Yuma AI

Where Yuma AI sits

Yuma is an AI support agent built specifically for ecommerce. It started life in 2023 as an experimental app on the Gorgias App Store, founded by Guillaume Luccisano — his third company after Socialcam and Triplebyte — and went through Y Combinator's Winter 2023 batch before raising a $5 million round to build out ecommerce support agents. That origin story shows in the product. Yuma does not try to be a horizontal customer-service platform; it tries to close tickets that Shopify and direct-to-consumer brands actually receive: where is my order, I want to return this, please change my subscription, does this come in blue.

The positioning has widened over time — the homepage now describes an "AI operating layer" spanning support, sales, social and chat, all sharing one knowledge base, brand voice and set of safety controls — but the centre of gravity is unchanged: high-volume, repetitive, data-driven ecommerce support conversations, resolved end to end inside the helpdesk the merchant already uses.

What it actually does

Four products sit on the shared layer. Support AI handles the ticket work: WISMO (where is my order), returns, refunds, exchanges, subscription changes, order edits and repetitive product questions. Sales AI answers pre-purchase questions and recommends products. Social AI replies to Instagram and Facebook DMs, comments, ad comments and reviews under moderation rules. Chat AI adds AI-native live chat to the storefront with escalation back into the helpdesk.

What distinguishes Yuma from a chatbot bolted onto a storefront is that it operates inside the helpdesk stack rather than around it. It works with Gorgias, Zendesk, Kustomer, Salesforce Service Cloud, Re:amaze, Gladly, Front, Sprinklr and Zoho Desk, and connects to the commerce systems where the work actually happens — Shopify, Recharge, Loop, Klaviyo, Yotpo, ShipStation, ShipBob, ShipMonk, DHL, AfterShip and more. That matters because a resolved ecommerce ticket is rarely just an answer. Checking shipment status, reading subscription state, validating return eligibility, updating an address, applying a tag, issuing a refund, creating a label — the agent has to do those things, not narrate them. Yuma's stated design goal is to read live commerce data, follow approved workflows, take real actions and escalate when a human should own the conversation.

The rollout model is deliberately conservative: a dedicated account manager connects the stack, maps the highest-value intents, configures brand voice and policy guardrails, and launches gradually against an agreed success target and a controlled share of eligible conversations. Hard limits stay on refunds, discounts and replacements; the CX team reviews AI replies before automation expands. There is a free 30-day trial on real ticket data, and no helpdesk migration is required.

On pricing, Yuma uses outcome-based pricing: you pay for conversations the AI fully resolves, and not for messages generated or conversations escalated. There is no public rate card — the pricing page is a demo-request flow — but the Shopify App Store listing does publish tiers at time of writing: a Small Merchant package at $850 per month and a Medium Merchant package at $1,200 per month, each including a package of automated tickets with overage fees beyond it. Third-party analysis describes the pricing page as quote-based, scoped to ticket volume, channels and implementation depth.

The vendor publishes strong outcome claims — up to 89% of conversations resolved autonomously, with case studies citing EvryJewels at 89% automation and 63% cost reduction across more than 150,000 tickets in a peak month, Javvy Coffee cutting first response from 24 hours to 12 minutes, and Glossier at 91% WISMO accuracy. These are vendor-published case studies; treat them as such.

My testing notes

Two observations from my time in the product stand out, and both are about control rather than raw automation.

First, the escalation boundary is where Yuma earns trust or loses it. In my testing, when I deliberately pushed conversations past the configured policy limits — a refund request outside the window, a discount demand — the agent stopped rather than improvising. It handed the conversation to a human inside the helpdesk with the full context: what it had checked, which policy blocked the action, and what the customer had asked for. In a category full of agents that invent refund policies under pressure, that behaviour is the single most important thing to verify, and Yuma's guardrail approach held in my runs. The counterpart is that the automation ceiling is set by your own policy configuration: if you lock down every action, the resolution rate drops, and if you open everything up, you own the errors.

Second, the helpdesk-native design changes the economics of trying it. Because Yuma operates inside Gorgias or Zendesk instead of replacing them, my setup test needed no ticket migration and no routing rebuild — the agent works the queues that already exist, and the "review before scaling" pattern means you can start with a narrow slice of eligible tickets. What I noticed in the configuration work is that the knowledge preparation is front-loaded: the agent's answer quality tracked the cleanliness of my documentation almost one-to-one. Ambiguous return-policy pages produced confident-sounding but inconsistent answers until I fixed the source text. Yuma does not repair your knowledge base for you; it faithfully reproduces what you feed it, plus whatever live order data it can read.

One more note on the social side: Social AI's moderation rules and brand-tone enforcement are a genuinely useful extension for brands where Instagram DMs have become a support channel nobody staffs at 2am, though I would keep tighter limits on public replies than on private ticket responses.

What it does not do

The first limit is category: Yuma is for ecommerce. If you sell B2B software, run a utility call centre, or need telephony and workforce management, this is not your product — there is no voice channel, no workforce scheduling, and no general-purpose ticketing beyond what your helpdesk already provides.

Second, pricing transparency is weak. The Shopify listing gives public anchors ($850 and $1,200 per month at time of writing), but the main commercial path is a demo and a custom quote, and the outcome-based model means your bill scales with success in a way that is only predictable if your ticket volumes are. Teams that need a fixed, auditable seat cost should model that before signing.

Third, the upfront knowledge work is real and non-optional. Every strong automation figure the vendor publishes comes from deployments with a dedicated account team, mapped intents and curated documentation. Self-serve buyers who connect a messy helpdesk and expect 80% automation will be disappointed.

Finally, the company is young. Founded in 2023, YC-backed, a $5 million round — the product is moving quickly (Deep Search, Guidelines and a social product have shipped recently), but the support and account-management machinery behind enterprise contracts is still being built out. The claims about self-improvement and autonomous resolution are vendor claims; independent benchmarks are scarce.

Verdict

Rated 3.9: core functionality depth is strong for its niche — end-to-end ticket resolution with real commerce actions, guardrails and helpdesk-native operation — value is reasonable at the published Shopify entry points against the human headcount it displaces though quote-based above that, usability is helped by guided onboarding and no-migration rollout but front-loads knowledge work, the integration ecosystem is unusually rich for a young company, and reliability signals are promising but limited by how new the vendor is. For an ecommerce brand drowning in WISMO and returns tickets on Gorgias or Zendesk, Yuma is one of the most purpose-built options I have tested. For anything outside ecommerce, or for teams unwilling to do the documentation work, the fit is poor.

Sources

• https://www.yuma.ai/
• https://yuma.ai/pricing
• https://yuma.ai/support-ai
• https://apps.shopify.com/yuma
• https://ycombinator.com/companies/yuma-ai
• https://www.ycombinator.com/launches/Lsn-yuma-ai-agents-for-e-commerce-customer-service
• https://yuma.ai/blogs/yuma-ai-raises-5-million-to-transform-e-commerce-customer-support-with-advanced-ai-agents
• https://eesel.ai/blog/yuma-ai-pricing
• https://yuma.ai/compare-yuma-ai
• https://yuma.ai/case-studies/how-evryjewels-achieved-70-automation-slashed-cost-by-63-and-process-over-150k-tickets-with-yuma