The real challenge in applying artificial intelligence (AI) to supply chains lies not in the sophistication of AI models but in the quality and integration of underlying data. A remark made by a Fortune 500 banking CEO at the recent Databricks Data + AI Summit encapsulates this truth: “The models aren’t the hard part. It’s the data.” This insight is especially relevant in the complex environment of supply chain and manufacturing industries, where data is notoriously fragmented across numerous sources such as ERP systems, IoT sensors, procurement platforms, spreadsheets, and third-party logistics providers.
Many organisations aiming to harness AI gains recognise that chasing cutting-edge algorithms without resolving their data foundations is futile. Instead, successful enterprises in sectors from fashion retail to semiconductors are focusing their efforts on establishing connected, AI-ready data ecosystems that enable real-time operational visibility and decision-making. Technologies like Google Cloud and Databricks on the Google Cloud Platform exemplify the infrastructure providing scalable analytics and unified data management, bridging structured, semi-structured, and unstructured data formats crucial for actionable AI.
The solution “Planning in a Box – Pi Agent,” built on Google Cloud and Databricks, is emerging as a potent platform designed to solve the data integration problem for supply chains. Its architecture centres on a master ledger that harmonises disparate data streams from an organisation’s ERP, CRM, warehouse systems, and sensor data, enabling AI agents with specific roles—such as demand forecasting, inventory optimisation, defect detection, and financial insights—to continuously learn and provide not just diagnostics but actionable recommendations. By operating natively within an organisation’s cloud environment, Pi Agent safeguards sensitive data while delivering faster, more responsive supply chain management, evolving planning from quarterly cycles to dynamic daily or hourly adjustments.
This approach aligns with broader industry trends underscored by multiple analyses revealing that supply chain AI struggles primarily due to siloed, poor-quality, or inaccessible data rather than lack of technology. For instance, the expansion of global trade, which has tripled the value of intermediate goods since 2000, has complicated logistics networks, making real-time monitoring critical yet difficult. Traditional tracking technologies—GPS, RFID, Transport Management Systems—offer partial visibility, yet fall short of comprehensive transparency across multi-tiered, multinational supply chains.
Advanced AI-driven visibility platforms termed “control towers,” employing machine learning and generative AI, strive to fill this gap by offering preemptive disruption alerts and optimisation insights. However, full end-to-end visibility remains elusive in practice, often due to a lack of willingness among companies to share sensitive supply chain data. Trust and collaboration barriers arise because stakeholders—suppliers, manufacturers, logistics providers, and retailers—use separate, often incompatible data systems, with concerns over competitive advantage, privacy, and security limiting data exchange. Blockchain has been proposed as a secure mechanism to facilitate data sharing, but widespread adoption is still confined to pilot phases, showing that cultural and organisational change remains as significant a barrier as technological readiness.
Moreover, firms face multiple challenges in scaling AI beyond proofs of concept. These include data ownership ambiguities, fragmentation caused by functional silos, outdated legacy systems, lack of a clear transformation strategy, and shortages of AI and data talent. The difficulty is not just acquiring data, but ensuring its accuracy, completeness, and relevance. Poor data input leads to flawed AI outputs, which can undermine trust in AI-driven decision-making—a concern highlighted by companies transitioning from experienced human planners to automated systems.
Alongside data access and quality issues, operational constraints such as the costs of AI infrastructure, bandwidth requirements, and ongoing maintenance weigh on organisations. Yet, when deployed effectively, AI solutions have demonstrated the capacity to optimise inventory levels, enhance demand sensing by incorporating macroeconomic trends and external signals like weather and promotions, improve defect detection and quality control with AI vision, and enable rapid scenario planning to anticipate tariff impacts, supplier delays, or demand surges.
Ultimately, the path to intelligent, autonomous supply chains is paved by investments in robust, integrated data foundations rather than merely the latest AI algorithms. Companies that move quickly to consolidate fragmented data sources and embed AI-driven decision layers within their core systems stand to outpace competitors, turning supply chain planning into a continuous, agile capability rather than a periodic task. The message is clear from industry leaders: AI success in supply chains starts with solving the data problem, and platforms like Planning in a Box – Pi Agent show how this can be translated into business advantage, delivering not just technological innovation but real operational impact.
Source: Noah Wire Services
