Introduction

Coresight Research is a research partner of Shoptalk Fall 2026, taking place September 29–October 1 at the Music City Center in Nashville, Tennessee.

The sessions at Shoptalk Fall 2026 are categorized into three major themes, which we presented in our guide to the event:

• Retail excellence
• Brand storytelling
• Understanding the customer

In this report, we present highlights from day two, which mainly covered customer loyalty, marketplaces and social commerce, AI and agentic commerce, and customer engagement.

Shoptalk Fall 2026, Day 2: Coresight Research Analysis

1. Brands Build on Customer Data and Communities They Own

Loyalty as a relationship, not a points ledger

Three vendors that pitched loyalty programs for womenswear brand DÔEN agreed that a program should be built around belonging and reciprocal value, to engage customers and deepen relationships with them, not to keep a points ledger or hand out discounts. Their common advice was to tie the program so closely to the brand that no other company could run it, and to give customers reasons to take part beyond buying. Regan Jayne, Co-Founder and CEO at loyalty platform Rediem, framed it as a “logo swap” test: if a loyalty page would suit a dog-food maker or a cosmetics brand once the logo changes, it is not tied closely enough to the brand.

Premium brands can reward customers without money, according to the panelists. Points are optional, Jayne said, because access, co-creation or recognition can carry high perceived value with little reliance on promotions. Christine Lee of loyalty platform Talon.One warned of a “discount trap,” in which promotions erode margin and train customers to expect lower prices, and named access, aspiration and advocacy as the alternatives.

The vendors favored starting small. Angela Clark, Founder at Eclipse Advisory Group, recommended an “invisible launch”: build the platform, quietly invite a selected group of the most engaged customers, watch what they use and value, and refine before spending on public marketing. Lee advised beginning with a smaller use case: “don’t let perfect get in the way of good.” Results should be measured on lifetime value, repeat purchase, retention, engagement and margin protection rather than enrollment or points redeemed, the panelists added. Loyalty also needs owners across several teams instead of a lone page in the site footer.

Collecting customer data directly

• The Vitamin Shoppe added DoorDash and Uber Eats for convenience and immediacy. It then began delivering orders from its own website, aiming to beat Amazon on speed, because the marketplaces keep the customer relationship and the data, said Sharon Leite, CEO at The Vitamin Shoppe.
• BRUNT Workwear asks customers about their trade, whether they work indoors or outdoors and the weather on the job. It combines the answers with session and past purchase data in machine-learning recommendations that feed cart upsells, product pages, reordering and lifecycle content. The aim, said Christopher Pond, Chief Growth Officer at BRUNT Workwear, is to show a boot buyer the rest of the collection.

Customer advocates and communities

Real customers and communities are the most durable discovery asset, according to speakers from Reddit and BRUNT Workwear on the digital discovery panel. Allison Tepley, Head of Global Audience Marketing at Reddit, cited the company’s work with Walmart on amplifying individual user-generated content, which brands have traditionally not amplified. “People love brands and they love products and they love sharing them,” Tepley added. Platforms and tools will keep changing, but the human perspective is becoming more valuable, and brands can seed and amplify conversation across many communities even though they do not control it, Tepley said.

On TikTok and Meta Reels, organic content now reaches viewers by niche interest, so a customer’s post about a product in their field can land in a prospective customer’s feed, and brands should encourage sharing, Pond said. Community, brand affinity and real-world experiences will matter more as commerce becomes more agentic, according to the loyalty panelists. In Jayne’s view, community becomes a stronger moat as AI-driven commerce makes product, price and distribution easier to copy.

We believe the speakers point to a discovery model built on trust, data and deliberate design. We believe retailers and brands should invest in programs that surface and amplify real customer content and advocacy across platforms, instead of depending on any single algorithm. They should also collect zero-party data and combine it with behavioral data to personalize across categories and the lifecycle, and use AI for customer research and listening, not only for promotion, without optimizing every journey for speed at the expense of inspiration.

