Yuma AI builds AI agents that resolve ecommerce customer conversations end to end inside the helpdesk a brand already uses. It was founded in 2022 by Guillaume Luccisano, a three-time Y Combinator founder, is headquartered in Boston with a second office in Barcelona, and says more than 140 ecommerce brands across North America and Europe use it. Its products are Support AI, Sales AI, Social AI, Chat AI and Voice AI, alongside Ask Yuma, AI Flows, Deep Search and Web Actions.
Since Yuma 2.0 launched in September 2026, one written Handbook runs every conversation across email, chat, SMS, WhatsApp, social DMs, outbound outreach and phone calls. Ask Yuma writes and updates the Handbook's playbooks and procedures in plain language, every change lands as a draft, and each playbook carries tests that replay as simulated conversations before a team publishes it. Hard limits cap refunds, discounts and coupons, every drafted reply runs through 15 to 20 quality control checks across multiple models, and questions the Handbook does not cover go to a person. Yuma says it answers in more than 180 languages and runs on models from OpenAI, Anthropic and Google.
Support AI reads live order and customer data and takes actions such as refunds, reships, address changes, cancellations, return labels and subscription skips. It works inside Gorgias, Zendesk, Kustomer, Salesforce Service Cloud, Gladly, Front, Re:amaze, Sprinklr, Zoho Desk and HubSpot Help Desk, and Yuma's integrations directory lists 62 tools across ecommerce platforms, subscriptions, returns, shipping, reviews and loyalty, including Shopify, WooCommerce, Recharge, Skio, Loop Returns, Klaviyo, Yotpo, ShipStation and Stripe. Web Actions, launched in September 2026, lets Yuma complete tasks in a real Chrome browser on sites with no API, such as carrier portals and 3PL dashboards. Voice AI answers the support line and makes order changes, cancellations and refunds while the customer is still on the call.
Pricing is outcome-based. Yuma bills only for full resolutions, meaning tickets it handled from start to finish with no human involved, counted once per ticket, and escalations are not charged. Every plan carries a 12-month commitment, and the published figures are illustrative annual prices, with the exact bundle built around a brand's forecast. Starter is $14,400 a year for 12,000 resolutions, about $1.20 each, with over-quota tickets at $1.50 on a monthly plan. Growth is $40,000 a year for about 48,000 resolutions, about $0.83 each, and adds no overage fees on annual plans, a dedicated account and technical manager, Deep Search, custom integrations and quarterly business reviews. Scale is $90,000 a year for about 120,000 resolutions, about $0.75 each, and above that volume the rate falls toward a $0.55 floor. Sales AI, Social AI, Voice AI and custom integrations are add-ons priced by scope, and Sales AI runs on its own subscription.
The first 30 days run free on the brand's own tickets, there is no implementation fee, and Yuma says most brands go live in one to two weeks, with automation rolled out gradually from 1% of eligible tickets to 100%. Every customer gets a dedicated account manager and a shared Slack channel. Named results on Yuma's site include 89% automation at EvryJewels during peak season and 79% at Petlibro.
Yuma suits ecommerce and DTC brands that want AI resolution inside their existing helpdesk on an annual, outcome-based contract. Teams that want monthly self-serve plans with AI inside the helpdesk itself should consider Gorgias, and support teams outside ecommerce should consider Intercom Fin.
Current state Q4 2026: Yuma 2.0, released in September 2026, moved every channel onto one tested Handbook, and Web Actions and Voice AI launched in the same quarter. The Yuma MCP server, launched April 15, 2026, works in Claude and ChatGPT and can now change Handbook policies and tests as drafts that a team member must publish in Yuma. Yuma holds a SOC 2 Type II report audited by Johanson Group LLP, is certified under the EU-U.S. Data Privacy Framework, and states that it may use customer data to improve its AI models unless a customer opts out. G2 named it a Fall 2026 High Performer for AI Customer Support Agents.
