Burberry
genAI copilot for consultants (search, image recognition, complete-the-look)
Penguin, the genAI clienteling copilot Burberry built in-house, delivered a 24% lift in average transaction value across customer service channels and equipped more than 100 consultants across every market.
Key points
- Penguin GenAI copilot for consultants: product search, street-to-shop image recognition, complete-the-look.
- Built in-house (transformers and LLMs fine-tuned on Burberry looks) by a three-person team.
- +24% average transaction value across customer service channels, 100+ consultants equipped.
- Evidence level B, live status confirmed (2025 DataIQ Awards).
Objective
Give Burberry customer service consultants a genAI copilot that speeds up product search, item recognition, and outfit suggestions, so they serve customers faster and raise transaction value.
The deployment
Penguin is a generative platform built in-house by a three-person data science team together with customer service. It combines three functions: multilingual natural language product search, street-to-shop image recognition that identifies a Burberry product in a photo and matches it to the catalog, and a complete-the-look feature that generates coherent outfit suggestions in line with the brand aesthetic from a single item. Penguin relies on transformers and LLMs fine-tuned on Burberry looks. In parallel, Burberry runs a suite of data models (intent, recommendation, product affinity) that powers store-led campaigns. The effort was recognized at the 2025 DataIQ Awards.
Results Proof B
Two award-winning 2025 DataIQ Awards submissions quantifying separately the clienteling copilot (Penguin, +24% ATV) and the data transformation (7M GBP incremental). Quantified case study, judged by an industry third party, not a figure from the brand alone.
How it works
Documented architectureThe stack in detail
- llm LLM fine-tunes sur les looks Burberry Language models adapted to the brand's curated corpus of looks for complete-the-look; the base model is not named in the sources.
- outil Reconnaissance d'image street-to-shop (transformers maison) Detection of a Burberry product in a photo and matching to the catalog, integrated into Penguin.
- outil Recherche produit en langage naturel multilingue Penguin's search function, built in-house on the product catalog.
- outil Suite de modeles data (intention, recommandation, affinite produit) In-house models that power store-led campaigns, with more than 90% of store-led campaigns using this framework.
How it runs, concretely
For ops teams-
1Consultant query customer service consultant
The consultant asks a product question in natural language or uploads a photo of an item to identify.
-
2Search and recognition AI
Penguin interprets the query or detects the product in the image (street-to-shop) and links it to the catalog.
-
3Outfit suggestion AI
From a single item, complete-the-look generates suggestions consistent with Burberry looks.
-
4Response to the customer customer service consultant
The consultant validates, adjusts, and delivers the response to the customer, staying in control of the relationship.
The up-to-date product catalog and the curated Burberry looks used for fine-tuning. Without this brand corpus, complete-the-look drifts away from the aesthetic and loses credibility.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- structured product catalog with images
- brand look corpus for fine-tuning
- opt-in customer data for personalization
Org prerequisites
- in-house data science team able to fine-tune LLMs
- customer service involved in the design
- change management on the consultant side
Possible stack
- open or API LLM with fine-tuning
- vision model (product detection)
- recommendation engine
- mobile/tablet consultant interface
The plan, step by step
- Step 1Structure the catalog (attributes, images) and curate a brand look corpus with the style teams.Deliverable: Training corpus validated by style leadership.
- Step 2Prototype the natural language product search on the catalog and have it tested by a handful of consultants.Deliverable: Prototype tested with documented consultant feedback.
- Step 3Add image recognition and the complete-the-look fine-tuned on the look corpus.Deliverable: Three-function copilot piloted on a limited scope.
- Step 4Roll out to consultants in one market with training; the consultant keeps control of the response to the customer.Deliverable: Adoption measured in a first market.
- Step 5Compare average transaction value across equipped and non-equipped channels, then decide on expansion to other markets.Deliverable: Average transaction value review and rollout plan.
First step: Prototype a natural language product search plus a complete-the-look on a catalog, tested with a handful of consultants.
The same pattern, proven in other industries
Amazon
genAI shopping assistant
Booking.com
genAI trip planning assistant
Flipkart
conversational genAI shopping assistant
Sources
- S1 Most Innovative Use of AI (Global) - Burberry (DataIQ Awards 2025) Secondary archive pending
- S2 Transformation with Data (Global) - Burberry (DataIQ Awards 2025) Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.