite imagery combined with government datasets. This innovative approach not only identified the pandemic preparedness of health facilities but also pinpointed the location of six new COVID-19 laboratories in Sierra Leone. The same technology was later expanded to projects related to malaria and family planning in collaboration with the governments of Côte d’Ivoire and the Democratic Republic of the Congo.
Yet, the challenges transcended mere logistics. Vaccine hesitancy, defined by the WHO as the “delay in acceptance or refusal of vaccines despite availability of vaccination services,” posed a significant barrier. Misinformation, mistrust in public institutions, and cultural differences led to reluctance in vaccine uptake in various regions. In response, Pendulum and PATH used AI to monitor digital traces like social media posts and identify areas where misinformation was rampant. This data-driven approach allowed governments and public health agencies to tailor their communication strategies to those regions, ensuring efficient vaccine distribution and uptake.
“Demand forecasting is not as simple as just getting a vaccine from point A to point B. You have to also make sure that someone is willing to walk into a clinic and get it,” says Hume Charm.
Learning lessons from the past and creating next-generation supply chains to prepare for an inevitable future where pandemics will be more common is a key area of focus in the WHO’s Immunization Agenda 2030: “Strengthen supply chains to ensure that high-quality vaccines are always available in the right quantity and form at the right time, in the right place and stored and distributed under the right conditions. Promote integration with other supply chains for more effective delivery of primary health care. Invest in systems and infrastructure to safely manage, treat, and dispose of vaccine waste to help reduce their environmental footprint.”
Ultimately, for AI to be a part of the pandemic-preparedness plan for vaccine supply chains requires a systemic paradigm shift: one that accepts both that humans may not have all the answers needed to allocate scarce resources in the most optimized, equitable way possible, and that training AI and machine learning to respond to the volatility of public health scenarios may reveal outcomes we’ve never imagined—outcomes that help health workers reduce senseless deaths.
“Each vaccine that you place at a given location at a given time is a bet about demand. And when you have a limited war chest of resources, you need each of those bets to be extremely high quality,” says Benjamin Fels, CEO and co-founder of Pendulum. “In every industry where AI and machine learning have been allowed to learn the answer, it has outperformed any other approach by orders of magnitude. For me, improving a supply chain where allocation is a matter of life or death is the most meaningful use of AI.”
