SECTOR INTELLIGENCE BRIEFING, BASED ON DOCUMENTED PUBLIC EVIDENCE, RESEARCH DOCUMENT
◈ Sector Intelligence Retail BreakPoint · AI Control Gap Assessment

AI Control Gap Intelligence:
Retail

What the documented evidence shows, and what BreakPoint would expect to find

Retail AI makes millions of pricing, personalisation, and credit decisions daily. Accountability for those decisions is assigned to functions that did not make them.

Briefing Summary
Document typeSector Intelligence Briefing
SectorRetail · E-commerce · Grocery
Evidence baseDocumented public sources
Regulatory triggerFCA Consumer Duty · ICO · CMA
Core tensionAI pricing + credit at scale, no named human authority
Key AI usesPricing · Personalisation · BNPL · Demand forecasting
BreakPoint instrumentAI Control Gap Assessment
01The Documented Picture
◈  This briefing draws exclusively on documented, publicly available evidence from UK regulators, parliamentary sources, and authoritative industry publications. The predictive findings in Section 04 are inferences from this evidence base, not assertions of findings BreakPoint has made in this sector.
◈ The scale problem

A large UK retailer's AI pricing engine makes millions of individual price decisions every day. A named commercial director is accountable for pricing strategy. No named individual holds authority over any individual pricing decision, because there are too many to hold. The gap between strategic accountability and decision-level authority is not a failure of governance design; it is what AI at retail scale structurally produces. The CMA's investigation into algorithmic pricing, the FCA's oversight of retail credit AI, and the ICO's guidance on automated decision-making all converge on the same point: scale does not dissolve accountability. It relocates it: and retail has not identified where it now sits.

£2.4bn
estimated UK retail AI market value by 2027
ONS / DSIT, 2024
CMA
launched algorithmic pricing investigation 2024, grocery and e-commerce primary focus
CMA, Feb 2024
£30bn
UK BNPL market: FCA regulation incoming; AI credit decisions in scope
FCA, 2024
ICO
published guidance on automated decision-making, meaningful human review required where decisions significantly affect individuals
ICO, 2023

AI is deployed across UK retail in five principal decision domains: dynamic pricing (including personalised pricing by customer segment); personalisation and recommendation engines; Buy Now Pay Later and retail credit decisions; demand forecasting and inventory management; and supplier relationship and contract terms AI. Each operates at a scale that makes individual decision-level human authority practically impossible, which is precisely why the governance question matters. When AI makes ten million decisions and a human reviews none of them, the accountability structure cannot be the same as when a human makes ten thousand decisions and a manager reviews a sample.

The CMA's 2024 investigation into algorithmic pricing and the incoming FCA regulation of BNPL represent the two sharpest regulatory pressure points in the sector. Both are examining the same question from different angles: when AI makes commercially consequential decisions at scale, who is accountable, and what does accountability mean when the AI makes more decisions than any human could review?

