AI-Controlled Supply Chain Decision Systems and Market Shaping

Introduction

AI-controlled supply-chain decision systems use machine learning, predictive analytics, automated procurement tools, demand forecasting, dynamic pricing, inventory optimisation, logistics-routing systems, supplier-scoring models, and platform algorithms to determine what is purchased, from whom, at what price, in what quantity, and through which distribution channel.

From a competition-law perspective, the important issue is not simply that AI is being used. The concern arises when control over an AI-driven supply-chain system enables an undertaking to shape market conditions, exclude suppliers or rivals, coordinate competitors, discriminate between trading partners, control access to essential inputs, or reinforce vertical market power.

Competition authorities increasingly recognise that algorithmic systems can facilitate coordination, discriminatory ranking, self-preferencing and exclusionary conduct. The OECD has specifically identified algorithmic collusion, hub-and-spoke arrangements and AI-enabled exclusion as emerging competition concerns.

1. Meaning of AI-Controlled Supply-Chain Decision Systems

An AI-controlled supply-chain system may perform several interconnected functions:

1. Supplier selection – ranking suppliers according to price, reliability, quality or risk.
2. Procurement allocation – automatically determining how much business each supplier receives.
3. Demand forecasting – predicting future demand and adjusting orders.
4. Inventory management – deciding stocking levels and replenishment.
5. Dynamic pricing – automatically modifying prices according to demand and competitor information.
6. Logistics allocation – selecting routes, carriers and warehouses.
7. Platform ranking – determining which suppliers or products consumers see.
8. Credit and risk scoring – deciding which suppliers receive favourable payment or financing conditions.
9. Contract optimisation – generating or adjusting commercial terms.
10. Market-entry decisions – predicting where a company should expand or withdraw.

The system therefore becomes more than an operational tool. It can become a mechanism through which market opportunities are allocated.

2. How AI Can Shape the Market

A. Supplier exclusion

An AI procurement system may consistently rank certain suppliers below affiliated or preferred suppliers.

If the system uses objectively relevant criteria, this may generate legitimate efficiency. But if a dominant undertaking deliberately designs the system to disadvantage independent suppliers, competition concerns may arise.

B. Self-preferencing

A vertically integrated platform may operate both:

• the marketplace through which suppliers reach customers; and
• its own competing supply or retail business.

The platform can then use its AI system to favour its own products.

This is particularly significant because the algorithm may influence visibility without an obvious contractual exclusion.

C. Information advantages

AI systems can process enormous quantities of:

• supplier prices;
• inventory information;
• consumer demand;
• competitor behaviour;
• sales forecasts;
• delivery performance; and
• commercially sensitive information.

A dominant platform possessing this information may gain an informational advantage over dependent suppliers.

D. Algorithmic coordination

Several competing firms may use the same algorithm or common software provider.

If the system incorporates competitors' commercially sensitive information and facilitates coordinated pricing or output decisions, traditional cartel rules may become relevant. The OECD distinguishes explicit algorithm-assisted collusion, hub-and-spoke coordination and the more difficult question of autonomous algorithmic coordination.

3. Principal Competition-Law Theories

A. Abuse of Dominant Position

Where an undertaking has substantial market power, AI-controlled supply-chain decisions may constitute:

• discriminatory treatment;
• refusal to supply;
• exclusive dealing;
• margin squeezing;
• tying;
• self-preferencing;
• unfair trading conditions;
• predatory strategies; or
• exclusionary access restrictions.

The AI itself does not create liability. The legal question remains what the undertaking does through the system and what competitive effects result.

B. Cartel and Concerted-Practice Risks

AI may reduce the need for direct human communication between competitors.

For example:

Supplier A and Supplier B independently adopt the same pricing algorithm, which continuously observes market prices and adjusts prices according to the same objective.

If the algorithm is merely independently optimising prices, parallel outcomes do not automatically establish an unlawful agreement. But if competitors intentionally use a common mechanism to exchange sensitive information or implement an agreed strategy, traditional cartel principles can apply.

The OECD has emphasised that established competition-law principles continue to apply when algorithms are used as instruments for implementing collusive agreements.

