Q&A: Sharon Gee, SVP, Product AI at Commerce
Agentic commerce is no longer a conference-stage prediction; it’s becoming daily shopping behaviour. For retailers, that shift creates a new imperative: understanding where they appear when AI assistants answer shopping questions, and how to influence that outcome.
Sharon Gee, SVP, Product AI at Commerce, has spent her career at the intersection of product data and new sales channels. In this conversation, she explains why agentic commerce is a new shopping behaviour rather than a new channel to bolt on, why the catalogue, not the checkout, is where retailers should start, and what best practice looks like in 2026.
Can you tell us a bit about your background?
My path into commerce was an unusual one. I trained as an opera singer and spent my early career touring as a spinto soprano in the US, Europe and Asia before moving into the business world. What I found when I got there was the same thing that drew me to performance: the craft of connecting with an audience. I cut my teeth on the agency and services side at Tribal DDB, Fluid and BORN Group, then moved into commerce software as Chief Revenue Officer at Skubana.
I joined BigCommerce in 2019 to lead omnichannel partnerships, and spent the next several years running our omnichannel business and then sales and partnerships at Feedonomics, which is where I really learned to appreciate the importance of quality product data, watching and supporting the largest brands in the world syndicate their catalogues to Google, Meta, Amazon and every new channel that emerged. Today I lead our product portfolios for AI across Commerce and our Feedonomics platform, which puts me right at the centre of the next channel shift: Agentic Commerce.
What’s striking, looking back, is that we’ve been here before. Businesses have been sending catalogue data to new channels for well over a decade, first Google product listings in the advertising era, then social, then marketplaces. At each inflection point there were new attributes and new methods of ingestion (new protocols) required. We are seeing this again in the agentic commerce era. Each time, the data wasn’t new; the channel and the rules of the channel were.
Agentic commerce is the next era in that same line. The difference is that this time, the reader on the other end of your product data isn’t only a human. Agents are customers too, and they need more context than before to answer a user’s query, which means we need to enrich the data further to ensure brands are discoverable on these new surfaces.
What does Commerce do, what’s your USP, and what’s special about your approach?
We’re an open commerce platform for B2C and B2B retailers, and increasingly we’re the layer that connects a retailer’s product data to every place shoppers, and now agents, are discovering and buying.
The way I’d put it is this: the data is the new storefront, the customer is the channel, and agents are customers too. A shopper might meet your brand on your site, via an AI assistant like Gemini or Copilot, on a marketplace, or in a messaging app. Our job is to make sure that wherever that encounter happens, the product data behind it is brand-approved, contextually rich, and connected to real-time inventory and price.
The combination of BigCommerce, Feedonomics and Makeswift delivers merchants an open platform, control of their owner experiences, plus deep product data enrichment, governance and syndication capabilities — this data sovereignty is the differentiator. It helps retailers do two jobs at once: reach agents, meaning be findable, recommendable and transactable wherever agents are shopping, and run with agents, using AI to help their own teams operate the business and offer personalized agentic experiences for their customers, on their owned channels.
Everything starts from the catalogue, because everything downstream depends on it. Agents can already read anything a human can read. That was never the problem. What most catalogues lack is the context an agent needs to answer a real question: occasion, finish, activity, compatible products. A catalogue that used to need a handful of fields now needs additional conversational attributes. And those attributes have to be brand-approved, not machine-guessed, which is why our approach builds in validation and a human in the loop, plus live inventory and price, because a recommendation is only as good as the moment it’s delivered.
I’m always wary of “one click and it just works” stories. The credible version includes the enrichment work, the approval step, and the systems that keep everything in sync.
What advantage does that add?
Visibility, first. Almost every conversation I have with a retailer right now starts with agentic discovery: when a shopper describes what they want to an AI assistant, like “I’m taking my kids camping for the first time, my daughter just turned six, what do I need?” That’s a question no single brand can answer and a category expert can. So how does the agent surface your product, and does it describe it correctly? Retailers are already winning or losing that answer today.
Then, for the right categories, transactability. Agentic checkout works today for products a shopper will buy without much deliberation, the things that fit in a box and ship easily, like apparel, footwear, beauty and small consumer electronics. High-consideration purchases will still hand off to the brand’s own site or agent, and that’s fine. Businesses stay in control with quality data either way: checkout eligibility is set per product in the feed, so you can enable the pillow and hold back the couch. And critically, the retailer stays merchant of record; the commercial relationship doesn’t move even if the front end shopper experience does.
How are retailers using your systems to gain competitive advantage, and what does best practise look like? Can you share a case study?
