In November 2022, a French engineer named Guillaume Luccisano put a small AI app on the Gorgias store. It answered ecommerce support emails. This was days before OpenAI released ChatGPT and taught the entire internet what a language model could do. Luccisano wasn't trying to start a company - he had already started two. But within days, dozens of merchants asked him to turn the prototype into a real product. So he did.
That product is Yuma AI, a member of Y Combinator's Winter 2023 batch. The pitch is narrow on purpose: automate customer service for large Shopify brands. Not "reimagine work." Not "the future of everything." Just the specific, repetitive, expensive grind of answering the same questions a few hundred times a day - where is my order, I want to return this, change my subscription, is this back in stock.
Reply is easy. Resolve is the business.
Most tools branded "AI support" do one thing: they draft a nice-sounding answer and hand it to a human, or paste it into a chat window. Yuma's argument is that the answer was never the hard part. The hard part is the refund that has to be issued, the subscription that has to be paused, the order that has to be looked up in Shopify and cross-checked against the shipping carrier. Yuma's agents are built to take those actions inside the merchant's connected systems, then confirm they're done.
That is why the company describes what it sells as end-to-end resolution rather than deflection. A deflected ticket is one a customer gave up on. A resolved ticket is one that's actually closed. The numbers the company publishes are about the second kind.
Who's actually using it
Yuma's customers are mid-market and enterprise ecommerce brands, most of them on Shopify. The names it points to are recognizable: Glossier in beauty, EvryJewels in accessories, Clove in footwear, Javvy Coffee, Petlibro, MyVariations. These are companies with real ticket volume - the kind where a 24-hour email backlog quietly costs repeat purchases.
The results the brands report are specific enough to be checkable. Glossier says Yuma hits 91% accuracy on "where is my order" (known in the trade as WISMO) and cut response times by 87%. EvryJewels reports 89% automation, a 63% reduction in support cost, and more than 150,000 tickets handled. Javvy Coffee says its response time dropped from a full day to twelve minutes.
The four products, one brain
Yuma started with support and has since spread across the customer journey. The through-line is a shared knowledge base: the AI answering an Instagram DM and the AI processing a return draw from the same source of truth, so the brand doesn't sound like three different companies depending on the channel.
Support AI
Resolves order status, returns, refunds, exchanges, subscription changes and product questions - with real actions, not just replies.
Sales AI
Answers pre-purchase questions and recommends products to nudge shoppers toward checkout.
Social AI
Handles Instagram and Facebook DMs, comments and reviews with brand-safe responses.
Chat AI
An AI-native live chat widget that sits directly on the storefront.
Two newer pieces sit on top. Flows, launched in 2025, is a no-code agent builder for merchants who want precise, reliable automations without an engineer. Ask Yuma lets a support lead build those automations from existing documents and diagnose problems in plain English.
A layer, not a replacement
One decision explains a lot about why Yuma gets in the door: it doesn't ask brands to abandon the helpdesk they already use. It plugs into them - Gorgias, Zendesk, Kustomer, Salesforce Service Cloud, Gladly, Re:amaze, Front - and into the commerce stack underneath: Shopify, WooCommerce, Magento, Recharge, Klaviyo, plus shipping tools like ShipStation and AfterShip. The AI becomes a layer on top of the tools a CX team already knows, rather than a rip-and-replace migration nobody has time for.
The founder who's done this loop before
Luccisano is on his third Y Combinator company. He started as an early engineer at Justin.tv, the livestreaming startup that became Twitch. He co-founded Socialcam (YC W12), sold to Autodesk for $60 million. Then Triplebyte (YC S15), a technical-hiring platform. Yuma is number three.
Third-time founders tend to pick unglamorous problems on purpose, and ecommerce support is about as unglamorous as software gets. The team is deliberately small - roughly 32 people split between the Boston/Cambridge area and Barcelona - and describes itself as "small in size, but mighty," with a stated obsession over product and shipping speed. In an AI market where models change every few months, that posture is the strategy: adapt faster than the ground shifts.
Where it sits in the market
The category is crowded and getting more so. Yuma competes with Gorgias's own Automate, with Siena AI, Intercom's Fin, Zendesk AI, Ada and DigitalGenius, plus the horizontal assistants any brand could bolt on. Yuma's wedge is depth in one vertical: it is built specifically for ecommerce, specifically for Shopify-scale merchants, and specifically for taking actions rather than suggesting them. Its funding - a $5 million round announced in late 2024, backed by Google's Gradient Ventures, Altman Capital Fund, Karman Ventures and AI Grant, with angels including Twitch co-founder Emmett Shear and YC's Garry Tan - buys it room to keep that focus.
The open question, the one that matters in ten years, isn't whether AI can handle a support ticket. It clearly can. It's what a customer-experience team becomes when the machine handles the boring 80% and humans get the 20% that genuinely needs judgment. Yuma is one of the companies betting that answer is worth building for.
The accidental launch
An AI support app hits the Gorgias store days before ChatGPT; merchant demand turns it into Yuma.
Y Combinator W23
Yuma joins the Winter 2023 batch and crosses one million resolved tickets and 100 Shopify merchants.
$5M seed + Ask Yuma
Raises $5M and ships a conversational interface for building automations in plain English.
Beyond support
Expands into Sales, Social and Chat AI and launches Flows, a no-code agent builder.