How Burberry uses data and AI across clienteling, design and the supply chain
Burberry has built customer models for its client advisors and talked publicly about predictive inventory and traceability pilots. This case study separates what has been published from what has not.
KEY TAKEAWAYS Summary by the editors
- Burberry won the Transformation with Data (Global) award at the DataIQ Awards 2025 for a Customer Science capability built on a single customer view and a composable Customer Decisioning Platform.
- According to the DataIQ award page, Burberry deploys six AI models across the customer journey, and client advisors use their outputs to choose clients, tailor local campaigns and prepare for interactions.
- The same page reports that AI-powered store-led campaigns generated £7 million of incremental revenue in 2024, and that over 90 percent of store-led campaigns now use the AI-driven framework; these are figures reported on an awards page, not audited results.
- Burberry's Director of Digital Delivery said in September 2024 that predictive AI is used for inventory, which he called critical because of variables such as weather, social media and celebrity product placement.
- Burberry has run a proof of concept with IBM for blockchain-based traceability, but published material describes it as a proof of concept, not a full roll-out, and describes no AI in the design process.
Burberry uses AI mainly to help store teams decide which clients to contact, with which products and when, and publicly describes predictive models for inventory as well. The strongest published evidence concerns clienteling: an award submission reports six customer-journey models and measurable campaign gains. Evidence for AI in design and in the supply chain is thin, and this article says so where it applies.
What has Burberry published about its use of data and AI?
Three types of source exist. The first is a 2025 industry award page from DataIQ that summarises the work of Burberry's data science team. The second is a conference talk in September 2024, reported by TechInformed, in which Jon-Paul Brett, Burberry's Director of Digital Delivery, described several digital initiatives. The third is older: a Harvard student blog post from the Digital Initiative platform that collects earlier public information, including a 2018 interview with Burberry's then SVP of IT.
None of these is an audited disclosure. The award page repeats outcome figures supplied for the award, and the student blog explicitly relays third-party claims, one of which (a 50 percent increase in repeat purchases by 2015) it labels as alleged. Readers should treat the numbers as indicative.
How does Burberry use AI for clienteling?
According to the DataIQ page, Burberry's data science team built what it calls a Customer Science capability. It rests on a single customer view and a composable Customer Decisioning Platform that unifies data from all channels into real-time customer profiles. On top of that sit six AI models placed at different stages of the customer journey, including intent models that estimate which customers are likely to engage and recommendation engines that suggest products.
The page adds that a newer model uses large language models to predict product affinities from sequences of customer behaviour, and that it outperforms traditional methods by a factor of two on retrieval metrics. The practical use is human: client advisors use the insights to select clients, adapt campaigns to local markets and prepare for interactions. The model proposes, the advisor decides and speaks to the client.
| Metric | Reported result | How to read it |
|---|---|---|
| Conversion rate of store-led campaigns | Doubled | Compared with a baseline the page does not define |
| Engagement | Tripled | Engagement is not defined on the page |
| Incremental revenue in 2024 | £7 million, an 11% year-on-year uplift | Reported as incremental, method not published |
| In-store appointment bookings | Tripled | Shows advisor activity, not necessarily sales |
| Manual effort in workflows | Reduced fivefold | Automation of campaign preparation |
| Share of store-led campaigns using the framework | Over 90% | Indicates broad adoption inside the business |
How does Burberry use predictive AI for inventory?
At Connected Britain in September 2024, Brett said predictive AI is used for inventory and called it critical, given variables such as weather, social media and celebrity product placement. No model type, accuracy figure or scope (which categories, which markets) was reported. The honest reading is that Burberry treats demand sensing as a use case, without evidence about how far it has been taken.
The same talk described machine learning and recommendation engines that predict customer preferences from available data, a virtual clothes rail that an associate can send to a customer so that selected items wait in the fitting room, and RFID-tagged products that send styling suggestions to a customer's phone. These are examples of connecting online data to store service, not of automated decision-making.
Does Burberry use AI in the supply chain and in design?
On the supply chain, the published item is a proof of concept with IBM for a blockchain solution covering sourcing of raw materials through to distribution. A QR code on a trench coat was shown to reveal its journey from cotton farm to spinning to manufacturing site. This is traceability technology, not AI, and it was described as a proof of concept. No source reviewed here reports AI in Burberry's design process or planning of production.
The 2018 interview summarised by the Harvard blog states plans to use machine learning to automate development, operations and testing, to improve scenario modelling for planning and logistics, and to strengthen security and fraud prevention. No timelines or results were attached, so these should be read as intentions.
What can other fashion companies learn from Burberry?
- Start from a unified customer view. The reported models depend on a single customer view and a decisioning platform, not on any one algorithm.
- Put the output in front of people who talk to clients. The value described is in better advisor decisions and preparation, with appointment bookings as a leading indicator.
- Measure against a baseline. Conversion and incremental revenue only mean something relative to a control group; ask what the comparison was.
- Separate pilots from deployments. Blockchain traceability and the in-store tech showcases are described as pilots or showcases, and should not be mistaken for scaled programmes.
- Expect short technology cycles. Brett said a connected magic mirror installed about a decade earlier stayed in stores for several years, whereas he now expects technology to need refreshing every year or two.
Most of the quantitative claims come from one awards page. It does not state the size of the test and control groups, the period over which conversion doubled, or the costs of the programme. Burberry's own annual report and investor material were not used for these figures, so no statement is made here about the contribution of AI to group revenue.
What data does a clienteling model like Burberry's need?
The published description points to several prerequisites. A single customer view means that purchases, appointments, online behaviour and service contacts are linked to one identity across channels. A decisioning layer then turns scores into actions, such as which clients an advisor should contact this week. Without the first step the models have little to learn from, and without the second the scores remain a report that nobody uses.
There are also governance questions that the sources do not address. The 2024 conference report mentions that recommendation engines can draw on customers' social media accounts. Using such data in Europe raises consent and purpose limitation questions under data protection law, and any company copying the approach should confirm what it may lawfully use. Burberry's own policies on this are not described in the sources reviewed.
How should executives judge a luxury AI case study?
Three checks help. First, identify the source: an awards page, a conference talk and a student blog carry different weight from an annual report. Second, look for a baseline and a time period, since a doubling of conversion from a small starting point is a different result from a doubling of a mature programme. Third, ask what is still human: in Burberry's case the described models inform advisors, who remain responsible for the client relationship.
Frequently asked questions
How does Burberry use AI?
The best documented use is in clienteling. Burberry's data science team built a Customer Science capability with six AI models across the customer journey, and client advisors use the outputs to select clients and prepare campaigns. Burberry has also said predictive AI is used for inventory, without publishing details.
Did AI really double Burberry's conversion rates?
The DataIQ Awards 2025 page reports that AI-powered store-led campaigns doubled conversion rates and tripled engagement. It does not publish the baseline or test design, so the figure is an unaudited claim rather than an independently verified result.
Does Burberry use blockchain in its supply chain?
Burberry has run a proof of concept with IBM for a blockchain solution that traces products from raw material to finished goods, including a QR code that shows a trench coat's journey. Published sources describe it as a proof of concept, not a company-wide system.
Does Burberry use generative AI for design?
None of the sources reviewed for this article report generative AI in Burberry's design process. The only generative or large language model use reported is in a customer affinity model that predicts product interest from sequences of behaviour.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
