Vinny O’Brien says e-commerce conversion rates have hovered around 3% for twenty years. He puts that to Darin Lynch at Shoptoberfest, and Darin’s answer is that a lot of B2B companies are measuring success in the wrong place.

Darin started Irish Titan 22 years ago. He sorts commerce history into rough eras. He calls the early one Neanderthal commerce, when sites were difficult for merchants to build and for customers to use. Then came a long stretch shaped by platforms and search engines, more trustworthy but still built around the merchant and the tech. Now a buyer can describe what they want to ChatGPT, Claude or a brand’s own agent and get to a product.

His case is that a B2B manufacturer or distributor shouldn’t judge commerce by how many orders go through its commerce platform’s checkout. Plenty of those orders come in through punch out or the ERP, and more will happen inside AI agents. He breaks agentic commerce into data enrichment, discovery and checkout, and wants a separate success metric for each.

He also talks about the Iowa farm his dad built up from 40 acres, and why he thinks Shopware’s US timing hurt it during the COVID rush. Darin is a Shopware agency partner who has emceed this event for the last two years, so weigh his view that Shopware has the best community of any platform with that in mind.

Vinny also asks about vocabulary. Every event defines agentic commerce differently, and merchants are left to sort it out themselves. Darin agrees and thinks the agencies that get through this period will help set standards, though he didn’t say how that would happen or who would lead it.

#watsonweekly #shopware #irishtitan #ecommerce #agenticcommerce #b2bcommerce

Salesforce agreed to pay about $5.7 billion in cash for Braze, the software brands like Instacart and Urban Outfitters use to send push notifications and texts. Rick works out what that comes to per customer. Salesforce hasn’t said how it will sell Braze alongside Marketing Cloud, the product a lot of Braze customers left to get there.

ChatGPT referral traffic dropped out of analytics reports for about a week in October. Profound traced the missing tracking tags to ChatGPT’s new Intelligent UI. The visits still arrived but were counted as direct traffic until the tags came back on October 16.

Walmart now sells Walmart+ and Sam’s Club+ together for about $120 a year, $19 less than Prime. Rick thinks it’s aimed more at Costco than Amazon.

Google’s universal cart is rolling out in Search and the Gemini app, with Nike, Sephora, Target and Walmart on board. Google has tried shopping checkout before and shut down Buy on Google in 2023. It hasn’t said what retailers pay.

Amazon is closing the storefronts it ran on amazon.com for Winn-Dixie, Metropolitan Market, Weis Markets and Morrisons. It wants those grocers to host Amazon lockers in their stores instead.

#Ecommerce #Retail #MarTech #RetailMedia #AgenticCommerce

UPS has now shed almost all of its Amazon volume. The 18-month wind-down took out about two million pieces a day and roughly $4.5 billion in expenses, leaving UPS with a smaller network that earns more per package but has less volume to set prices with. Rick looks at who held the leverage in that split, and whether a UPS focused on healthcare logistics is turning into a specialty carrier.

BJ’s Wholesale Club plans to cut about 20% of its products over the next couple of years, taking older clubs from around 7,500 items to 6,000 or 6,500. The last time BJ’s tried this, sales fell and it added items back. Lululemon and Under Armour have made similar cuts, as has Dollar General.

On October 8 the U.S. Court of International Trade certified a class of importers who paid IEEPA tariffs between February 2025 and February 2026 and haven’t filed their own suits. The court estimates about $30 billion in refunds would otherwise stay with the government, and its order says most of the importers who haven’t used Customs’ CAPE refund system are small. If you sell online and imported anything in that window, call your customs broker this week.

Albertsons is asking Kroger for between $8 billion and $9.3 billion over their collapsed 2022 merger, against a $600 million termination fee in the contract. The trial in Delaware began October 19.

On October 29 Rick hosts Agentic Island, a two-hour virtual summit on agentic commerce sponsored by Shopware. It includes a fireside chat with Shopware co-founder Stefan Hamann and sessions with agencies that help brands prepare for AI shopping agents. It starts at 9 a.m. Pacific. Register here: [link]

Investor minute

• Confido raised a $55 million Series B
• Marlo raised $5 million in seed funding
• Ultra Robotics raised a $50 million Series A
• Stord secured a $400 million credit facility led by Citi
• The Knot acquired Zola for an undisclosed price

Rick thinks Starbucks has lost its way. Jess thinks Brian Niccol’s turnaround is working, and she points to four straight quarters of growth.

Niccol has put handwritten Sharpie names back on cups and is spending $1.2 billion to make stores more comfortable. Rick says a better seat means nothing when you wait 25 minutes for burnt coffee. Jess says consistency matters more than quality, and that cake pops and pink drinks keep parents and Gen Alpha coming back.

Rick asks whether Starbucks should accept being the McDonald’s of coffee or go upscale with Starbucks Reserve. He also floats a Yum Brands-style parent company that buys other chains. Jess wonders if it needs a Krispy Kreme to take on Dunkin’. Rick calls Niccol’s six-point strategy a zero-point plan.

He thinks the board could push Niccol out before they cover Starbucks again.

Watson Weekly Newsletter at watsonweekly.com

Shopify has bought shop.com from Market America and won’t say what it paid. Rick Watson and Jessica Lesesky use the deal to work out where Shopify fits once AI agents do more of the shopping.

