AI and Supply Chains: From Reaction to Anticipation

In 2021 the consultancy AlixPartners estimated that the shortage of electronic chips would cost the global car industry about 210 billion dollars in revenue and lose about 7.7 million vehicles of production, an upward revision of its May estimate of 110 billion dollars. A component measured in millimetres halted a factory on another continent. This is the field that artificial intelligence is now entering.

Three rivers

Three rivers run through a supply chain: the river of goods, the river of money, and the river of information. The third is among the most common points of confusion: decisions built on late or scattered data, among suppliers who cannot see one another. A McKinsey survey (2024) of 88 senior supply-chain executives found that only about 60% had comprehensive visibility of their tier-one suppliers, and that visibility into deeper tiers had fallen for the second year running. If the third river is weak, any technology that strengthens it changes the picture.

What does artificial intelligence add?

Its most prominent contribution is the move from reaction to anticipation: estimating demand before it appears, spotting signs of failure, and proposing an alternative route when one closes. In a 2021 analysis, McKinsey reported that early adopters of artificial intelligence in supply-chain management had improved logistics costs by about 15%, inventory levels by about 35% and service levels by about 65%, compared with slower-moving competitors. That figure concerns early adopters and is published by a consultancy; it is not a controlled experiment, so we read it as a sign of direction, not a promise for every firm.

Reality is slower than the headlines. In the same 2024 survey, two-thirds of respondents were moving forward with advanced planning and scheduling systems, yet only 10% had completed deployment, while interest in artificial intelligence tools for demand planning is rising.

Efficiency or resilience?

The more a chain is tightened towards maximum efficiency and its stocks are cut, the less able it is to absorb a shock. Recent crises show this. And memory is short: in McKinsey's 2024 survey, 73% said they were making progress on dual sourcing and 60% were working to regionalise their chains, yet 46% expected to reduce or eliminate safety buffers, 47% would keep them as they were, and only 7% intended to raise them.

Artificial intelligence is therefore a double-edged tool: it helps manage complexity, and it tempts firms to tighten a chain beyond what it can bear. We should not leave it to the algorithm alone to decide how much safety stock to hold; that is a balance between cost and safety which belongs to management, and management bears the consequences.

Decisions and data

A growing share of daily decisions, such as freight pricing and load allocation, is left to systems. When a decision touches food or medicine, it is not enough that the algorithm "works": those responsible must understand why it decided as it did, must be able to switch it off, and anyone harmed must know who is answerable. We set out this principle in the article "Responsible Artificial Intelligence: Four Principles for Governance and Trust". And an algorithm without data is worth nothing, while the data of containers, ports and consumption is a form of wealth. Whoever gathers data from an entire chain may see what its individual parties cannot. So the question of who owns trade-flow data, where it is stored and under what controls it is shared should be settled early.

In Saudi Arabia

The Kingdom rose to 38th of 139 countries in the World Bank's Logistics Performance Index in the 2023 edition, from 55th in 2018 (Arab News, 2023). It launched the National Transport and Logistics Strategy on 29 June 2021, whose goals include placing the Kingdom among the top ten countries in this index and raising the transport and logistics sector's contribution to GDP from 6% to 10%, as the newspaper reported.

We take the view that linking ports, customs, transport and warehouses in consistent data is a prerequisite for any useful artificial intelligence, and that training national talent to understand the systems they operate is better than importing closed ones. The index alone is not evidence of artificial intelligence's effect on the sector, since it measures logistics performance in general.

A resilient chain is one in which trust is kept: contracts honoured, data told truthfully, and provision made for the unexpected. These are meanings that precede technology and outlast it.