02The Governance Gap, What the Evidence Shows
CMA · Algorithmic Pricing Investigation · Feb 2024CMA investigating whether AI pricing algorithms harm consumers, accountability for pricing decisions is the central question
The Competition and Markets Authority launched a formal investigation into algorithmic pricing in UK retail in February 2024, with grocery and e-commerce as primary sectors of interest. The CMA's concern is not that AI pricing is inherently anti-competitive, it is that when multiple retailers use AI pricing tools supplied by the same vendors (or trained on similar data), the resulting pricing patterns can have competition effects that no individual human decision-maker intended or authorised. The accountability gap is structural: the AI produces the pricing outcome; no named human holds authority over the individual price determination; the commercial director holds authority over pricing strategy but not over any specific price. The CMA investigation is testing whether this accountability structure satisfies consumer and competition law obligations.
FCA · BNPL Regulation · 2024 to 25Buy Now Pay Later AI credit decisions coming under FCA Consumer Duty, affordability and creditworthiness AI in scope
The FCA's incoming regulation of Buy Now Pay Later products brings AI-driven retail credit decisions within Consumer Duty scope. BNPL affordability assessments, currently made by AI with minimal human review, will be required to demonstrate that they produce good outcomes for consumers, including consumers in financial difficulty. The AI's credit decision logic, the training data that shaped its outputs, and the named human authority over its parameters will all be in scope for FCA examination. UK BNPL providers have not yet been required to demonstrate decision-level human authority over their credit AI. That requirement is imminent.
ICO · Automated Decision-Making Guidance · 2023Meaningful human review required where AI decisions significantly affect individuals, personalised pricing and credit decisions are in scope
The ICO's 2023 guidance on automated decision-making under UK GDPR requires that where AI makes decisions that significantly affect individuals, those individuals must have access to meaningful human review, not a review that simply ratifies the AI's determination, but one where a human with genuine authority over the outcome applies independent judgement. In retail, personalised pricing (where an individual consumer is offered a different price from another based on AI profiling) and AI credit decisions (including BNPL) fall within this scope. The ICO has noted that "meaningful human review" does not mean a human reads a screen before clicking confirm, it means a human with the authority and the information to reach a different conclusion.
Which? / Citizens Advice · 2023 to 24Documented consumer harm from AI pricing and personalisation, loyalty penalty AI and personalised pricing under scrutiny
Which? and Citizens Advice have documented consumer harm from AI-driven pricing practices in UK retail, including the "loyalty penalty" pattern where AI identifies that customers unlikely to switch are systematically offered higher prices, and personalised pricing where AI segments customers by predicted price sensitivity. In both cases, the harm is AI-generated and systematically applied. The human accountability structure, commercial director responsible for pricing strategy, does not include a named individual responsible for the AI's customer-level pricing differential. Consumer organisations have called for named accountability at the decision-level AI output. No retailer has published such a structure.
DSIT AI Opportunities Action Plan · Jan 2025Government explicitly identifies retail as priority sector for AI adoption, regulatory framework for AI in retail not yet matched to adoption pace
The DSIT AI Opportunities Action Plan (January 2025) identifies retail as a priority sector for AI-driven economic growth, with specific focus on demand forecasting, personalisation, and supply chain optimisation. Government endorsement of AI adoption in retail is running ahead of the regulatory framework needed to govern it. The AI Act equivalent in the UK retail context, a framework specifying human authority structures for AI that significantly affects consumers, does not yet exist. This creates a window in which retailers are being encouraged to adopt AI faster than the governance standards to manage it are being developed.
03The Regulatory Framework
Framework
Core Requirement
Where AI Exposure Sits
FCA Consumer Duty (retail credit)
BNPL and retail credit AI must demonstrably deliver good outcomes. Affordability AI in scope. Named human authority over AI credit parameters required when FCA regulation takes effect.
BNPL credit decisions. Retailer credit card AI. Affordability assessment AI. Buy-in-instalments AI. Any retail AI that determines a consumer's access to credit.
ICO, UK GDPR Art 22
Meaningful human review for automated decisions significantly affecting individuals. "Meaningful" means genuine authority to reach a different outcome, not confirmation of AI output.
Personalised pricing AI. AI credit decisions. Customer profiling AI used in pricing or product availability decisions. Any AI where individual consumer outcomes differ based on AI assessment.
CMA Competition Act
Algorithmic pricing cannot produce competition law outcomes even if no individual human intended them. AI pricing governance must demonstrate that competition effects were foreseeable and human authority was exercised.
Dynamic pricing AI. Competitor price monitoring AI. Personalised pricing by segment. Any AI pricing tool where market-level pricing patterns are an emergent output of AI decisions.
Employment Rights Act / Working Time Regulations
Workforce scheduling AI that affects workers' hours, pay, and conditions is subject to employment law. AI-generated rotas and zero-hours management AI must be governed by named human authority over outputs affecting individual workers.
AI shift scheduling (grocery, logistics, fulfilment). AI demand-driven staffing. Zero-hours hours allocation AI. Any AI that shapes the working pattern or pay of hourly-paid retail workers.
04What BreakPoint Can Predict
Epistemic note: The following are indicative and predictive inferences from the documented evidence base above - not assertions of findings BreakPoint has made in this sector. Each prediction identifies a specific gap type that the evidence strongly suggests is present. BreakPoint's role is to determine, for a specific organisation, whether the gap exists and its severity. Note that BreakPoint is not limited to these factors however, and will often surface unexpected risks or risks that were not fully articulated
✕ Predicted ABSENTNamed authority over individual AI pricing decisions, strategy accountability exists; decision-level authority does not
