There are some really cool products coming out of Google Labs right now. Quick background: Labs exists to translate Google DeepMind research into real workflows. Not demos - actual use in day-to-day work. Their latest release is Pomelli. Pomelli isn’t another AI tool that writes copy or spins up campaigns. There are plenty of those, and they mostly just create more content someone still has to align with. Pomelli focuses on the coordination layer. Marketing falls apart because context is scattered: • Strategy in decks • Messaging in docs • Creative in Figma • Performance in dashboards • And the real rationale in people’s heads Pomelli holds the shared narrative and keeps it consistent across channels and teams a as things change. That’s the distinction: Most AI tools produce output. Pomelli maintains coherence. Marketing is just the first domain where this is obvious. But the underlying shift is larger: If a model can sustain shared context, it can reshape any multi-team workflow. That’s the actual significance. Check out their demo!

Exciting to see our Agentic Commerce APIs in action! Lane just demonstrated the seamless end-to-end flow we’ve been building - from AI agent discovery through native payment execution. This is exactly the kind of frictionless commerce experience that will define the next era of payments. Proud of Reshmi Suresh, Harsh Mehta, and Simon Farrow for making this integration so developer-friendly and easy to use for merchants’

One of the most underrated parts of Google's UCP is how it extends something merchants already know and are familiar with: the Merchant Center. Historically, the merchant center was mostly about a single, canonical product listing. You uploaded a feed with your SKU, title, image, price, availability, maybe couple of basic attributes. That data powered Shopping ads, comparison surfaces, and marketplaces. It was descriptive but not enough for a product to truly match complex agentic user intent. To make accurate recommendations, agents need more granular and standardized information. Shoppers now give AI agents far more context - dozens of words about who they are, how they’ll use a product and what they prefer - but most product catalogs were never built to absorb or respond to that level of detail. UCP changes that. With UCP, the product model becomes multi dimensional. Instead of one flat listing, products can now carry multiple structured attributes and tags that describe not just what the product is, but how it can be used, sold, restricted, bundled, delivered, returned, substituted, and a whole lot more. As an example, Before a jacket SKU might look like: 'Men's jacket - $150 - In stock - Free returns' Now: - Color = black, navy, olive - Sizes = S - XL - Navy + Size M = out of stock - Black + Size L = next day eligible  - Olive = UK only - Return window = 30 days  - Condition = unworn, tags attached  - Promo = expires in 48h  + other descriptive fields including substitutes, shade variants, touch and feel, etc. This to me is the real unlock. Excited about our partnership with Google and enabling UCP for Worldpay merchants. https://lnkd.in/etfdRCD8

Amazon quietly rolled out a new ad surface inside its AI shopping assistant - and it’s… interesting. Instead of the usual sponsored listings, the AI now answers your question and slips in a sponsored response. So if you ask, “Does Belkin have a magnetic power bank?” the AI gives you an answer… but the top recommendation is actually an ad. On one hand, this feels like a logical evolution. If users shop through AI conversations, then ads will follow the conversation. But as a user, I’m torn. When I ask an AI for help choosing a product, I’m doing it because I want a neutral guide, not a slot machine that injects paid recommendations into what looks like an objective answer. And in a conversational interface, the line between “helpful suggestion” and “sponsored nudge” gets blurry very quickly. A simple example: If I ask the AI, “What’s the best power bank for travel?” and it confidently recommends one - I shouldn’t have to pause and wonder whether it’s because of product quality or because someone paid for that position. Discovery becomes less transparent, and trust becomes harder to earn. If AI becomes the layer we rely on to simplify decisions, it also becomes the layer that needs the highest level of clarity around “why this recommendation?” Right now, I’m curious - but cautiously so. A helpful assistant shouldn’t feel like it’s whispering on behalf of whoever bid the most.

Big milestone at Lane: we completed a live, end-to-end purchase natively inside an AI agent, powered by Worldpay Agentic Commerce APIs and Lane. No web scraping or clunky URL redirects. Just Lane-generated, merchant-owned APIs that allow the agent to achieve perfect product search and seamless checkout.

Pretty wild chart from a16z speedrun / Ramp The top 1% of companies are spending 600x+ more on AI per employee than the median company. The interesting bit for me is how uneven adoption is becoming. At ~$12 per employee per month, the median company is probably still largely buying copilots and experimenting. At thousands of dollars per employee, companies are likely using APIs, coding agents and increasingly putting AI directly into workflows and products. That could create a very different adoption curve from SaaS or cloud. The companies using AI heavily learn where it works, redesign more workflows around it and then have more confidence to deploy it elsewhere. The gap can compound rather than simply close over time. Of course, spend ≠ value. Some of this will be wasted, and AI-native companies skew the numbers. But if even part of that 600x difference translates into productivity, the bigger question may soon be less ‘are we adopting AI?’ and more ‘how much of the company actually runs differently because of AI?’ The more AI spend scales, the more bullish I get on the orchestration layer. The advantage won’t be access to intelligence, but how effectively you route and deploy it.

