Network dynamics in the age of AI

In our highly (inter)connected world, with the growing impact of AI on almost every facet of business, organizations must redefine, cement, and extend not only their business models but also their sources of competitive advantage.

Sustainable competitive advantage refers to a company's ability to maintain its market position and profitability over the long term, despite competition¹. Various terms describe this concept, each highlighting different aspects of competitive durability and uniqueness. Here are some of the commonly used terms:

Any one company can leverage any number of these to build its competitive advantage, and where things get interesting is in the potential interplay between them. We will focus on supply chain networks and how they can be a source of multiple avenues for competitive advantage, from creating economic moats and lifting barriers to entry, to competitive differentiation and strengthening network effects.

Building Supply Chain Networks for Competitive Advantage

In simple terms, network effects mean that the more something is used, the more valuable it becomes. However, for our purposes, we want to use the term Network more liberally to mean a structure with edges and connections. This structure can take the form of social networks (like LinkedIn) or supply chain networks, logistics, and partnerships that become stronger through the interaction of their nodes.

A company’s supply chain and network of partners and suppliers can be an immense source of competitive advantage. Take, for example, ASML, the world’s leading (and only) manufacturer of high-end lithography equipment. This equipment is fundamental in producing advanced semiconductors, like NVIDIA’s GPUs.

ASML’s competitive advantage comes as much from its technological IP as it does from its highly complex supply chain, encompassing over 4000 suppliers, custom packaging, transportation, and services structure, which makes that technical IP economically viable in the first place. Many companies supplying components and materials for ASML machines have long-term agreements and, in many cases, exist exclusively to supply ASML.

Any company wishing to compete with ASML has not only the enviable task of inventing technology capable of pushing the boundaries of physics but would also have to deal with the almost equally complex task of establishing a network of partners and suppliers required to bring that technology to the market at scale. Needless to say, the barrier to entry here is enormous, and to a large degree, is due to the vast, established ecosystem that underpins ASML’s technology in the first place.

Another example of the power of supply chain network effects and derived competitive advantage is Amazon, whose switch from bookstore to global retail and cloud juggernaut could not have been possible without a highly sophisticated partner, logistics, and supply chain network.

In the next sections, we will explore the areas where modern technologies like AI can impact and further advance a firm’s network-derived competitive advantage. The focus is on supply chain networks, as we consider this the domain with the most to gain from leveraging AI and data sharing innovations.

Improving supply chain network interactions with AI Agents

In a previous blog, we discussed the potential impact of AI on an organization's internal processes and operations. A logical next step we explore here is how a business can extend this potential beyond the internal boundaries of the organization and into its network. One of the main ideas gaining mainstream traction in the current landscape of AI is the concept of AI agents. These agents are, in general, specialised models often augmented by tools and other components that work together, each executing its designated task to achieve a global objective. This idea is also known as Compound AI systems and starkly contrasts with having a single, monolithic general model.

One of the most promising applications of these agents or compound AI systems is streamlining and improving interactions within and across networks, notably supply chain networks, where hundreds or thousands of handshakes between systems are often done manually. A few of the areas where these systems can change the way these interactions and integrations take place are:

• Plant Operations: Automating the transition and triggering the required processes across activities such as inventory management, assembly, packaging, and delivery.
• Inventory management: By integrating a set of agents that leverage internal and external data for forecasting and can link this to inventory levels, and trigger relevant work orders and stock actions. For example, recommending alternative suppliers or regions less affected by tariffs, enabling rapid scenario planning, dynamic sourcing, and optimal inventory positioning to minimize costs and disruptions.
• Logistics: Analyzing factors like traffic patterns, weather conditions, and delivery schedules to determine the most efficient routes, reduce fuel consumption, lower transportation costs, and ensure timely deliveries.

