Tapestry Patents Mira, Its In-House AI Decision Engine for Coach and Kate Spade
Tapestry has secured a US patent for Mira, the proprietary AI platform it has built to unify data across Coach and Kate Spade and accelerate decision making for merchandising, inventory and consumer teams. The patent covers Mira's core system architecture and is the group's first in AI, following an earlier patent on its Global Data Fabric - the data layer Mira sits on top of.
Built in-house by Tapestry's Data and Analytics team under CDAO Fabio Luzzi, the platform connects merchandising, inventory, product, consumer and financial data into a single queryable layer, with role-based access controls and governance built in from the start. Tapestry says Mira is already in use for assortment planning, inventory optimisation and reacting to emerging consumer trends, compressing work that previously took days into seconds or minutes. CEO Joanne Crevoiserat framed the platform as evidence that an 85-year-old fashion company can operate at the speed of a tech-native one.
Why it matters: The substantive question isn't whether Tapestry is becoming a tech company - it isn't - but whether building proprietary infrastructure beats licensing from the established planning and analytics vendors most retailers use. The bet is that a system tuned to Tapestry's specific merchandising, inventory and consumer model, with governance baked in, will outperform what software vendors can offer off the shelf. That's a real trade-off: in-house builds carry maintenance cost and key-person risk, while packaged software brings vendor roadmaps, integrations and benchmarking across peers.
The patent itself is unlikely to have much defensive value, what matters is whether Mira delivers measurable margin or speed advantages over what Tapestry could have bought. If it does, expect more brands to revisit the buy-vs-build question rather than ceding the decision layer to software vendors.
Amazon Licenses Alexa for Shopping to Outside Retailers, With Kate Spade First in Production
AWS has launched the Agentic Shopping Assistant, a packaged version of the technology behind Alexa for Shopping - the AI assistant formerly known as Rufus, which Amazon says drove nearly $12 billion in incremental sales on its own platform last year. The product gives retailers the architecture, starter code and implementation support from the AWS Generative AI Innovation Center to launch a branded conversational shopping experience in around 60 days rather than building from scratch. Each deployment is tuned to the retailer's catalogue, customer data and brand voice, generating personalised product recommendations as visual cards with images, pricing and quick replies.
Kate Spade is the first retailer in production, having launched its AI Gift Concierge in April - built on Anthropic's Claude Haiku 4.5 via Amazon Bedrock - which questions shoppers about occasion, recipient and style to identify gift suggestions. Additional retailers are in testing, and the product sits within an intensifying field that includes Google's open commerce standard with Shopify, Microsoft Copilot's checkout integrations with brands like Ralph Lauren, and OpenAI's work with Walmart and Shopify inside ChatGPT.
Why it matters: Amazon is doing what Amazon does: turning internal infrastructure into a paid platform for everyone else, the same playbook that built AWS twenty years ago. The strategic logic makes sense - if conversational commerce becomes the default discovery layer, Amazon would rather own the underlying stack than watch retailers build it on Shopify, Google or Microsoft. The trade-off for brands are uncomfortable though: a 60-day path to a working assistant comes with deep dependence on a competitor's infrastructure and learnings drawn from Amazon's own shoppers.
The Kate Spade detail is the most revealing part - Tapestry has spent the week positioning Mira (see above) as proof of its in-house AI capability, yet its customer-facing conversational layer runs on Amazon's rails through Bedrock and Claude. That split is likely to become the norm rather than the exception: proprietary infrastructure for internal decisioning, licensed platforms for customer-facing AI, because the economics of training and operating a frontier-grade shopping agent are overkill for almost any single retailer.
Gémo Reports 2.3x Conversion Lift From Algolia's AI Search
French omnichannel retailer Gémo has credited Algolia's AI search platform with a 2.3x increase in conversions and a claimed 160x return on investment, with the technology now driving a third of the retailer's digital revenue. The deployment centres on Algolia's AI Ranking, which reorders results based on real-time shopper behaviour, alongside Merchandising Studio and rules-based controls that let category managers boost products and tune results around seasonality and campaigns without engineering support. AI Synonyms reduces zero-result queries by interpreting varied phrasing of the same intent.
Gémo, which positions itself around affordable apparel for families, operates a large catalogue that the retailer says had previously made consistent search relevance difficult to achieve. Category manager Elsa Souillart framed the work as groundwork for conversational discovery experiences to come.
Why it matters: Strip away the vendor's ROI figures - 160x is a marketing number, not an audited one - and there's a genuinely useful example here about where AI is actually earning its keep in fashion retail. While the headlines go to conversational agents and agentic checkout, the unglamorous work of making on-site search return the right product remains one of the highest-leverage applications available, and one of the few with a measurable conversion line attached.
Gémo's own framing is interesting: it treats improved search as the foundation for conversational discovery rather than the endpoint, which is the correct sequencing. A shopping agent built on top of a catalogue that can't reliably match a query to a product will simply fail more fluently.