On 9 September 2026, I attended Visualsoft’s SYNC event: a day devoted to ecommerce, customer experience and the technologies changing how products are discovered, bought and delivered.

The event was hosted by customer-experience expert Kate Hardcastle MBE, and the speakers ranged from Shopify, TikTok Shop, ITV and JD Sports to founders and operators from REFY, MYRQVIST, Universal Works, Paradox London and Calla Shoes.

I expected artificial intelligence to dominate the day. It did, to an extent. There were AI shopping agents, AI-generated adverts, AI-assisted product discovery and questions about whether brands will employ entirely synthetic influencers.

Yet the more interesting theme was the tension running beneath all that technology.

Commerce is becoming more automated. At the same time, the businesses people trust are often the ones protecting their humanity most carefully.

The practical challenge for marketers is therefore twofold: make the business easier for machines to understand, while giving people more reasons to care about it.

That may sound contradictory. By the end of SYNC, it felt more like the job description.

Visualsoft CEO Chris Fletcher opened with one of the day’s most arresting observations: Visualsoft had seen conversion from AI traffic running at five times the level of traffic from Google.

Fletcher connected this with a form of trust transfer. If a consumer has spent time explaining a problem to an AI assistant and trusts the recommendation it receives, the visitor may reach the retailer with far more confidence and intent than somebody clicking an ordinary search result.

Four tests for an AI-ready retailer

Shopify’s Ben Homer gave businesses four questions with which to assess their readiness for AI commerce:

1. Can AI understand what you sell?
2. Can the agent rely on you?
3. Can AI transact on your website?
4. Can AI answer on behalf of the business?

They look straightforward. They are not.

Can AI understand what you sell?

A person can infer a surprising amount from an attractive website. We recognise visual hierarchy, interpret lifestyle photography and fill small gaps using context and common sense.

An AI agent has to reconstruct the proposition from the evidence available to it. It needs structured, consistent information about products, variants, prices, availability, sizing, ingredients, compatibility and intended use.

This is why the humble product page is becoming strategically important again. Long-tail information wins because shoppers do not use AI only to type shorter versions of Google searches. They explain situations.

They might ask for a shirt suitable for a summer wedding in Italy that works with a particular body shape, fits inside hand luggage and can arrive before Thursday. Or a product that meets several dietary, ethical, practical and budgetary requirements at once.

The agent can only recommend confidently if the retailer has supplied enough context.

Can the agent rely on you?

Shopify’s second test concerns trust evidence. Is the product genuinely available? Can the retailer fulfil the order? Is the delivery promise credible? Are the reviews, safety information, policies and returns terms clear?

Humans tolerate a degree of ambiguity and may continue browsing until they feel reassured. An agent may simply exclude an option that cannot be supported by sufficient evidence.

That makes trust signals more than conversion-rate decoration. They are becoming part of the information layer through which a business can be recommended at all.

Can the agent complete the transaction?

Google’s Universal Commerce Protocol was another important part of the discussion. Co-developed with Shopify and other commerce businesses, UCP is an open standard intended to connect AI experiences with retailers throughout discovery, checkout and post-purchase support.

Google describes native checkout within AI Mode and Gemini, while allowing the retailer to remain the merchant of record and retain its customer relationship and business logic.

This matters because a checkout is not merely a payment box. It contains delivery rules, discounts, subscriptions, loyalty schemes, age checks, product restrictions and all the awkward exceptions accumulated through years of operating a real business.

The agentic future cannot work by politely ignoring those complications.

Can AI support the customer afterwards?

The final test takes us beyond acquisition. Can an assistant accurately answer “Where is my order?”, explain the returns process or report a delivery problem?

This is where marketing, ecommerce, fulfilment and customer service stop being separate diagrams and become one customer experience.

The overall message was clear: AI readiness does not begin with launching a chatbot. It begins with sorting out product information, policies, evidence and systems.