Rick thinks Shop Pay matters most. If agents skip the storefront, Shopify may be left owning the checkout and not much else, and he thinks that could still be a very good business. Jess sees Shopify going after PayPal’s share of checkout. They also get into the Shop app, which most people only open to track a package, and Shop Cash rewards that seem to disappear before anyone spends them.

Shopify’s revenue per employee has climbed fast since 2023. Rick connects that to Tobi Lütke telling staff to use AI and to regular reviews of whether each team has the right number of people. He compares the culture to Amazon’s old order of priorities, where employees came after customers and sellers. Jess wonders how that fits with Harley Finkelstein’s appetite for big agentic bets, and whether those bets will help Shopify more than the merchants who build on it.

Get the newsletter at watsonweekly.com.

#watsonweekly #shopify #shoppay #paypal #agenticcommerce

PartsTree has sold replacement parts for lawnmowers and other outdoor power equipment for about 20 years, mostly on software its own team built. By this year the code had become something nobody wanted to touch. Some of it was written more than 20 years ago, and only one person could fix it when it broke.

Vinny O’Brien sat down with Doug and Drew from PartsTree and Chris from Altus Nova at Shoptoberfest 2026 in Nashville to talk about the move to Shopware. It was Altus Nova’s first Shopware project. PartsTree’s rule had always been to change systems between late October and January, because May and June sales run about four times January’s. They went live in mid-July anyway, after years of watching their own suppliers stumble through mid-season migrations.

Chris explains why the team gave up most of its custom logic for platform defaults and kept customization for the few things core to how PartsTree works. Doug and Drew talk about bringing staff along when the switch landed in their busiest months, and about how much of the business runs on what long-serving employees know and never wrote down. Doug says he doesn’t know yet how AI agents will change the way customers buy parts. Drew has a warning from data cleanup, where an AI-written product description turned a blower into a Glock pistol. Chris thinks the bigger opening is B2B, such as shared accounts that let a landscaping crew in the field and their mechanic work out a repair together.

The new site now ships weekly releases. The team says the main metrics match or beat the old site’s, and some secondary ones are still being tuned.

I asked the retailers and brands at Retail Club to raise a hand. About ten or eleven went up. When I asked who had good data, one hand stayed.

This was our first Watson Live! The Retail AI Debates at RetailClub Presented by KIBO Commerce. Speakers argued assigned sides, some they don’t personally hold, and the room voted with red and blue plates after each round.

Debate one asked whether retailers should rebuild workflows around AI before their data is ready. Gireesh Sahukar took the yes side and won the vote. His point was that chasing clean data first never ends, and that bad inventory data starts with someone scanning blueberry yogurt instead of strawberry. Kelly Goetsch of Pipe17 argued against, partly from conviction, and pointed to Target at one point running seven master systems of record for inventory.

Casey Golden of CI&T and Kevin Reese took on whether companies will regret AI-driven layoffs. Casey said retail never wrote down what anyone actually does, so cutting people removes the context AI would need. She used Dollar General pulling back on self-checkout as an automation regret that had nothing to do with AI. Kevin won the room by arguing that companies will regret how they ran layoffs much more than the decision to make them.

For the third debate I stepped in against Danny Smith, who leads agentic commerce at Stripe and worked on Rufus at Amazon. I argued retailers should build their own agents first because it forces you to connect inventory, product data, service and delivery dates, and I don’t think you should hand that data to Sam Altman. Danny’s case was that shoppers want one agent that knows them across every store, so money spent on a branded chatbot nobody visits is wasted. I won the vote, and Danny had already said I would because I was the host.

We closed with Anatoli Iakimets from KIBO Commerce, who argued for starting now with a vendor and keeping the budget small until the technology settles.

This Watson Live event was presented by KIBO Commerce, and supported by Power Digital and Ekyam AI.

Most of the room admitted their data is a mess and still voted to start. I don’t know yet whether that was the right call for retailers whose inventory counts are already wrong.

Anthropic has filed its S-1, and Rick Watson and Jessica Lesesky use the filing to ask whether software companies are in real trouble. Reported targets put the raise at $100 billion or more at a valuation of up to $2 trillion, which would top SpaceX. OpenAI, meanwhile, has pushed back its IPO and is reportedly raising $30 billion privately at about $1.4 trillion.

Jess goes back to the software selloff earlier this year, when investors decided Claude Code and Cowork would kill SaaS, a prediction that hasn’t come true yet. Rick thinks the old buy, build or partner decision now comes down to build, because marketing teams believe they can vibe code their own tools. Jess expects systems of record like NetSuite and Salesforce to hold up and worries more about the small and mid-tier apps someone could rebuild overnight. They also get into what a moat means when code is cheap, and a website that lets prospects check whether your product is just a markdown file.

The IPO race splits them. Rick says OpenAI is making a mistake by waiting while Dario Amodei takes Anthropic public. Jess thinks Sam Altman is playing it like Apple, letting someone else go first and arriving later with something better.

Subscribe to the newsletter at watsonweekly.com.

#watsonweekly #anthropic #openai #saas #claude #ipo #ecommerce