C. Vertical Foreclosure

AI may be used to make access to a distribution network conditional upon:

• exclusivity;
• minimum-volume requirements;
• use of affiliated logistics;
• use of an affiliated payment system;
• acceptance of algorithmically imposed prices; or
• compliance with restrictive platform rules.

This can prevent alternative distributors or suppliers from reaching sufficient scale.

4. Market Shaping Through Algorithmic Allocation

The most important conceptual issue is algorithmic allocation of commercial opportunity.

Traditional competition law normally examines contracts, prices, market shares and observable conduct.

AI adds another layer:

Data → Model → Ranking → Allocation → Market outcome

For example:

Supplier data

↓

AI risk/quality model

↓

Supplier ranking

↓

Procurement allocation

↓

Supplier revenue

↓

Supplier investment capacity

↓

Future market structure

Thus, a procurement algorithm can potentially influence not merely today's transactions but which suppliers survive and expand tomorrow.

5. Case Laws

1. United States v. Amazon Marketplace Sellers / Online Poster Sellers

This group of enforcement examples demonstrates how automated pricing software can implement an underlying agreement between competing sellers.

The sellers agreed not to undercut one another and used automated repricing software to implement the arrangement on Amazon's marketplace. The algorithm therefore functioned as an execution mechanism for collusion, rather than eliminating the underlying antitrust issue.

Principle

An undertaking cannot avoid cartel liability merely because the agreement is implemented through software rather than human instructions.

Relevance

The same reasoning can apply to AI-controlled supply chains where competing manufacturers use a common algorithm to coordinate:

• procurement prices;
• wholesale prices;
• production quantities;
• delivery terms; or
• allocation of customers.

2. Trod Ltd / GB Eye Ltd — UK Online Poster Pricing

The UK competition enforcement concerning online poster sellers is an important algorithmic-collusion example.

Competing sellers used automated repricing software after agreeing not to undercut one another. The software continuously monitored prices and implemented the agreed pricing strategy.

The case illustrates the distinction between:

lawful automation of independent decisions

and

automation used to implement an unlawful agreement.

The OECD identifies Trod/GB Eye as an early example of algorithm-facilitated coordination.

Supply-chain relevance

The same mechanism could theoretically operate in procurement or distribution:

competitors agree on commercial parameters → AI monitors deviations → AI automatically restores the agreed market position.

3. Google Shopping — European Commission

The Google Shopping decision is highly relevant to AI-controlled ranking systems.

The European Commission found that Google systematically gave favourable positioning to its own comparison-shopping service while competing comparison-shopping services were subject to algorithms that reduced their visibility.

The UK's competition-policy discussion of algorithms identifies Google Shopping as a seminal example of algorithmic self-preferencing.

Principle

An algorithmically generated ranking can constitute a competition concern when a dominant undertaking uses control over a platform to favour its own service and disadvantage rivals.

Supply-chain relevance

The same principle can apply to:

• supplier ranking;
• warehouse allocation;
• procurement marketplaces;
• logistics platforms;
• manufacturer visibility; and
• AI-generated product recommendations.

4. Amazon Marketplace — European Commission

The European Commission's Amazon investigation examined Amazon's dual role as:

1. marketplace operator; and
2. retailer competing with third-party sellers.

The Commission highlighted Amazon's access to non-public data generated by independent sellers and the importance of the Buy Box in determining seller visibility and sales.

Competition significance

This creates a structural conflict:

Platform controller

→ receives competitor/supplier data

→ operates algorithmic ranking

→ competes against those suppliers

→ determines commercial visibility.

Supply-chain relevance

An AI-controlled supply-chain platform could similarly use third-party supplier information to improve its own competing products or services.

This can produce a data-feedback loop:

supplier data → AI optimisation → platform advantage → increased sales → more data → greater AI advantage.

5. FTC and States v. Amazon

The US FTC and state attorneys general sued Amazon alleging a series of practices designed to maintain Amazon's market power, including conduct affecting sellers, pricing, competition and marketplace dynamics. The allegations include mechanisms that allegedly prevent rivals and sellers from competing effectively on price and other dimensions.

Competition significance

The case illustrates how a large platform's interconnected technological and commercial mechanisms may be examined collectively rather than as isolated algorithmic decisions.