2. Marketplaces Add Customers, and QVC Reports 1 Million From TikTok

Marketplace sales

Marketplaces bring retailers customers they do not reach on their own, provided assortment, content and delivery fit each platform, panelists at the marketplace session found. Sally Beauty Holdings reported that 75% of its delivery-marketplace business is incremental, said Natalie Lockhart, Chief Strategy and Digital Officer at Sally Beauty Holdings. It treats DoorDash, Uber and Instacart as a third channel next to stores and owned e-commerce. Delivery demand went beyond the low-stakes items Sally expected, such as lashes: it sells a lot of hair color, and DoorDash orders peak at about 5:00 p.m. on Fridays as customers get ready to go out.

RG Barry Brands found little overlap between marketplace shoppers and its own direct-to-consumer channel, which brought net new consumers, said Corinne Baker, SVP, Ecommerce DTC and Marketplaces at RG Barry Brands.

Platform-specific tactics

Each marketplace rewards different behavior, so strategy has to differ by platform. Some favor frequent product introductions, which calls for staggering launches to keep exposure continuous, while others raise a product’s ranking as sales accelerate, which puts pricing and media support at the center, Baker said. Content has to differ as well, because what works on TikTok differs from what works on DoorDash. On any marketplace, The Vitamin Shoppe holds a firm line on accurate ingredient and manufacturing information, said Sharon Leite, CEO at The Vitamin Shoppe. Panelists found that lasting growth also takes reliable data and investment in owned channels.

Live and social commerce

Live shopping works when content and selling style fit each platform and market, and when merchandising, fulfillment and customer service can support the sales, according to speakers on the video and social shopping panel. Kind Patches found that aggressive selling and urgency raised engagement and sales in the US, whereas UK audiences responded to a calmer style. Conversion is its biggest measure. “Just try it and push all your energy into trying something and pivot, pivot and pivot,” said Magnus Hjörne, Co-Founder and CEO at Kind Patches.

TikTok brought QVC Group 1 million new customers in 2025, a mix of existing fans and people who had not known the company, said Krystyna Taheri, SVP, Social Commerce at QVC Group. QVC treats TikTok Shop as a full retail business rather than a marketing channel, with studios built for TikTok, omnichannel merchants, private-label capabilities, five warehouses and a customer service team staffed by people. It tests hosts, categories, prices and timing, and gives each new product a 28-day cycle to reach mass velocity.

Live shopping takes far more operating investment than switching on a platform AI feature, so retailers interested in it should start before demand matures, said Joe Yakuel, Founder and CEO at WITHIN.

• Look out for our forthcoming report on video commerce, including livestreaming e-commerce.

3. Panelists Call Agentic Storefronts Overhyped

Agentic storefronts and product data

Consumer behavior has not caught up with agentic storefronts, panelists agreed, but retailers should prepare product data, feeds and infrastructure now so they can take part as the channel matures. Joe Yakuel from WITHIN called the storefronts overhyped because autonomous shopping agents account for little conversion compared with the attention they get. Rebecca Maffei, Chief Information Officer at Fashionphile, agreed adoption is early but said the data and infrastructure work supports channels beyond agentic commerce, even if some experiments are discarded.

The clearest near-term priority, panelists said, is what retailers already control: product data, feeds, search, customer data, apps and websites. The website remains a brand’s main owned entry point for loyal and prospective customers and the place where large language models and shopping agents find structured information, in Yakuel’s view. Panelists converged on product feeds as a practical investment, because feeds already supply Google Shopping, performance media, social platforms and retargeting, and they improve readiness for AI discovery. Low-cost pilots are worthwhile because the questions shoppers ask reveal unmet needs, according to Laura Cantor, VP, Marketing and Ecommerce at New York & Company.

Panelists treated GEO and AEO as an extension of good product information and third-party citations, not a standalone fix. Third-party editorial and “best of” citations can decide whether a retailer appears when shoppers search a category without naming a brand, and sentiment in those sources can shape how the brand comes across.

Data quality, governance and human review

Illustrating the importance of data quality, an automated pricing run across roughly one million SKUs at RG Barry Brands missed empty margin cells in the data. Items priced at $400 sold for $50, and thousands of orders went unfilled, said Corinne Baker, SVP, Ecommerce DTC and Marketplaces at RG Barry Brands. The company now cleans its data before automating further, Baker added, because automation does not make data safer, and it wants clean data and the right controls as it introduces AI.

AI speeds the work but does not replace engineering, governance or human review, panelists said. Vibe coding shortens the path from idea to prototype, but a prototype that must serve employees or customers at scale still needs security, privacy and maintenance work, according to Maffei. AI-generated video is harder to deploy than static images because of authenticity and legal concerns, and human review remains standard before AI-created work goes live, Yakuel added.