Every major UK retailer has a named commercial director, pricing director, or equivalent with accountability for pricing strategy. No named individual holds authority over any individual AI pricing determination, because the AI makes millions of them and the governance structure was designed for a world where pricing decisions were made by category managers, not algorithms. The CMA's investigation is testing whether this structure, strategic accountability without decision-level authority, satisfies competition and consumer law. BreakPoint's framework provides the first formal answer: it is ABSENT at the point of individual decision, REAL only at the level of policy approval. The regulator's question and BreakPoint's determination are the same question asked from different directions.
Evidence basis: CMA algorithmic pricing investigation Feb 2024; ICO automated decision-making guidance 2023 (meaningful human review); Which?/Citizens Advice 2023 to 24 (documented consumer harm from AI pricing decisions); DSIT AI Opportunities Action Plan 2025 (adoption ahead of governance).
▲ Predicted ASSUMEDBNPL credit AI: affordability logic reviewed for policy compliance, no named authority over the model's individual credit determination
BNPL providers' credit AI is approved against a credit policy that specifies acceptable parameters. Individual credit decisions, is this customer offered credit, at what limit, on what terms, are then made by AI within those parameters. The named accountability is for the policy, not the decision. The FCA's incoming Consumer Duty obligation will require good outcomes at customer level, not just policy-level governance. When a BNPL AI declines a financially vulnerable customer, or extends credit to one who cannot manage it, the accountability question is: who held authority over that individual determination? The answer, currently, is: no one.
Evidence basis: FCA Consumer Duty (BNPL in scope when regulation takes effect); ICO UK GDPR Art 22 (meaningful human review for credit decisions); FCA BNPL consultation 2024; Citizens Advice evidence on BNPL harm to financially vulnerable consumers.
▲ Predicted ASSUMEDPersonalisation AI: customer segmentation and differential treatment governed at model level, individual treatment decisions are AI-only
Personalisation AI governs which products individual consumers see, which promotions they receive, and, in AI pricing environments, what price they are offered. The governance structure approves the segmentation model and the personalisation logic. It does not assign named human authority over the AI's treatment of any individual customer. Under ICO guidance, where personalisation AI produces decisions that significantly affect individuals, differential access to products, prices, or promotions, meaningful human review is required. No retailer has published an account of how individual personalisation decisions are subject to meaningful human review. That is because the architecture of personalisation AI does not accommodate it.
Evidence basis: ICO automated decision-making guidance 2023; UK GDPR Article 22 (significant effect threshold); Which? loyalty penalty reporting 2023 to 24; CMA digital markets investigation (personalisation and differentiated treatment); DSIT AI governance framework gap (no retail-specific AI decision standard).
◆ Predicted ConditionDemand forecasting AI: scope has extended from stock optimisation into supplier contract terms and de-listing decisions
Demand forecasting AI was approved to optimise inventory levels and reduce waste. The same models now inform, and in some implementations directly drive, decisions about supplier order volumes, contract renegotiations, and de-listing recommendations. These are commercially consequential decisions for suppliers, particularly SME food producers, and they are being shaped by AI that was not governance-approved for that scope. The Groceries Code Adjudicator has documented supplier concerns about retailer algorithmic purchasing decisions. The governance sign-off for demand forecasting AI covers its original scope. The operational scope is broader and the accountability structure has not expanded to match.
Evidence basis: Groceries Code Adjudicator annual reports (algorithmic purchasing decisions, supplier concerns documented); CMA investigation (supply chain AI in scope); DSIT AI Opportunities Action Plan 2025 (supply chain AI as priority investment area); pattern consistent with scope creep Condition across all sectors.
✕ Predicted ABSENTWorkforce scheduling AI: no named authority over individual shift assignments that determine workers' income, AI rotas are not reviewed at worker level
AI shift scheduling tools used in grocery, logistics, and fulfilment determine individual workers' hours, and therefore their income, on a week-by-week basis. The governance structure assigns accountability to the HR or operations function for the scheduling policy. No named individual holds authority over the AI's individual shift assignment for any specific worker. For zero-hours and variable-hours workers, the AI's determination of their working pattern is the determination of their income. Employment law: including the right not to be unfairly disadvantaged, applies to these decisions. The human accountability structure does not extend to individual AI-generated rota decisions.
Evidence basis: Employment Rights Act / Working Time Regulations; TUC reports on AI in workplace scheduling 2023 to 24 (documented worker harm from algorithmic management); ICO employment guidance on automated decisions; Trades Union Congress evidence to BEIS committee on AI and work.
05Source References
CMA 2024
Competition and Markets Authority, Algorithmic Pricing Investigation, launched February 2024. Grocery and e-commerce focus. Accountability for AI pricing decisions under competition and consumer law. gov.uk →
FCA: BNPL
FCA, BNPL regulation consultation and incoming Consumer Duty application, 2024 to 25. AI affordability and credit decisions coming into FCA scope. fca.org.uk →
ICO: ADM
ICO, Guidance on Automated Decision-Making, 2023. Meaningful human review requirement. Applies to personalised pricing and retail credit decisions. ico.org.uk →
DSIT 2025
DSIT, AI Opportunities Action Plan, January 2025. Retail identified as priority AI adoption sector. Governance framework development trailing adoption. gov.uk →
Which? / CA
Which? and Citizens Advice, reports on AI pricing and loyalty penalty, 2023 to 24. Documented consumer harm from algorithmic personalisation and differential pricing. which.co.uk →
GCA
Groceries Code Adjudicator, Annual Reports 2023 to 24. Supplier concerns about algorithmic purchasing decisions and de-listing AI documented. gov.uk →
TUC: AI
Trades Union Congress, AI, Surveillance and Algorithmic Management, 2023 to 24. Workforce scheduling AI and worker income harm. Employment law accountability gap. tuc.org.uk →
Luminary Diagnostics · BreakPoint: AI Control Gap Assessment

Retail AI makes ten million pricing decisions a day.
No human makes any of them, but a human is named as accountable for all of them.

BreakPoint maps the gap between strategic accountability and decision-level authority: the gap that the CMA, FCA, and ICO are all, from different angles, arriving at. The question is who answers it first.

Book a BreakPoint conversation →
Luminary Diagnostics  ·  BreakPoint Sector Intelligence · Retail
luminary-diagnostics.com · © 2026 Luminary Diagnostics Ltd