Minimal AI

Minimal AI's core thesis is that e-commerce support is a workflow problem, not a chat problem — and that shows in the product. The most concrete differentiator is the action layer: the homepage names Shopify (cancel order), Monta (block shipment), Magento (create invoice), Firmhouse (upgrade subscription), Returnista (generate return form) and a generic API for anything else. Fulfilling a return isn't a drafted reply, it's a backend write, and few competitors in this price range advertise that breadth. The second differentiator is the AI Manager. You describe policies in natural language — the site's example is "a customer can only cancel their order if it was placed less than 12 hours ago in Shopify and has not been shipped according to Monta" — and the agent applies them across systems. That's meaningful because it means support leads, not engineers, own the logic. Strengths cluster around operations maturity. 95+ languages with per-brand tone-of-voice (the site's sample is "Be direct and solution driven... always use the informal 'you' form") is a real lever for European multi-market retailers. Per-topic routing — automated for order status, draft or human for product complaints — is the right primitive; most tools only let you toggle AI globally. Self-learning with a weekly, one-click-approved improvement list is a reasonable compromise between black-box auto-tuning and weeks of manual retraining. And the pricing page publishes per-ticket economics at every thousand-ticket increment from 1,000 to 20,000, with a stated comparison to the ~€3.40 average human cost per resolved ticket. That's unusually transparent for this category. Weaknesses are structural. First, you need an existing helpdesk — Minimal is explicitly not a standalone widget — so it's a layer, not a replacement, and that means two vendors in the loop. Second, the entry price of €489/month assumes roughly 1,000 tickets; for a store doing 300 tickets/month the per-ticket math is brutal. Third, the ROI banner numbers (90% automation, 93% CSAT, 80% lower cost per ticket, 5x more tickets per agent) are vendor-reported and the case studies behind them aren't itemised here; treat them as ceilings you'll approach after a training period, not day-one results. Fourth, the published price ladder only shows the top-level ticket-volume tiers — nothing on the page about SSO, audit logs or data-residency add-ons, so enterprise security buyers should verify directly. Where it fits: e-commerce brands on Zendesk, Gorgias, Intercom, Trengo or Front with no in-house ML team, shipping 1,000+ tickets/month, ideally spanning multiple languages or stores. Where it doesn't: SaaS, healthcare, financial services — anything non-e-commerce where a backend action could be a compliance event. And teams that want every reply human-approved won't get value from the action layer, which is where the cost sits.