The future of commerce has arrived, and it would first like somebody to tidy the product catalogue.

Unified commerce only matters when it solves a real problem

Homer illustrated “unified commerce” with a refreshingly ordinary story.

A customer bought a shirt online. It was the wrong size, so they visited the brand’s physical shop to exchange it. The correct size was in the store, within touching distance, but the exchange could not be completed cleanly because the ecommerce platform and retail system did not share the right information.

From the company’s perspective, there were two systems and a process problem. From the customer’s perspective, the brand had their money, possessed the correct shirt and was somehow incapable of putting the two together.

That is what poor integration feels like in real life.

Shopify’s increasing focus on point of sale is therefore not simply a technology land-grab. The objective is to create one usable view of the customer, inventory and transaction across online and physical retail.

The innovation space also demonstrated how physical shops may become more interactive. A product could be placed on a counter and, through a tag communicating with a nearby sensor, the screen behind it would immediately display the relevant features and information.

It was smooth, slightly futuristic and far more useful than installing a large screen that spends its life playing the same brand film on a loop.

Delivery is part of the product

Douglas Holm of Swedish footwear brand MYRQVIST joined Ingrid to explain how delivery had become a competitive and commercial lever.

The international comparison was particularly useful. Swedish consumers may expect free delivery to a pickup point without necessarily expecting free delivery to their home. They are also more accustomed to being offered a broad selection of delivery methods. British shoppers have been trained differently.

That alone is a warning against copying another market’s ecommerce playbook without examining the local expectations beneath it.

MYRQVIST described express delivery as its tenth-best-selling “product”. According to the figures presented, it had increased the express price from £9 to £15 without a corresponding rise in its own cost, while its wider delivery work produced lower shipping costs and substantially reduced operational effort.

The precise numbers belonged to this case study, but the principle is widely useful: shipping is not merely a necessary expense to hide at checkout. Speed, location, certainty and convenience all carry value—and different customers value them differently.

The footwear category adds another complication: returns.

A customer may order several sizes or styles knowing that most will go back. The first sales report can therefore create a rather flattering picture. Revenue arrives immediately; the returns, processing costs and corrected profitability follow later.

It is a bit like celebrating a goal before VAR has finished drawing its collection of suspiciously precise lines. The ball went into the net, but the number that counts is the one left on the scoreboard.

Brands often focus on scoring more goals—more orders, more revenue, more customers—without checking how many were actually onside. Managing profitability by customer behaviour, fulfilment cost and returns can matter more than the initial order value.

REFY co-founder Jenna Meek provided one of the most compelling stories of the day.

Before REFY, Meek built festival-beauty business Shrine and had worked in product development with fashion brands including Burberry. She later co-founded REFY with model and creator Jess Hunt, who had around 1.2 million followers when the brand launched.

One of Meek’s earlier breakthroughs came through Topshop’s Oxford Circus store. Rather than waiting indefinitely for a conventional listing, her business secured an activation through which customers spending £50 could receive a free makeover. It attracted queues, attention and proof that consumers wanted the proposition—helping create the commercial opportunity that had previously been difficult to obtain.

The clever part was not merely offering something free. The activation allowed the brand to demonstrate demand in the retailer’s own environment.

Meek described REFY’s focus through three areas:

• Brand foundation: simplifying the identity and building a recognisable lifestyle.
• Product: releasing a small number of products based on genuine customer needs.
• Community: remaining close to the people who buy, discuss and represent the brand.

REFY’s lifestyle positioning was built to generate a simple reaction: “I want that life.”

It sounds almost rebellious in a business culture obsessed with speed, scale and the suggestion that every unclaimed market represents a failure of ambition.

REFY resisted opportunities until the product, infrastructure and brand experience were ready. That discipline supported expansion into retailers such as Selfridges and Sephora without turning the brand into a collection of disconnected launches.

