How Pentland Brands Is Using AI to Replace Product Photo Shoots and Transform Operations

Pentland Brands Stopped Photo Shoots. AI Made It Possible.

There is a gap between what brands know AI can do and what they have actually committed to. Most have run a pilot, published a case study, and moved on. A smaller number have done something harder: restructured how the business creates, negotiates, and makes decisions, then lived with the consequences.

Pentland Brands is in that second group. At Shoptalk Europe 2026 in Barcelona, a senior leader from the company, known in the session as Anna, gave a frank account of what it looks like to push AI adoption past the proof-of-concept stage into operations. The session was titled “Preparing for a Future of AI-first Discovery.” What emerged was something more specific and more useful: a field report from an organisation that has made commitments it cannot easily reverse.

The most striking of those commitments concerns product imagery. Pentland, which operates as both a wholesale and DTC business, regularly has to sell products to partners months before those products exist in physical form. For Decathlon, JD Sports, Amazon, and others placing orders six months out, the selling packs, including the model imagery and product visuals, are now entirely AI generated. “We’ve stopped all shoots,” Anna said.

The brief was bottom-up, not top-down

The decision to go this far did not come from a centralised mandate. It came from a structured programme of distributed experimentation. When Anna joined Pentland, the business had launched an AI entrepreneurs programme: an open invitation to employees at every level, from SVPs to operational staff, to train intensively on multiple AI tools and develop business use cases. The CEO framed the goal directly: “We want AI to be a virus that infects the business.”

The programme produced many use cases, most of which fell into two categories. The first was operational efficiency, compressing timelines that previously ran to weeks into hours or days. The second was structural complexity: problems that had previously required coordination across multiple global teams and extended time horizons.

For the creative team, the product imagery use case sat firmly in the second category. The challenge was not just speed. It was coverage. In fashion wholesale, samples are often unavailable at the moment you need to sell. Supply chains slip, factories run late, and photo shoots cannot happen without product. AI-generated imagery solved a logistical problem that existed well before anyone started discussing the technology.

Getting to quality took longer than expected

The honest version of this story involves a prolonged period of failure. Reaching a standard good enough to present to wholesale partners took significant time and multiple partner changes. “It looked fake for a long time,” Anna said. “It wasn’t realistic enough to evoke genuine inspiration.” The specific challenge for a brand like Speedo, where selling imagery might require models in swimwear at a pool location, illustrates how demanding the brief was. Generic AI output was not sufficient. A specialised partner, Grasswold AI, was eventually selected to handle selling pack production.

The next phase moves from selling packs to consumer-facing imagery: generating product images directly from CAD files for use on DTC portfolio pages, bypassing the sample-and-shoot process entirely.

Negotiation tools changed how deals happen in the room

The second use case Anna described was less visible but arguably more revealing about where AI can shift operating behaviour rather than just speed it up. The business built a negotiation tool, described as a toggle, that allowed teams to model pricing, volume, and SKU combinations in real time. The change was not just the tool itself. It was that teams began using it in the room with suppliers, transparently, rather than retreating to run spreadsheet scenarios between meetings. The back-and-forth reduced. Decisions moved faster.

For senior marketers and commercial leaders, this is the use case worth examining. The value was not that the AI did the analysis. The value was that it changed the social dynamic of the negotiation.

Where the business acknowledges it is behind

Anna was direct about the areas where Pentland has not made equivalent progress. Data access is the most significant. Despite running both wholesale and DTC operations, the business has not yet built the semantic data layer that would allow anyone across the organisation to query data and make decisions from it. Definitions of revenue and margin differ across business lines. The infrastructure required to use AI for decision support at scale has not been put in place.

The second acknowledged gap is ROI quantification. The programme has generated anecdotal evidence of time and cost savings, and the exec team has seen the use cases. But translating that into documented return on investment has proved difficult. The business is still working on it.

Both gaps are worth naming because they are common. Most organisations running AI programmes at scale face the same two blockers: data that is not clean or accessible enough to support AI-driven decisions, and an ROI case that is hard to close when experimentation is distributed and outcomes are qualitative.

What this means for senior marketers

The Pentland case makes a specific argument about how AI adoption actually happens in large organisations. It does not start with a centralised technology strategy. It starts with a permission structure: an invitation to experiment broadly, with executive backing that treats the cost of experimentation as acceptable even before the ROI is proven. What follows is consolidation, a review of the use cases that emerged, an identification of the overlapping problems, and a decision about which two or three investments can satisfy the ten requirements that distributed experimentation surfaced.

The lesson is structural as much as technical. What Pentland has built is not primarily a set of AI tools. It is a process for turning distributed experimentation into consolidated capability. The organisations that will make the most of AI are the ones that figure out how to run that cycle faster.