JEDDAH: As artificial intelligence shifts from experimentation to real-world deployment, Saudi supply chain and enterprise leaders are embedding advanced tools into operational decision-making, forecasting and resilience strategies.
For Yahyah Pandor, vice president and general manager for MENAT at Blue Yonder, this transition reflects a structural shift rather than a passing trend.
In an interview with Arab News, Pandor detailed how AI is transforming forecasting, localization strategies, ERP integration and supply chain resilience across the Kingdom.
Saudi Arabia’s fast-expanding retail and FMCG sectors face a complex demand landscape shaped by omnichannel behavior and macroeconomic shifts.
“For Saudi retail and fast-moving consumer goods, a layered approach, not a single model, offers the strongest results. The most effective stack today is machine-learning forecasting, demand sensing, lead-time and variability prediction, and AI decision layers embedded inside planning workflows,” Pandor said.
He emphasized that volatility in the Kingdom is both macroeconomic and behavioral, with consumers increasingly adopting omnichannel habits.
“In practice, that means combining external signals like weather, promotions, competitor pricing and social trends with explainable AI, so forecasts can improve replenishment, inventory and service decisions,” he explained. “When forecasting is connected to execution, and not isolated in dashboards, it delivers the greatest value through better availability, lower inventory distortion and faster response to volatility.”
As Saudi Arabia accelerates economic diversification under Vision 2030, localization is reshaping supply chain strategy.
“To support localization under Vision 2030, supplier selection needs to move from lowest-cost sourcing to multi-objective optimization,” he said.
“Practically speaking, AI models should rank suppliers not only on price and quality, but also on local-content contribution, supplier-development potential, resilience, lead-time reliability and long-term capability transfer.”
DID YOU KNOW?
• Saudi firms are now combining weather, promotions, pricing and social trends with AI to improve forecasting accuracy.
• The most advanced supply chain systems in the Kingdom are using multiple AI layers, not a single mode.
• AI in Saudi supply chains is shifting supplier selection from lowest cost to multi-factor scoring.
Pandor argued that localization must be embedded into supply chain design itself.
“So, any network design should embed localization directly into the optimization objective to ultimately decide where to place inventory, postponement nodes or light manufacturing capacity based on service, risk and domestic value creation. Organizations that do this well will treat localization as a design variable, not a reporting exercise after the fact.”
AI-powered digital twins are also gaining traction, particularly in logistics and infrastructure.
“Adoption is undeniably emerging in Saudi Arabia. However, the public evidence today is stronger in logistics, infrastructure and public-service environments than in disclosed retail or FMCG case studies,” he said.
Pandor cautioned against overstating predictive precision.
“Yet, I would avoid overstating predictive precision. Public Saudi disclosures are still thin on a single ‘accuracy’ number,” he noted.
“So, the honest answer is that adoption is real, but most organizations are measuring value through faster scenarios, earlier risk detection and better decisions (not publishing a neat precision score).”
For enterprises operating complex legacy ERP systems, the challenge is architectural rather than mathematical.
“Embedding advanced AI into legacy ERP estates is very achievable, but it is rarely straightforward. The main difficulty is not the model; it is the architecture around it,” Pandor said.
“AI needs structured access to ERP data such as orders, inventory, suppliers and schedules, plus controlled ways to write decisions back into workflows.”
He added that phased modernization offers the most practical route.
“It consists of cleaning up master data, exposing APIs, adding orchestration and observability, and then deploying domain AI into planning, procurement and inventory,” he said. “Seamless ERP and MES integration reduces duplicated data and improves real-time planning quality.”
While fully autonomous, “self-healing” supply chains remain a long-term goal, Pandor said Saudi organizations are currently in a guided-autonomy phase.
“We are seeing early movement toward autonomous planning, but the market is still in a guided-autonomy phase rather than true self-healing, lights-out planning,” he said.
“The near-term value for Saudi organizations can be found in always-on exception management, parameter tuning, demand sensing and governed decision layers on top of APS or IBP, not in removing planners from the loop.”
Opinion
This section contains relevant reference points, placed in (Opinion field)
In practice, AI is augmenting planners rather than replacing them — for now.
“In essence, adoption is beginning, but it requires high technical maturity before autonomy can be trusted at scale,” he added.
Pandor identified operational readiness — not algorithms — as the main barrier to scaling AI.
“The MLOps gap is significant. It is less about algorithms, and more about operational readiness,” he said.
“For Saudi enterprises, the challenge is rarely whether leadership believes in AI, but whether the business can industrialize it.”
He advocates a shared platform model to scale effectively.
“The operating model that scales best is a shared platform model, including one common data backbone, alongside one governance layer for models, controls and monitoring, and business-domain teams that can deploy use cases locally.”
Amid geopolitical uncertainty and supply disruptions, resilience engineering is becoming central to enterprise strategy.
“Technically speaking, probabilistic modeling and reinforcement learning solve different layers of the resilience problem,” Pandor said. “Probabilistic models help planners move from single-point assumptions to confidence ranges for demand, lead times, supply availability and recovery time.”
“Reinforcement learning is useful for one layer above that, inside a simulator or control-tower environment, where the system can learn which response policy works best under disruption.”
He highlighted the scale enabled by modern platforms.
“Our platforms process more than 25 billion AI predictions a day, which demonstrates the scale now possible when sensing, prediction and action are connected.”
He added: “As a result, we’re confident in saying that resilience improves when automation is bounded, explainable and tied directly to execution.”
As Saudi Arabia continues to localize industries, modernize infrastructure and expand non-oil growth, AI is moving from pilot programs into the operational core of enterprise planning.