Accessibility means involving the person with the problem

Another story concerned a customer with a disability who could not open the packaging. REFY did not merely log the feedback and ask the packaging supplier for a slightly different prototype. The customer was invited in to help redesign it.

It was a small story compared with global expansion and multimillion-pound sales, but arguably more revealing.

Customer-centricity is easy to proclaim when the customer agrees with the meeting already in the diary. It becomes meaningful when the company changes the process and gives the customer influence.

A recall does not have to become a lasting betrayal

Meek also discussed an SPF product recall that cost the business heavily. The lesson was not that recalls are secretly wonderful marketing opportunities. They plainly are not.

It was that trust is often shaped by how a company responds when something goes wrong. Clear action, responsibility and openness can strengthen loyalty, whereas hesitation and defensiveness allow the mistake to become a judgement on the whole brand.

REFY’s wider community approach included field teams focused on building relationships rather than operating purely as field salespeople, and visually distinctive activations such as artists creating watercolour portraits in Paris.

The connecting principle was participation. People were not treated only as an audience waiting for the next advert.

In a breakout session, Amy White of Paradox London and Alex Wright of Yotpo discussed Paradox’s new loyalty programme.

The business wanted to reduce its dependence on discounting and encourage customers to see it as more than an occasional purchase for weddings and formal events. Its answer was a three-tier structure: Pearl, Sapphire and Diamond.

The programme combined familiar mechanics—money off a second order, referrals and double-points events—with early access, birthday benefits and physical experiences. Diamond members could be invited to meet the brand through events such as an afternoon-tea bus tour.

Refer-a-friend was already working particularly well, while the tiers gave Paradox a much clearer view of how its most engaged customers were developing.

The most interesting feature was the ability to upload receipts from third-party retailers, such as John Lewis, to earn points. Paradox could recognise purchases made outside its own website while learning more about customers who would otherwise remain largely invisible.

This is the difference between a loyalty currency and a loyalty strategy.

If every tier simply provides a slightly larger voucher, the programme is mainly discounting dressed in ceremonial clothing. Meaningful loyalty can include recognition, access, community and experiences that would be difficult to place in a percentage-off banner.

Calla Shoes makes footwear for people with bunions and other foot problems. Its session with Attentive (Holly Gaffney and Kat Trenouth) focused on connecting email and SMS around the customer journey.

One promotional sequence used:

• Email on Friday.
• SMS on Saturday.
• Email on Sunday.
• SMS on Monday.

Written like that, it sounds less like sophisticated customer experience and more like a brand that has obtained your number and is thrilled about it.

The important part was the segmentation underneath the sequence.

Recent purchasers were suppressed. Existing customers could receive “treat yourself” messaging, while prospective buyers received a different reason to act. Founder stories and useful information helped build an emotional relationship around the commercial messages.

This is often missing from discussions of personalisation. Marketers concentrate on selecting the correct content, product and send time, but relevance can also mean silence.

Calla also treated promotions realistically. One remark acknowledged that marketers might not like the answer, but promotions remained the largest revenue generator. The mature response is neither to pretend discounts do not work nor to use them indiscriminately. It is to understand where they create incremental value and where they merely subsidise a purchase that was going to happen anyway.

A simple Black Friday recommendation was to change the sign-up message in advance: invite visitors to register specifically for early access to the forthcoming offers. It aligns the value exchange with what the customer actually wants at that moment.

Niels Floors of ChannelEngine, Giulia Yang of TikTok Shop and Jo Hunt of Debenhams framed ecommerce around three successive waves:

1. Marketplaces.
2. Social commerce.
3. Agentic commerce.

Debenhams: an old name with a radically different model

Debenhams provides a remarkable context for that discussion. The name’s retail lineage stretches back to 1778. The historic department-store chain ultimately entered liquidation, and Boohoo acquired the brand and website in 2021 before relaunching Debenhams online. The wider group subsequently adopted the Debenhams name, with the marketplace becoming central to its strategy.

