Delight AI

Delight AI is Sendbird's customer-facing revenue agent. It handles buying conversations and support requests across chat, SMS, WhatsApp, email and voice, completes actions like bookings, upgrades and refunds, and connects to Salesforce, Zendesk and similar systems.

Delight AI is a customer-facing revenue agent built by Sendbird, aimed at companies that treat support conversations as a sales surface rather than a cost centre. Unlike AI SDR tools that work for the sales rep, Delight speaks directly to the shopper on the site, the guest part-way through a booking, or the member inside an app, and carries that buying conversation through to a confirmed order. The same agent also answers ordinary service questions, which is where it differs from a pure sales assistant: it recognises the upsell or recovery moment hiding inside a support request. In practice the platform is organised around a five stage loop. An agent memory layer holds identity, preferences and intent so customers are not asked the same question twice. An omnipresence layer decides when to reach out, and an actionbook of book, upgrade and renew tasks lets the agent act rather than just chat. An autonomous supervisor keeps a long-running case moving across days (coordinate, follow up, complete), while a trust layer enforces boundaries, approvals and human oversight. That last piece matters for the harder jobs: a damaged order that needs a replacement reserved, a refund approved by a team lead and a carrier redelivery booked before Friday can be coordinated across systems and outside partners in one thread. Channels cover chat, SMS, WhatsApp, email and voice, with one continuous thread so a conversation can start in chat and finish on a call. The agent connects to Salesforce, Zendesk, ServiceNow, Intercom, Freshworks, Sprinklr and NICE CXone, pulls knowledge from Confluence, Notion and Google Drive, and reaches commerce, booking and payment systems through their own APIs. Named deployments span retail, travel, airlines, healthcare and software, with BJS, Norse Atlantic Airways and Mixpanel published as case studies. This is enterprise infrastructure rather than a self-serve tool. Buyers evaluating a helpdesk AI or a support chatbot should compare it against those on price and simplicity; teams whose revenue runs through live customer conversations should compare it on whether the agent can actually complete the transaction.