Retailers who win across channels aren’t just doing something new; they’re doing something consistently. The same discipline that makes a listing perform on Google Shopping makes it perform on social, a marketplace, or an AI surface: clean data, accurate attributes, and real-time signals that answer a real query. Feed management has always been the foundation of multi-channel growth. The channels just keep multiplying, and agents are simply the newest reader on the other end of your product data.
New Balance, working with their agency Brave Bison, is a good example. Their old feed tools couldn’t keep pace with ambitious growth targets. Data updated too slowly, accuracy slipped at scale, and out-of-date stock meant ads showing sold-out products, clicks with no sale, and a real risk of account suspensions.
After switching to Feedonomics, Brave Bison refreshed listing cadence in Google Merchant Center, fixed localised inventory and sale pricing across markets, and then enriched the catalogue itself, importing best-seller data, grouping products with custom labels, and testing attributes from shorter mobile titles to leading with model names like Fresh Foam and FuelCell. Over a year that drove cost down 38%, conversion up 15%, revenue up 22% and ROAS up 95%, and let New Balance scale from a few core markets to 13.
Another great example is Euro Car Parts, the UK’s number one car parts supplier, which turned to Feedonomics to get its product data AI-ready at scale. With over 100,000 products drawn from fragmented sources, manual preparation simply wasn’t feasible, so Feedonomics automated the enrichment of complex product data, turning a generic “Brembo front disc pad” into a far more discoverable “Brembo front disc brake for Nissan.”
The team started with their top 100 best-sellers and scaled to 25,000 enriched listings, boosting AI search visibility and driving a 20% increase in ROAS. This built on an already strong foundation of feed quality: Feedonomics improved GTIN coverage from 58% to 74% in just six months and helped maintain a Google Merchant Centre accuracy rating above 99%, earning Euro Car Parts recognition from Google as having one of the best feeds in retail. It’s a clear demonstration that enriched, AI-ready data isn’t just about cleaner feeds; it’s about staying visible and competitive as search grows more conversational and AI-driven.
Are there other companies you partner with?
The AI surfaces themselves, first and foremost. Retailers today can be discovered in OpenAI and Google AI surfaces like Gemini, and complete checkout on Google UCP and on Copilot via PayPal, alongside the established marketplace and social channels. We also work with a broad ecosystem of agencies and technology partners who handle everything from catalogue strategy to fulfilment. If the data is the oil, the protocol is the pipeline. The plumbing matters, but the retailer shouldn’t have to think about it. Our job is to keep them connected as new surfaces and standards emerge.
What challenges are retailers facing in 2026?
The biggest is that discovery has moved onto surfaces retailers don’t own, while the back end still has to work, where the retailer remains merchant of record. Shoppers now describe an aesthetic, an occasion, a use case, not just keywords, and an agent synthesises the answer rather than ranking a list. The brands that win are the ones whose data lets that agent answer confidently and accurately. A catalogue rich enough in context to answer a real question, brand-approved so every answer stays on-voice, and connected to live inventory and price, that’s what drives discovery and conversion, everywhere your shoppers are.
For multibrand retailers there’s a genuine disruption to be honest about: the discoverability they used to provide across many brands is now partly happening inside the answer engine. But their edge is real, category and cross-category context plus loyalty.
How will you address these challenges and turn them into successes?
By treating the catalogue as the strategic asset it now is. Most product catalogues don’t hold the context a shopper needs to make a confident purchase decision. The gap isn’t readability, it’s the ability to generate the conversational attributes needed to drive AEO and GEO. Those aren’t just data problems. They’re lost sales and broken brand trust. Retailers who treat their product data as a living asset are the ones converting across every channel they enter. Google and other channel partners have introduced new ‘Conversational Attributes’ specifically to ensure they have the data needed for their AI to make good recommendations. If you aren’t optimizing and delivering this new data, you’re missing opportunity.
The second move is the owned brand agent; trained on your catalogue, customers, orders and policies, deployed wherever you choose: your site, your app, messaging, in store. The intelligence travels with the brand; the surface is a deployment choice. That’s the part no third-party platform can replicate for you.
Any final thoughts?
Websites aren’t dead, they’re changing. The site is where a brand’s uniqueness and its owned agent live. And none of this should feel like a leap into the unknown. Brands have been sending catalogue data to new channels for over a decade; the need for data isn’t new, the channels and the protocols are. You already know how to do this and partners can help. The shoppers have simply started talking to your brand the way they’d talk to a person, and now your data has to be able to talk back.
To find out how Commerce can help your retail operation, visit them online here or connect with them here.
