Built to Let Go: key takeaways from the evening

Retail leaders are ready to hand more of their trading over to AI. What holds them back is rarely ambition. It is the state of their data, the level of trust they can place in the output, and a harder question about what their brand should stand for once machines are running much of the site.

That was the thread running through Built to Let Go, the private dining experience for enterprise leaders hosted by fusefabric and Shopify on 1 October at The Lucky Cat by Gordon Ramsay, 22 Bishopsgate. Senior ecommerce and technology leaders from some of the UK's best-known retail groups gathered to explore a single idea: what happens when day-to-day trading and merchandising is ran by an AI layer, and how far should businesses be prepared to let it go?

The conversation was led by Simon Hamblin, fusefabric Co-Founder and CEO, with guest speakers Ana from Pentland Brands and Raine from Crew Clothing helping to open up each part of the discussion. Both leaders have recently brought multi-brand portfolios onto Shopify, which gave the evening a grounded, practitioner's perspective from the outset. 

The dinner followed over many fantastic courses. The first discussion named the problem, then we explored solutions, and dessert turned to the ultimate benefit. What follows are the themes that resonated most around the table.

Five takeaways from the evening

1. The bottleneck lies in the inputs, not the work 

There was little doubt in the room that AI can already take on a meaningful share of trading and merchandising tasks. The more pressing constraint is everything that feeds those tasks.

When product information has not been enriched, stock has not yet landed in the warehouse or imagery is still sitting with the photography team, even the most capable algorithm stalls. Teams end up chasing colleagues to get the site ready, and the promise of automation never quite arrives.

The consensus was clear. Automate enrichment first, then move on to the next layer.

Some businesses are already seeing the benefit. Product uploads that once took a single person the best part of an hour across several legacy systems are now largely handled by AI, and with greater accuracy than manual entry ever achieved.

2. Shopify is reshaping how teams are built

Replatforming is changing team structures as much as technology stacks. On legacy platforms, content, trading and promotions tend to sit in separate specialist teams, largely because the systems demand it. On Shopify, a single brand ecommerce team can own the entire online journey.

When a task that once took two hours can be completed in minutes, the case for a dedicated specialist weakens and teams naturally become multi-skilled. Several guests expect this shift to accelerate as more of their estates move across.

The next step is AI absorbing the operational load. Keying promotion codes from spreadsheets into the platform, for example, can be a full-time role. The logic already exists, and an agent is well placed to apply it.

3. Autonomy will go far, but brand still needs a human hand

As Simon Hamblin, fusefabric Co-Founder and CEO, put it to the room, retailers are likely to see trading operate close to fully autonomously within the next few years, pushed as far as the technology allows. 

The table broadly shared that view on direction. Search, merchandising and personalisation moving to AI was treated as a given.

The caveat is newness. For seasonal and fashion-led brands, launches, campaigns and sponsorships carry the brand proposition, and there is understandable caution about letting go of them. An A/B test will always find a winner. It cannot tell a business what it wants to be known for.

Guests also reflected on the instinctive merchandisers whose eye for a page could never be fully captured in rules, however hard teams tried. That judgement remains worth protecting.

4. Trust is the real adoption curve

One retailer, turning over in the region of £120 million, shared a candid example. A trial of AI to support buying and size curves was ultimately paused, because without full confidence in the model every output would have needed manual checking and sign-off. At that point, the existing process was just as efficient.

The alternative they pursued is instructive. Natural language processing is now applied to granular returns data, identifying statistically significant patterns in customer feedback and surfacing them to product development before a new style is signed off. Nothing reaches the customer directly, yet returns fall because the product itself improves.

Governance follows a similar curve. One business described opening AI tools to everyone at once, then pausing access, and now reintroducing them step by step as teams build a shared understanding of what is safe.

5. Agentic commerce means designing for two audiences

The dessert course turned to agentic commerce and a tension every brand will need to resolve. Years of A/B testing and conversion optimisation have taught retailers that less is more for human shoppers. AI agents need the opposite: rich, crawlable content that answers every question about materials, fit and construction.

AI visibility tools are beginning to make this measurable, showing where a brand ranks well in AI-led discovery, where competitors are ahead and which content gaps to close. Guests discussed how larger retailers have already used these insights to expand their content footprint and grow their visibility as a result.

Many expect product listing pages to diminish over time, with richer product pages carrying more of the load for humans and agents alike. In the meantime, tools such as Shopify Sidekick are already part of everyday trading for some teams, used constantly for insight and data questions.

On whether agentic commerce will dominate, the room was measured. Most saw it as an additional way to shop rather than a replacement, with certain categories set to benefit far more than others. ChatGPT is already emerging as a source of new customers, even as some retailers continue to block AI agents outright.

What this means for enterprise retailers

The direction of travel is not in doubt. Agentic merchandising is coming, and the leaders in the room are already automating the work that sits beneath it.

The retailers who get there first will be those who strengthen their foundations now: enriched product data, joined-up stock and market logic, teams structured to own the full customer journey, and a clear view of which brand decisions should remain in human hands.

Being built to let go, in other words, means being built well enough to trust what you hand over.

fusefabric partners with enterprise retailers to build exactly those foundations on Shopify, from replatforming through to the data and operating models that allow AI to take on more of the trading. To continue the conversation, get in touch with the team.

Our thanks to Ana and Raine for helping to shape the discussion, to our partners at Shopify, and to every guest who joined us at the table.