Supply-chain relevance

Where AI controls:

• seller visibility;
• pricing;
• fulfilment;
• advertising;
• inventory;
• customer access;

the competitive effects may be cumulative.

6. Meituan — China

China's SAMR investigated Meituan's conduct in the online food-delivery platform market.

SAMR concluded that Meituan had abused its dominant position through practices including differential treatment, delayed merchant onboarding, exclusive arrangements, deposits and technological mechanisms involving data and algorithms. SAMR stated that these mechanisms supported the platform's "choose one from two" exclusivity practice and restricted competition.

The authority imposed a fine of RMB 3.442 billion and required extensive rectification.

Principle

Algorithmic and technological mechanisms can be part of an exclusionary strategy even where the underlying conduct also involves contracts and commercial incentives.

Supply-chain relevance

This is particularly important for AI-controlled procurement and logistics ecosystems because:

exclusive contracts + data + algorithms + platform power

can collectively make suppliers dependent on one ecosystem.

7. Alibaba — China

In 2021, China's SAMR imposed a major penalty on Alibaba concerning its "choose one from two" exclusivity practices.

The investigation examined Alibaba's use of platform power to restrict merchants from dealing with competing platforms. The decision was significant because the conduct affected the ability of merchants to participate across competing ecosystems.

Supply-chain relevance

The case demonstrates the importance of multi-homing.

If AI-controlled supply-chain platforms make suppliers choose between competing distribution systems, the platform can potentially:

• prevent suppliers from reaching rival markets;
• increase switching costs;
• deprive competitors of supply;
• increase dependency; and
• reinforce network effects.

8. Leegin Creative Leather Products, Inc. v. PSKS, Inc.

The US Supreme Court's decision in Leegin concerned resale-price maintenance rather than AI.

Nevertheless, it is relevant to AI-controlled supply chains because it illustrates the competition-law importance of vertical restrictions on pricing and distribution.

The Court moved away from treating resale-price maintenance automatically as unlawful per se and subjected it to rule-of-reason analysis.

Relevance to AI

An AI supply-chain system can automatically impose or monitor:

• minimum resale prices;
• recommended prices;
• discount restrictions;
• distribution restrictions.

The fact that an AI system executes the restriction does not determine its legal character.

6. AI Supply-Chain Risks by Category

7. Essential-Facility Dimension

An AI-controlled supply-chain platform can become especially important where competitors cannot realistically replicate the underlying infrastructure.

Examples could include:

• dominant logistics networks;
• critical distribution platforms;
• major warehouse networks;
• national procurement exchanges;
• essential data repositories;
• dominant cloud-based supply-chain infrastructure.

A refusal to provide access is not automatically unlawful merely because a facility is important. Competition law generally requires a demanding analysis of dominance, indispensability, exclusionary effects and legitimate business justifications, depending on the applicable jurisdiction.

8. Data as a Competitive Input

AI supply-chain systems make data strategically important.

A dominant platform may possess:

• real-time inventory data;
• supplier performance histories;
• consumer demand forecasts;
• competitor prices;
• delivery information;
• transaction-level data.

The resulting advantage can create data-driven barriers to entry.

A new competitor may have adequate financial resources but lack sufficient historical data to train an equally effective model.

This produces a possible:

Data → AI → efficiency → market share → more data → stronger AI

feedback loop.

Competition analysis therefore increasingly needs to consider whether an AI advantage is contestable or becomes self-reinforcing.

9. Algorithmic Discrimination

An AI system can discriminate without an explicit human instruction.

For example, a dominant procurement platform might give:

• faster onboarding to affiliated suppliers;
• better search placement to its own products;
• favourable payment terms to selected suppliers;
• higher procurement volumes to preferred partners;
• lower commissions to strategically important firms.

The legal analysis should examine:

1. Who controls the algorithm?
2. What data does it use?
3. What objective is being optimised?
4. Who benefits?
5. Who is disadvantaged?
6. Is the discrimination systematic?
7. Does the undertaking possess substantial market power?
8. Are there objective efficiency justifications?
9. Could equally efficient rivals compete despite the system?

10. Hub-and-Spoke Supply-Chain Risks

A particularly important model is:

Supplier A
↘
Common AI Platform
↗
Supplier B

If the common platform receives commercially sensitive information from both competitors and uses it to coordinate their commercial decisions, it can potentially become the hub facilitating coordination between competing firms.

