Ninety percent of UK retail decision-makers are exploring AI agents, and a third are implementing them across chatbots, forecasting and personalisation, according to research from Eversheds Sutherland and Retail Economics. Yet, despite billions invested, 96% of executives report no ROI. Why?
The issue isn’t a lack of ambition but application. Most AI deployments remain point solutions, optimising single tasks while leaving processes fragmented and manually coordinated. Retail operations are patched rather than redesigned end to end, pressured by high customer expectations and tight margins. Quick fixes often seem safer than systemic changes.
Without embedding AI across process chains, retailers risk missing AI’s full ROI potential.
Beyond Front-End AI: The Rise of End-to-End Process Chains
Discussions about AI in retail often focus on customer-facing applications, like chatbots and personalised recommendations. These innovations only scratch the surface.
A bigger transformation is underway: AI is moving from pilots to integrated process chains, where agents orchestrate decisions across workflows. AI acts as the glue, enabling faster throughput, consistency and resilience amid fluctuating volumes and rising complexity.
This approach helps retailers stay agile during disruption. For example, a beauty company launching a limited-edition SPF set can use AI to oversee its full supply chain. If an ingredient or shipment is delayed, AI can advise how to reroute stock, update promotions, and reschedule staff. By overseeing the chain, AI ensures availability, consistent quality, and resilient performance during supply disruptions and seasonal peaks.
Scaling Logistics AI: Orchestration Across Mixed Automation
“Production AI” refers to systems deployed at scale in real-world operations, enabling dynamic configuration of automation and robotics. It allows mixed environments and complex handovers to function seamlessly.
Operational excellence depends on translating digital decisions like availability, substitutions, returns, into reliable outcomes across warehouses, stores and carrier networks. Process chains aren’t new, but production AI greatly expands their scope, helping retailers manage faster delivery, broader assortments, higher returns and sharper volatility.
A key principle is vendor agnosticism: automation from different manufacturers must collaborate with each other and human workers, avoiding lock-in to proprietary tech stacks.
As retailers prepare for summer launches, production AI can coordinate robots picking items, conveyors routing packages and vision systems performing quality checks. This coordination ensures accurate fulfilment, efficient stock allocation and smooth customer experiences during peak demand.
Arvato is developing an IT platform to orchestrate such technologies allowing fulfilment to adapt with changing volumes and service promises, while maintaining reliability, and cost efficiency. Our recent acquisition of Unchained Robotics strengthens vendor-independent robot integration by making deployment and connectivity easier across manufacturers, reducing complexity and speeding up interoperable automation.
The Analytics Loop: Better Data, Better Execution
Retail data underpins forecasting, fulfilment, and reliable service, but high-quality, labelled operational data often falls short.
Synthetic data, artificially generated to mimic real scenarios, accelerates AI training by simulating rare or hard to capture situations. Vision models and robots thus learn from millions of realistic variations in lighting, packaging, orientation, and damage, boosting reliability during assortment changes or process disruptions. Once deployed, advanced automation generates richer operational data, like exception logs, cycle times and quality signals, feeding continuous improvement. This creates a self-reinforcing analytics supercycle: better models drive stronger execution, and stronger execution produces data that enhances future models.
Augmenting Teams, Not Replacing Them
Retail’s main challenge is coordination, not technology. Point solutions optimise individual tasks, but value is lost at handovers, where exceptions and shifting priorities derail performance. Real ROI relies on end-to-end, AI-enabled networks that preserve context, coordinate decisions, and turn insights into action, improving reliability, speed and cost efficiency.
Human oversight remains essential. AI adoption should include targeted training, enabling employees to collaborate with AI, manage exceptions and ensure technology strengthens daily operations and supports consistent outcomes.
