AI Shopping Agents Are Here: What Firmly's $15M Round Means for Your Online Store

JPMorgan and Mastercard just put real money behind an idea that sounds futuristic but is quietly becoming plumbing: your next customer might not be a person scrolling your website at all, it might be an AI agent shopping on their behalf. Seattle startup Firmly raised a $15 million Series A on November 5, 2026, led by Squarepoint Capital, with J.P. Morgan Payments, Mastercard, Samsung NEXT, ARK Invest, Alumni Ventures, Benhamou Global Ventures, Granada Funds, Hike Ventures, The Mercantile Companies and Versus Ventures all participating. That takes Firmly's total funding to roughly $20 million. The exact cheque size from JPMorgan or Mastercard was not disclosed, and neither was Firmly's valuation, revenue, or merchant count. But the interesting part for an online store owner isn't the funding figure. It's the direction two of the world's biggest payment companies are pointing in.

Alongside the funding, Firmly launched two products. The first, Ask Agent, is a conversational shopping tool merchants can put on their own website. It lets a shopper type or speak a natural-language request, pulls answers from a real-time product catalog, handles checkout right inside the chat, and can also answer standard support questions. The second, Firmly Connect, is a no-code layer that plugs a retailer's existing commerce stack into AI agents, marketplaces, social platforms, and shoppable ads, so a merchant doesn't have to build a separate custom integration for every agentic channel that shows up.

In practical terms, Firmly is trying to do for AI shopping agents what a payment gateway did for cards twenty years ago: one connection, many downstream channels. Whether or not Firmly specifically becomes the standard, the problem it's solving is real and growing.

Because they're already building the rails underneath it. Mastercard launched Agent Connect in September 2026, a service meant to link merchants, AI agents, and payment providers through a single integration rather than dozens of bespoke ones. According to GeekWire's reporting, Firmly handles the merchant-facing side of that integration, and Mastercard is actively recruiting merchants to join. That gives Firmly a built-in distribution channel, and it gives Mastercard a working partner to actually onboard stores, rather than just an API nobody uses.

For a card network and a bank, the incentive is obvious. If shopping increasingly happens through an AI agent instead of a browser tab, whoever controls that checkout moment controls the transaction. Mastercard and JPMorgan would rather be inside that moment than locked out of it.

Strip away the jargon and it's simple. Instead of a shopper opening your website, scrolling, filtering by size and price, and clicking 'add to cart,' they tell an AI assistant what they want in a sentence, something like 'find me a cotton kurta under Rs 1,500 in size M that can arrive by Friday.' The agent searches multiple stores, compares options using whatever data it can read, and in the more advanced setups like Ask Agent, completes the purchase without the shopper ever landing on your product page.

This is a different relationship than a chatbot widget answering FAQs. The agent is acting as the customer's proxy. It reads your catalog the way a machine reads it, not the way a human reads it, and it needs to trust your checkout enough to hand over payment details on the shopper's behalf.

Today, most small and mid-sized stores built their checkout for a human clicking through screens: add to cart, enter address, pick payment method, confirm. An AI agent doesn't click through screens. It needs a structured, predictable way to complete the same steps programmatically, confirm stock, apply the right price, process payment, and get back a clear success or failure response it can relay to the shopper.

That's a meaningfully different engineering requirement than a nice-looking checkout button. It means your store needs dependable product and order APIs, accurate real-time inventory so an agent doesn't try to buy something that's already sold out, and a payment flow that can confirm a transaction without needing a human to squint at a screen and tap 'pay now.'

This is the part most merchants can start on right away, independent of which agent platform eventually wins.

If your checkout only works through a visual webpage, an agent can't complete a purchase through it. You want an API or webhook-based path that can create an order, apply a valid payment method, and return a clean confirmation, ideally the same backend your website checkout already uses, just exposed in a way software can call directly. You also want clear, immediate order-status updates, because an agent will often need to tell the shopper 'your order is confirmed' within the same conversation, not five minutes later by email.

Be careful here, because this is where enthusiasm often runs ahead of fact. Firmly's own announcements and GeekWire's reporting do not state whether Ask Agent or Firmly Connect supports UPI, Indian GST invoicing, rupee settlement, Indian logistics partners, or Hindi and other Indian-language product discovery. None of that should be assumed just because the underlying idea is sound. What's genuinely useful for an Indian SMB today is the pattern, not the product: get your catalog machine-readable, keep stock and price accurate in real time, and build a checkout that doesn't depend on a human clicking through a webpage. Whichever agent platform eventually becomes dominant in India, a store built this way will be ready for it.

The stores that win the agentic shopping era won't be the ones with the prettiest product pages. They'll be the ones whose data a machine can actually trust.

This is exactly the territory ODIV's multi-agent-systems service was built for. Firmly's approach, treating product discovery, agent access, checkout, and merchant-side connectivity as separate but connected pieces, is the right mental model for any Indian business preparing for AI-driven shopping, support, and ordering. ODIV builds that for you: a trained AI agent that reads your real catalog and inventory accurately, a checkout and order flow exposed in a way agents and automations can call reliably, and the connective layer between your existing systems, your website, and the chat or voice interfaces your customers actually use, without you having to hand-build a separate integration for every new channel that shows up next year.

Our engineers work hands-on in modern AI build tools like Lovable and Claude Code alongside conventional engineering discipline, which means we get a working system in front of you fast, then harden it so it's secure, properly integrated with your inventory and payments, and maintainable after launch. That combination is why commissioning this from ODIV typically lands in a fraction of the time and cost of a traditional, fully hand-coded custom development project. If you run a store and want to understand, concretely, what preparing for AI shopping agents would look like for your catalog and checkout, start a chat with us on WhatsApp and we'll walk through it.

And if what you actually need first is simpler, getting your product catalog, cart and payments working smoothly inside WhatsApp itself for customers who already chat with you there, that's a job for ODIV Engage's WhatsApp commerce tools, worth a separate conversation.