Panelists split on whether a company needs a Chief AI Officer. An AI-specific role can leave other executives thinking AI belongs to someone else, Yakuel said, while a dedicated leader gives governance, accountability and investment a clear lane, according to Maffei.

Loyalty panelists also cautioned against adopting AI for its own sake, and Christine Lee of Talon.One said the company uses it mainly to explain reporting and campaign performance.

4. Speakers Say a Funding Pitch Starts from the Decision Sought

The pitch as a decision tool

A pitch works as a decision-making tool for founder and investor, not as a sales tool, said Joy Allen-Altimare, President at Vella Bioscience. Aron Bohlig, CEO and Managing Partner at ComCap Holdings LLC, credited her with “the frame.” The frame is to start from the decision sought, then work backward to the proof that supports the thesis, the earned insight that positions the team and the investor’s own business model and process.

Investors work from a checklist. Sequoia Capital’s venture pitch deck guide supplies seven of its nine items, and Bohlig adds “proof” and “edge.” At seed stage, anecdotes can serve as proof, but by Series A investors expect operating figures and customer data points. Early investors buy “possibility,” while larger venture funds buy “probability.”

Finding investors who fit

Laurel Orley of Daily Crunch said the company has raised about $8 million, starting with about $1 million from friends and family. Food and beverage venture funds set minimums of about $20 million, so she turned to Launch Tennessee, a state-backed fund. About one in 10 people responded to her cold messages on LinkedIn and Instagram, and the replies brought a six-figure investment from Bethenny Frankel and another from a former Poppi executive who now sits on the board.

Vella raised a little under $10 million, mostly from consumer packaged goods and retail investors, and now wants biotech investors ahead of a Series A, Allen-Altimare said. In her view, biotech funds are often larger and a data-based case on women’s midlife health, including cardiovascular risk after menopause, is more rational than an emotional one. “All money is not good money,” Allen-Altimare said, advising founders to treat each investor “as a marriage,” interview them and ask what strategic problem they solve beyond money.

Learnings beyond startups

We believe the session’s advice, although aimed at startups, applies to anyone making a funding or budget case, including internal pitches to a CEO or board. Founders and executives should open with the decision requested and work backwards to the proof, insight and investor-fit questions. Broad market claims should be replaced with specific operating data and customer evidence, and the stage of the business sets how much is needed. We believe they should choose funders for strategic fit as well as capital, and plan for the relationship to last years. We believe they should treat follow-up as part of the pitch: build a repeatable cadence and keep prospects informed even after a “no.”

5. From AI Copilots to AI Operators: Speed of Response Becomes the New Advantage

Fragmented systems continue to slow response times

Retailers’ ability to respond quickly to changes in demand, inventory or supply-chain conditions is being constrained by fragmented systems and the manual work required to reconcile them. Several speakers noted that operational agility increasingly depends on reducing the time between identifying a problem and taking the next action.

Kelly Breakstone Roth, Co-Founder and CEO of Prysmic, described operational agility as “the speed between the unexpected and the next best action.” Her argument was that retailers have digitized individual parts of their operations but still rely on people as “middleware” between fragmented ERP, warehouse, inventory, procurement and finance systems. When conditions change, teams therefore spend valuable time reconstructing what happened before they can decide what to do.

Prysmic is attempting to compress this gap using agents that identify margin leakage, tied-up capital and lost-sales risk, propose alternative plans and execute them after human approval. The company reported up to 20% of margin recovered, up to 30% of inventory projections corrected and approximately 50 hours per week returned to customer teams. Roth stressed that these are not “set and forget” systems: an agent should be treated like a new employee, with clearly defined ownership, ongoing training and increasing autonomy as it earns trust.

Roth also warned against simply automating existing workflows. Some processes reflect years of workarounds around fragmented systems; reproducing them exactly with AI can preserve inefficiency rather than remove it.

Warehouse visibility is becoming more Important

Sankalp Arora, Co-Founder and CEO of Gather AI, made a similar argument from the warehouse perspective. He cited a $1.7 trillion annual problem caused by discrepancies between recorded and actual inventory and said 52% of inventory tracking remains manual. Gather AI uses cameras mounted on forklifts and drones to create a near-real-time view of warehouse inventory without requiring new fixed infrastructure.