That distinction matters: this was not one unchanged 250-year-old company casually replacing tills with a marketplace plugin. It was the reinvention of one of Britain’s most recognisable retail identities under new ownership and a radically different model.

The marketplace expands the available product range without requiring Debenhams to buy and hold everything itself. According to the panel, the most successful sellers still need the fundamentals: a strong product, appropriate price, effective imagery and story, reliable availability, continuous improvement and a willingness to test the platform’s growth tools.

New route to market; stubbornly familiar need to be good at retail.

TikTok Shop and demand created by discovery

Traditional ecommerce often begins with intent. The customer knows broadly what they want and searches for it.

TikTok can reverse that sequence. Content introduces a problem, behaviour or product that the viewer had not intended to investigate. The familiar reaction is: “I didn’t know I needed that.”

This is demand created by discovery.

The platform’s scale and growth figures presented at SYNC were substantial, but the more useful lesson was qualitative. Raw, relatively unfiltered video can reduce the distance between a customer and a brand. It feels less like a campaign being delivered and more like participation in somebody else’s enthusiasm.

That does not make TikTok right for every organisation. Yang was clear that a brand without stories worth telling may not be a natural fit.

SharkNinja was discussed as a business that has used discovery and product innovation to move customer attention beyond the air fryer and towards categories such as coffee machines and ice-cream makers. The content does not simply capture existing category demand; it helps consumers imagine a use they were not previously considering.

AI is useful when the customer does not know what to search for

Search engines have traditionally rewarded marketers for mapping language to known categories. Generative AI changes the interaction because the customer can explain a complex situation without knowing the commercial vocabulary for its solution.

Yang described AI as particularly useful when people do not yet know what they want.

That is more significant than it first appears. The customer journey can begin with the problem, move through inspiration and product discovery, and increasingly proceed towards the purchase inside the same conversation.

It also reinforces the importance of taxonomy and product data. If a product cannot be understood in relation to needs, attributes and circumstances, it may remain invisible even when it is technically listed.

The old battle was to rank for the words customers typed. The next battle may be to supply the evidence an agent needs when the customer no longer types the product name at all.

Mat Brown of Universal Works and Axel Larsson of Depict introduced a valuable note of caution.

Universal Works had moved from a predominantly wholesale background into ecommerce and acknowledged that parts of the operation had remained manual for a long time. Automation could improve that. But Brown also described the occasions when the brand deliberately overruled automated product ordering.

Fashion is not only a process of measuring what people already want. A retailer may decide to feature a limited edition, introduce something unfamiliar or attempt to set a direction. If historic demand controls every decision, the algorithm can become very good at repeating yesterday.

Universal Works has also made a point of keeping customer interactions human. If customers believe they are speaking to a person, they really are. That is not necessarily an argument against all AI support; it is a reminder that the method of service communicates something about the brand.

The company had experimented with AI in other areas, including helping customers visualise clothing being worn and moving. Static product photography cannot always show how fabric behaves when a person walks.

The balanced lesson was not “AI bad, humans good”. It was to decide where automation improves the experience and where it erases something customers value.

As more businesses use the same models, tools and optimisation systems, distinctiveness risks being averaged away with great efficiency.

Eilish Anderson, Head of Influencer, Social and YouTube at JD Sports, brought the argument back to culture and execution.

JD has an unusually valuable source of customer insight: many of its in-store employees overlap directly with its target market. The company has involved apprentices and shop-floor colleagues in the conversation, giving them a voice even when their answers do not flatter the existing plan.

That last part is essential. Asking for insight while accepting only the answers leadership already likes is not research. It is a meeting with decorative young people.

Anderson described creator selection in terms of values and genuine fit. JD wants to understand what a creator stands for, educate them about the brand and involve them in the business rather than purchase an isolated post.

They notice when someone who has never shown interest in a product suddenly develops a deep attachment to it for approximately one contractual deliverable. They also challenge brands on representation, free products and the experiences routinely given to influencers rather than customers.