The European Commission's horizontal-cooperation framework and recent OECD analysis recognise the competition risks of common third-party algorithms where commercially sensitive information is shared.

11. Autonomous AI Coordination

The most difficult future problem is when:

• no competitor communicates with another;
• no explicit cartel exists;
• each firm independently deploys AI;
• each algorithm observes market behaviour; and
• the algorithms gradually converge on similar strategies.

This raises a fundamental legal question:

Can competition law address coordinated market outcomes when there is no traditional human agreement?

There is currently an important distinction between proven algorithm-assisted coordination and merely parallel autonomous behaviour.

The OECD notes that, particularly in the EU framework, traditional Article 101 analysis generally requires some form of communication or concurrence of wills; autonomous parallel conduct therefore raises difficult questions of evidence and legal characterisation rather than automatically constituting a cartel.

12. Efficiency Defence

AI can generate genuine pro-competitive benefits.

For example:

• reduced transportation costs;
• better inventory management;
• fewer stockouts;
• lower wastage;
• improved supplier matching;
• faster delivery;
• reduced procurement costs;
• improved forecasting.

Therefore, competition analysis should distinguish between:

AI optimisation that makes markets more efficient

and

AI optimisation that uses market power to restrict competition.

An efficient algorithm is not inherently anticompetitive.

13. Compliance Framework for Businesses

Companies deploying AI-controlled supply-chain systems should maintain:

1. Algorithm governance

Document:

• objectives;
• variables;
• optimisation criteria;
• decision rules;
• model changes.

2. Competition-risk testing

Test for:

• discriminatory treatment;
• exclusion;
• self-preferencing;
• competitor-data use;
• coordinated pricing;
• exclusive dealing.

3. Data segregation

Competitively sensitive information from independent suppliers should not automatically become available to an affiliated competing business.

4. Human oversight

High-impact decisions involving exclusion or access should have meaningful review mechanisms.

5. Audit trails

Companies should preserve:

• input data;
• model versions;
• decision outputs;
• changes to parameters;
• human interventions.

6. Explainability

A company should be able to explain why a supplier was:

• rejected;
• downgraded;
• charged more;
• allocated less inventory; or
• denied access.

14. Competition-Law Test

A useful analytical framework is:

Step 1 — Identify the market

↓

Step 2 — Identify the AI-controlled decision

↓

Step 3 — Identify the undertaking's market power

↓

Step 4 — Determine whether competitors/suppliers are dependent

↓

Step 5 — Examine algorithmic inputs

↓

Step 6 — Examine discriminatory or exclusionary outputs

↓

Step 7 — Determine whether competitors' commercially sensitive information is used

↓

Step 8 — Test for coordination or information exchange

↓

Step 9 — Assess foreclosure and competitive effects

↓

Step 10 — Examine legitimate efficiencies and objective justifications

↓

Step 11 — Consider appropriate remedies

15. Possible Remedies

Competition authorities may consider:

• prohibition of discriminatory algorithms;
• access obligations;
• data-access remedies;
• interoperability;
• algorithmic auditing;
• separation of datasets;
• restrictions on use of competitor data;
• non-discrimination requirements;
• modification of ranking systems;
• termination of exclusivity;
• behavioural commitments;
• structural separation in exceptional circumstances.

Remedies must be designed carefully because excessive disclosure of algorithms can itself create cybersecurity, trade-secret and manipulation risks.

Conclusion

AI-controlled supply-chain systems are potentially market-shaping infrastructure, not merely business-management software.

The central competition-law concern is the transition from:

AI as an efficiency tool

to

AI as a mechanism for controlling access to suppliers, information, distribution, pricing and customers.

The cases involving online algorithmic pricing, Google Shopping, Amazon, Alibaba and Meituan demonstrate different dimensions of this problem: algorithm-assisted coordination, self-preferencing, exploitation of platform data, exclusionary vertical arrangements and technological enforcement of market restrictions.

For competition law, the critical question is therefore not "Was AI used?" but rather:

"Did control over the AI-enabled supply-chain system materially alter the competitive conditions under which suppliers, distributors and rivals can participate in the market?"