The company reported reducing weekly inventory-counting hours from 4,400 to 800 and cutting pallet emergencies by 90%. Arora described Gather AI as a “ground truthing system,” intended to show operators what is actually happening in a facility rather than relying only on what systems indicate should be happening.

Structured data is still useful, but less of a moat now

As AI becomes better at using unstructured data, companies may no longer need years of data-cleaning and systems transformation before they can begin. Arora argued that historical investments in structured data still help, but their advantage is shrinking because digitization is becoming cheaper and integration faster.

Jason Brenner, SVP, Digital Portfolio at FedEx, made the same point. FedEx generates roughly two petabytes of data per day and captures approximately 200 data points on each package, but Brenner said structured data is becoming “less of a moat.” FedEx’s recently launched Global Trade Navigator illustrates how the company is trying to move intelligence further upstream, using AI to classify products, estimate duties and taxes and simplify documentation before a shipment reaches the border.

Human oversight remains central to deployment

The sessions also challenged the assumption that AI value comes primarily from replacing labor. Instead, several speakers described a model in which automation works best when it augments employees and gradually earns the right to act with greater autonomy.

Melissa Wong, Co-Founder and CEO of Zipline, positioned AI as a way to raise the performance of frontline staff. Zipline’s AI coach combines task and behavioral data with factors such as traffic, weather and inventory to recommend actions to individual employees. Wong’s objective is to move the performance distribution of stores “to the right” by scaling the judgment of high-performing managers.

Wong argued that the strongest store managers often improve performance wherever they go because of how they contextualize information, communicate and make decisions. The difficulty has historically been scaling that judgment across hundreds or thousands of locations. AI may provide a way to codify and distribute some of those practices without eliminating the human interaction that makes stores valuable.

6. Authenticity Remains Important as AI-Generated Content Increases

For younger shoppers, credibility depends heavily on whether brand communication feels real, specific and transparent. A panel of teens and Gen Alpha shoppers noted that AI imagery makes it harder to judge what a product will actually look like, described generated visuals as repetitive and objected to replacing designers, artists and models with synthetic content. Two panelists closed by explicitly telling brands to “stop with the AI ads.”

The panelists described trust as something built through imperfection. Creators who acknowledge a product’s flaws, “speak their mind” and have demonstrated long-term affinity with a product were regarded as more credible than polished endorsements. One shopper said a review that names no flaws is itself a reason for distrust.

A separate brand-storytelling session reached a similar conclusion from a commercial perspective. Paul Archer, Founder and CEO of Duel, argued that real customer stories and advocacy will matter more as AI-generated content expands.

Duel structures creators, affiliates and brand fans into communities using missions, challenges and incentives. Archer said the growing volume of authentic customer language could have value beyond social platforms, potentially influencing how recommendation engines and LLMs surface brands and products.

Brand voice remains difficult to automate safely

Customer-facing communication remains an area where several speakers were cautious about AI replacing human judgment. Tone, humor and context can be difficult to manage in sensitive or fast-moving conversations.

Glow Recipe offered an example of why human judgment still matters in brand communication. Christine Chang, Co-Founder and Co-CEO, described the brand’s response to the simultaneous launch of similar eye patches by several beauty companies. Glow Recipe combined humor with product education, explaining its ingredients and product intent while participating in a wider conversation that eventually drew in consumers and other brands.

Chang said the company had tested AI-generated customer responses but abandoned the experiment quickly because the systems could not reproduce the wit, tone and judgment required for sensitive conversations.

The point was not that AI lacked utility, but that brand voice and community management involve context, nuance and judgment that the company did not believe could yet be outsourced safely.

7. Social Commerce Is Combining Discovery, Selling and Product Feedback

TikTok Shop starts with discovery rather than search

Social commerce is changing the point at which shopping begins. Instead of starting with a search for a known product, consumers can encounter products in entertainment content and move directly into a transaction.

Patrick Nommensen, President, Strategic Initiatives at TikTok Shop, contrasted traditional e-commerce with discovery-led commerce. Traditional e-commerce tends to work well when shoppers already know what they want. TikTok Shop, by contrast, allows consumers to encounter products while watching content with no initial purchase intent.