I've written a version of this post before. And the month after that. Every time I think the agentic commerce landscape has settled enough to explain, three new frameworks drop, two networks announce something, and someone invents a new acronym. "Find me a silk scarf, under $150, deliver by Friday." For that to really work and not just open a browser tab, a few problems need solving at once. The merchant has to be readable by machines, not just humans. Most websites are built for eyes. An agent scraping buttons and dropdowns is slow and brittle. WebMCP is a merchant saying: I'm open for agent business. Then someone has to run the transaction across any merchant, any agent, any platform. Without a shared language it's N integrations times N retailers. Google UCP is that shared language. The checkout equivalent of HTTP. Then you need proof the agent bought what you actually asked for - not something that just fits the parameters. Mastercard Verifiable Intent creates a cryptographic record of your original instruction and checks the final cart against it. Not "was a card used" but "is this what the person wanted." Then the cart needs locking the moment you say yes. Google AP2 does this. Item, price, merchant - all signed. The agent can't quietly swap anything out after you've approved. Then the credential has to be safe for an agent to hold. You're not handing an AI your card number. Visa Intelligent Commerce provisions a token scoped to this agent, within limits you set. Not your card - a derivative of it. And the merchant needs to know the agent at the door is legitimate. A well-behaved AI agent looks a lot like a bot. Cloudflare Web Bot Auth lets the agent cryptographically prove it's acting for a real person. No CAPTCHA. Just a signed header. Finally, your card company needs to understand why you bought something, not just that you did. American Express ACE matches your intent to the outcome - and if the agent got it wrong, you're covered. Nobody else has committed to that yet. One sentence in. One delivered scarf out. The reason it looks like alphabet soup is that we're watching seven different problems get solved simultaneously, by different players, in public. WebMCP - discoverability. UCP - interoperability. Verifiable Intent - trust. AP2 - authorization. VIC - credentials. Web Bot Auth - identity. ACE - liability. Next month there'll probably be three more. And that’s what makes this space exciting. Proud to be part of the Global Payments Inc./ Worldpay team helping merchants navigate and enable all of this - and a whole lot more. Reshmi Suresh James Fry Simon Farrow Ankur Sharma Elizabeth Sheena Cherian Stephanie Torget Shane Mellor-Wolstenholme

Everyone’s arguing about whether AI means people will build instead of buy software. Most companies didn’t buy software because it was impossible to build. They bought it because running software is a liability. Ownership comes with operational drag, decision debt, and a long tail of maintenance that compounds over time. So no, this doesn’t end in everyone rolling out their own tools. But it also doesn’t end well for a lot of SaaS companies. The real impact of AI is that it removes scarcity on the supply side. Building stops being the constraint. Distribution, trust, and adoption become the bottleneck. When that happens, categories don’t collapse but they overpopulate. You still have demand. You just have far more vendors chasing it. And that’s what will dramatically change. Features converge faster, switching becomes easier to justify, and pricing power erodes by default because alternatives are always one click away and good enough is everywhere. Margins get squeezed and incumbents can no longer justify sky high prices. Growth starts leaning harder on GTM, brand, bundling, and distribution leverage rather than product advantage alone. I’m very bullish on AI. That’s exactly why I’m sceptical of software companies whose moat was mostly that they got there first and built something hard.

I'll be joining Reshmi Suresh Scot Wingo and Idan Levin 🔵 for a discussion on how agentic commerce is reshaping the shopping landscape. A lot of the conversation about agentic commerce gravitates toward payments and checkout. In reality, that's the easy part (for now). The harder problems are upstream: - how agents interpret intent - how back-end systems surface inventory/products - how decisions get scoped, constrained, and delegated - how authority is expressed across systems - how workflows move from human-led to system orchestrated We'll be discussing how different frameworks are emerging across the ecosystem - from human in the loop approaches like ACP and UCP, to more autonomous patterns like AP2, and the broader coordination layers such as MCP that makes cross system agent operations possible. Look forward to a practical conversation on what this means for merchants, platforms, and infrastructure teams building in this new landscape. P.S. We'll probably make some interesting predications too ;) Join us live on February 4th! https://bit.ly/4sFUfZX Worldpay ReFiBuy.ai Retailgentic nekuda

From $2T in Payments to AI Agents: Inside Worldpay & OpenAI’s New Partnership Worldpay has partnered with OpenAI to support the Agentic Commerce Protocol (#ACP)—allowing AI agents to make payments safely across merchants, without acquirer lock‑in. https://lnkd.in/gPsT3CbM 1️⃣ Why It Matters Worldpay processes $2 trillion annually and powers 1 million+ merchants. That scale makes its ACP adoption a turning point for agentic commerce. Combined with its #Trulioo partnership to create a Know Your Agent (#KYA) standard, it introduces real infrastructure‑level trust for AI‑driven transactions. For merchants, ACP support arrives with almost no friction—no new SDKs, no acquirer changes. Worldpay’s credential layer lets any connected merchant accept agent purchases using existing rails. 2️⃣ How It Works When an agent checks out, Worldpay handles token issuance, credential refresh, and fraud defense—then routes the payment through normal rails. The experience stays seamless for both merchants and consumers. 3️⃣ Strategic Implications Merchants gain access to a new, AI‑native commerce channel where product discovery and checkout converge. Worldpay becomes the bridge linking traditional payments with the emerging agent economy. 🔍 Next Question: Who will own the agent’s credential layer—the wallet, the acquirer, or the AI platform? That answer will define the future of trust in autonomous payments. #AI #AgenticCommerce #Payments #FutureOfCommerce #Worldpay #OpenAI #Fintech #KYA #Trulioo #ACP