The strongest partnerships therefore begin before the campaign. The creator already belongs somewhere near the subject, audience or culture.

A five-creator World Cup campaign bigger than one channel

JD’s work with a five-creator group referred to during the session as the Bov Boys provided the fullest example.

The creators were involved in brainstorming rather than simply handed a completed script. For the World Cup campaign, they stayed in a JD-branded house, meaning the brand remained visible when they streamed through their own channels. Airport advertising and other physical placements extended the idea across the journey.

Although Twitch was part of their world, the campaign’s YouTube performance became especially important. JD reported that the creators’ superfans migrated to its own YouTube channel, sentiment was overwhelmingly positive and average watch time increased by four minutes.

The lesson is not that every retailer now needs five young men in a branded house. It is that the audience relationship belonged to the creators before it benefited JD. The brand earned access by building something the creators and their fans could inhabit naturally.

AI influencers do not remove human questions

Anderson was also asked whether JD would use AI-generated creators.

Her answer was sensibly open rather than theatrical: never say never, but the business must remain answerable to its consumers and preserve a human touch.

The questions raised were more interesting than whether a synthetic influencer can pass for a photograph:

• Why did the brand choose that skin colour?
• Why that hair, face or body shape?
• What does the selection communicate about who the brand considers desirable or representative?
• Why create a fictional person instead of employing a real one?
• Will every AI-assisted post require a disclosure, and how will that affect trust?

AI does not remove creative responsibility. It moves more of it behind the screen.

With a human creator, some characteristics belong to an existing person. With a synthetic creator, every characteristic is—or appears to be—a brand decision. Audiences can reasonably ask what those decisions mean.

This may become one of the defining issues of AI marketing. The ability to generate anything does not free a brand from explaining why it generated that particular thing.

The breakout session on streaming television - with Anthony O'Neil of The Video Ad Agency and Sean Robinson of ITV - showed how the boundary between brand advertising and performance marketing continues to blur.

Connected television was once discussed primarily as a reach channel. Streaming platforms now offer richer targeting, interactive formats and more direct measurement.

Amazon can combine viewing behaviour with shopping data. ITVX can use commerce-focused targeting and formats designed to create a response rather than simply an impression. A “Show me more” prompt can capture interest and trigger a follow-up email. Pause ads occupy the screen when viewing stops, while home-screen skins place the advertiser directly within the browsing experience.

The speakers described budgets ranging from relatively modest tests to major campaigns. The sensible approach was familiar: run an influence test, support it with retargeting, measure the result and scale what works.

There was also discussion of fully AI-generated advertising. High-quality still product images can increasingly be turned into moving creative without the cost structure of a conventional shoot.

That will make television-style advertising accessible to more businesses. It will not make the creative idea optional.

Cheaper production merely allows more people to make a forgettable advert.

So, what did Visualsoft SYNC really say about the future of ecommerce?

I was pulled aside during the day to be interviewed by Visualsoft about the event. My feedback was glowing because the event deserved it: the programme combined large platforms with operators who could speak honestly about implementation, mistakes and the decisions behind growth.

The technology was impressive, but the human stories stayed with me:

• A customer invited in to redesign packaging she could not open.
• Retail employees being asked what their generation actually thinks.
• Creators helping shape a campaign instead of reciting it.
• A fashion brand overriding the algorithm because taste sometimes needs to lead demand.
• A company saying no to opportunities until the timing was right.
• Loyalty members receiving genuine experiences rather than another voucher.

At the same time, the technical foundations are becoming unavoidable. Businesses need product data that agents can interpret, evidence they can trust, checkouts they can navigate and post-purchase systems they can use without inventing the answer.

Perhaps that is the central lesson.

The winners in agentic commerce will not necessarily be the businesses that sound most like AI companies. They will be the ones that make themselves completely legible to machines without becoming interchangeable to people.