Sarah Potempa, Founder and CEO of Beachwaver Co., highlighted the difference between traditional televised selling and livestream commerce. On QVC, she might have had 10 minutes to present a product; on TikTok Shop, a founder or employee can broadcast for an hour or even 24 hours, demonstrating products, explaining ingredients and interacting directly with customers.

This gives brands more room to educate customers, answer objections and tell product stories before the transaction, rather than relying on a short advertisement or static product page.

Livestreaming creates an ongoing feedback loop

The same interaction that supports sales can also generate product and customer insight. Livestreaming allows brands to hear questions, objections and preferences in real time rather than relying only on post-purchase research.

Potempa said some Beachwaver customers had placed 30 or even 100 orders and were asking what else they could buy, supporting expansion into adjacent areas such as haircare, fragrance and body care. She also described using community feedback to shape products and collections.

Nommensen referred to the comment and content layer as an “innovation and R&D engine,” arguing that brands can collect feedback in real time that might otherwise take six months to a year.

Left to Right: Patrick Nommensen, President, Strategic Initiatives at TikTok Shop; Sarah Potempa, Founder and CEO of Beachwaver Co.; and Neil Saunders, Managing Director, GlobalData
Source: Shoptalk Fall

 

8. Physical Retail Continues To Serve Social and Trust Functions

Physical retail retains a role for younger consumers because it allows them to assess products directly rather than relying entirely on digital claims and imagery.

Several panelists in the teen and Gen Alpha session said they prefer stores because they find it difficult to know whether online product claims are accurate. They value being able to see and touch products, browse without pressure and ask staff for help when needed.

They also described store layout, fitting rooms and knowledgeable employees as important parts of the experience.

Contributors described desired store features including seating, comfortable fitting rooms, hooks, mirrors and activities rather than static displays. One panelist noted that her local mall bans teens after 6 p.m., illustrating the limited number of places where younger consumers can gather.

Younger shoppers want products that allow them to participate in trends without looking identical to everyone else. Thrift stores, Depop, vinyl records, books and printed photos are products and channels that feel more personal and distinctive. Several panelists said young consumers dislike seeing the same products repeatedly or arriving somewhere in the same outfit as other people. They want to “stand out but also fit in.”

This gives resale and thrift an advantage beyond price. These channels can offer both perceived quality and greater individuality.

9. Agentic Commerce Introduces a New Friction Problem: Knowing Which Bots To Let In

Retailers need to distinguish authorized agents from malicious bots

As AI agents begin to participate in shopping, retailers need a more nuanced approach to automated traffic. The challenge is no longer simply to block bots, but to identify which ones are acting legitimately on behalf of customers.

Ori Snir, Chief Product Officer, Prove, described the need to distinguish legitimate agents from malicious bots and connect an agent to a verified person and the permissions that person has granted.

This is different from the traditional approach of blocking automated traffic. In agentic commerce, some automated traffic may be desirable if it represents a legitimate shopper.

Payment infrastructure is being adapted for agent-led transactions

Agent-led commerce also requires payment systems that can confirm authorization and complete transactions efficiently. Ashye Marcus, Global Head of Retail GTM, Stripe, described using secure payment tokens and its Link wallet to enable authorized purchases by AI agents. The company also reported a 12%–13% conversion lift in an A/B test from dynamically presenting relevant payment methods at checkout.

Retailers may face a growing protocol-management burden

The emergence of multiple agentic-commerce standards creates an integration challenge for retailers. Firmly, a commerce technology company, positions itself as an integration layer between merchants and emerging AI-shopping channels. The company noted that protocols are changing quickly and may be implemented differently by platforms such as Google and Meta.

Retailers could therefore face a significant maintenance burden if they integrate each channel directly. Firmly’s value proposition is to absorb that complexity on the merchant’s behalf.

Product data affects whether retailers appear in AI-led discovery

As discovery shifts toward conversational interfaces, product data increasingly affects whether AI systems understand and surface a retailer’s assortment.

Marcus advised retailers to test whether their products appear when relevant questions are put to an LLM.

She described a retailer that was not being properly associated with jeans because its product descriptions used the term “denim” instead.

The implication is that product data is becoming a distribution issue as well as a merchandising issue. Retailers need descriptions and attributes that are legible not only to shoppers and search engines but also to AI-based discovery systems.