ASOS partners with Microsoft on AI Stylist, agents now write 15% of code

TL;DR: ASOS chief executive Jose Antonio Ramos and chief technology officer Przemek Czarnecki have detailed the retailer’s AI rollout in a Microsoft case study. Conversational “AI Stylist” runs on fine-tuned large language models; agents handle around 50% of customer enquiries; AI agents write roughly 15% of ASOS code. The two companies have jointly identified 93 agentic use cases across the value chain. ASOS has consolidated data into Microsoft Azure Datalake with PowerBI dashboards.

The piece, published on Microsoft’s UK Stories platform, is unmistakably a vendor-narrated case study, and the figures it cites are self-reported by ASOS rather than independently audited. Even with that caveat, the percentages are unusually concrete for a UK retailer. Czarnecki frames the agentic deployments as bounded — “you put agents in well constrained contexts, and you always keep humans in the loop” — a stance the case study leans into in light of recent industry incidents involving unguarded coding agents.

What’s actually deployed

AI Stylist is the customer-facing surface: a dialogue-style shopping experience that takes occasion, context and past preferences as inputs. Czarnecki said ASOS partners with Microsoft “on fine-tuning the large language models specifically for fashion”. On the operational side, ASOS is using Dynamics 365 Finance & Operations and back-office agents to automate invoice processing, inventory management and purchasing. Microsoft 365 Copilot is deployed across approximately 3,000 employees. The company says its idea-to-shelf cycle has compressed to roughly three weeks, with 3D design tools reducing physical sampling.

The 93 use cases is the most striking number. Most UK retailers struggle to articulate even ten genuine agentic deployments. ASOS’s framing — agents in customer service, code, finance ops, supply chain, plus trend-scouting from social and internal data — points to an organisation with the data foundation to ask the question seriously rather than experimentally.

Looking forward

For UK retail competitors, the ASOS framing — that AI is mostly a velocity and personalisation play, not a cost-cutting one — is the more honest read. Ramos’s claim that AI Stylist could become “more inspirational than shopping offline” is marketing, but the underlying integration of personalisation, agents and centralised data is the real workstream. UK SMEs in retail considering similar deployments should treat the 50%-of-enquiries-handled-by-agents figure cautiously: that ratio relies on a heavy investment in the human escalation path and on retained capacity for the complex 50%, not on full agent replacement.