Networking and collaborating across organizational boundaries

A crucial aspect of the type of supply chain networks we have been discussing is that they are made up of multiple parties. Even if one party may ultimately act as the overall integrator (e.g., assembling and distributing the final product), many other parties with their own sub-assembly and sub-distribution participate in this process. Effective integration and coordination between these parties is, therefore, crucial for success.

However, challenges emerge because of the difficulty in integrating various evolving technology stacks, data silos, protocols and organizational processes that delay the availability of information and hinder the possibility of making the best decisions at every stage. Given this reality, building strong networks for collaboration presents itself as a fundamental element to improve and streamline complex supply chains.

One of the first barriers to overcome is how to effectively and efficiently share data (information) among network partners. Today, much of this data remains locked away in on-premises systems and proprietary formats that do not integrate well (if at all) with each other. Additionally, many datasets are so siloed and segregated across tools that unified governance is impossible. When the setup is riddled with technical limitations and enforcing any kind of organizational process around data sharing is nearly impossible, companies create barriers to competitive advantage.

The path forward, as shown in recent years, is moving toward common open data standards, such as Iceberg and Delta, that enhance interoperability across systems and organizational boundaries. The emergence of these formats has also given rise to the development of open sharing protocols, like Delta Sharing, which allows organizations to securely share data internally and externally, across regions, clouds, and even on-premises sources through the use of federation - without the necessity of physically moving or duplicating data.

These technologies enable many game changing business advantages, from the creation of private exchanges (multiple parties can host and make data available to selected partners) to collaboration through cleanrooms (ephemeral private environments that allow parties to work on a common dataset without having to physically share or even disclose the data with each other).

Circling back to competitive advantage, we can now extend our view on how building strong networks - and the dynamics and effects those networks generate - can be evaluated and measured. Here are some metrics often used in network analysis (most commonly social network analysis) but applied to the specific context of supply chain:

• Degree Centrality: This measures the number of direct connections a firm has within the supply chain network. Firms with high centrality are often central to information flow and resource allocation, making them crucial for network stability and efficiency.
• Betweenness Centrality: This metric identifies firms that act as bridges between different network parts, controlling the flow of goods or information. High betweenness centrality can indicate a firm's strategic position, but also poses risks if disruptions occur.
• Eigenvector Centrality: This measures a firm's influence based on its direct connections and the connections of its partners. Firms with high eigenvector centrality have significant transitive influence, meaning their performance can affect other critical firms in the network.
• Closeness Centrality: This assesses how quickly a firm can reach all other firms in the network. Firms with high closeness centrality are well-positioned to respond quickly to changes or opportunities across the supply chain.
• Network Density and Intensity: These metrics describe interconnectedness and communication frequency among supply chain partners. Higher network density and intensity can enhance collaboration and reduce communication barriers, contributing to a competitive advantage.

By applying these metrics, businesses can identify strategic positions within their supply chain networks, optimize collaboration, and enhance their competitive advantage through effective network management. The data and AI technologies discussed in this blog can strategically improve an organization’s position across the above metrics.

Conclusion

Creating these stable edges between partners, suppliers, and other parties in the network plays a critical role in improving and developing the next generation of supply chain and logistics. Data accessibility and sharing via a modern platform stack, augmented by the use of AI, will allow companies to:

• Make better decisions: With relevant, timely data available, intelligent agents and automated processes can act.
• Reduce operational costs: Streamline platform and joint network architecture to reduce integration costs and minimize or eliminate the need for manual intervention.
• Increase competitiveness: By enabling the development of strong integrated networks, accelerating innovation, and unlocking new business models.

While we focused on collaboration between companies and partners in this post, this thinking clearly also applies to internal management strategy between company business units and divisions in support of cross functional communications and collaboration.

To learn more about how Databricks can enable better sharing and collaboration visit https://www.databricks.com/product/delta-sharing

Reach out to the authors to discuss how AI can support the next generation of supply chain networks.

¹ Source: All Revenue is Not Created Equal: The Keys to the 10X Revenue Club | By Bill Gurley.