Delight.ai
Also known as: Sendbird, Sendbird AI Agent
Enterprise AI customer agent from Sendbird for sales and support across the customer lifecycle, running omnichannel agents with a dedicated memory platform and the Trust OS governance layer for observability, control and staged release.
Delight.ai is the enterprise AI agent platform from Sendbird Inc., built on the communications infrastructure that already powers in-app chat, voice, and video for thousands of the world's largest apps. Sendbird was founded in 2013 in San Mateo by John Kim, Harry Kim, Brandon Jeon, and Forest Lee. It went through Y Combinator and pivoted from a social app for parents into a developer chat API.
The company has raised about $220 million and reached a $1 billion valuation with its $100 million Series C in March 2021, led by ICONIQ alongside SoftBank Vision Fund, Tiger Global, and Y Combinator. The platform now carries more than seven billion conversations a month for customers including Hinge, Hyundai, Match Group, Noom, Redfin, Yahoo, Virgin, and Weight Watchers.
The agent platform carries the Delight.ai name and runs on its own domain, while Sendbird's company, documentation, and status pages stay on sendbird.com. The Agent Memory Platform gives agents what the company calls a living memory and a business brain. It absorbs signals from every conversation and action to build an evolving understanding of each customer that goes beyond static CRM data, and it keeps history after a session closes.
Omnipresent agents carry that context across in-app chat, web, email, SMS, WhatsApp, social, and voice, and they reach out proactively when a customer goes quiet or switches channel. Agent Steward takes full ownership of complex, long running issues end to end, and a Voice AI Agent handles multilingual real time calls, including outbound.
Trust OS covers observability, oversight, and control as one framework. Openbook logs every conversation, decision, and tool call for review. Hallucination and unsafe content detection flags factual errors and off policy responses automatically. Kill switch and escalation controls let an operator pause, redirect, or kill any flow instantly, with a one click handoff to a person.
Operations teams author agent behavior directly through an Actionbook Editor, editing refund thresholds, escalation logic, and tone without an engineering ticket. Releases are staged. Staging environments mirror production, test suites built from real conversation IDs return pass or fail results, rollout starts with ten percent of traffic, and A/B testing measures resolution rate and CSAT before a change is committed.
Sendbird grew up as a developer platform, so the agent embeds natively in a customer's own product through APIs and SDKs for iOS, Android, JavaScript, React Native, and Unity, and it connects outward to Zendesk, Salesforce, and Freshworks. The security program covers SOC 2 Type 2, ISO 27001, HIPAA/HITECH, GDPR, and CCPA, with a live trust center. Customers get granular role based access control, full audit logs, retention settings they choose, and regional hosting that leaves them in control of where their data resides.
Customers build their own agents, and there is no template library or agent marketplace. There is no model selection or bring your own key path, and the service is cloud only with no self hosted deployment.
Vendor details
Canonical URL
https://delight.ai
Category
Customer support agent
Subcategory
Omnichannel AI agent platform
Funding status
Independent, headquartered in San Mateo, California, founded in 2013 by John Kim, Harry Kim, Brandon Jeon, and Forest Lee, and a Y Combinator company. Raised about two hundred twenty million dollars in total, reaching a one point zero five billion dollar valuation with a one hundred million dollar Series C in March 2021 led by ICONIQ Capital, with SoftBank Vision Fund, Tiger Global, Shasta Ventures, and Y Combinator participating. Its communications infrastructure powers more than four thousand applications and routes billions of messages a month for customers including DoorDash, Match Group, Noom, and Paytm.
Company status
first party product
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Delight.ai connects to Zendesk, Salesforce and Freshworks for live agent handoff, with Salesforce also supported as a knowledge source. Third party knowledge bases plug in so agents surface the right information in real time. Agents reach customers through in-app chat, web, email, SMS, WhatsApp, social messaging and voice on Sendbird's communications platform, which ships cross platform SDKs for iOS, Android, JavaScript, React Native and Unity. AI agent documentation lives at sendbird.com/docs/ai-agent, alongside a community forum, a status page and a GitHub organization. There is no MCP server or agent manifest.
In practice
Your chatbot forgets everything the moment a conversation ends, so customers repeat themselves constantly. Sendbird's agents keep persistent long term memory across chat, voice, SMS, and email, so every interaction builds on the last.
Support is fragmented across a web widget, an app, SMS, and social, each with its own bot. Sendbird deploys one AI concierge that follows the customer across all of those channels with shared context.
You want AI agents embedded natively in your product, not bolted on. Sendbird's communication APIs and SDKs let engineers wire agents into the app directly, with Agent Steward resolving issues autonomously end to end.
Sources & related URLs
Research sources
Agentic Index coverage score
11.5 / 14 capabilities · 82%
| Integrations & Tool Calling | Full |
|---|---|
|
Agents take action as well as answer, and Trust OS logs every tool call they make. Help desk and CRM connections cover Salesforce, Zendesk, ServiceNow, Intercom, Freshworks, Sprinklr and NICE CXone for handoff and case work, and commerce, booking and payment systems connect through their own APIs, which is how agents in retail and travel raise basket size and booking value. Every connection runs on Sendbird's communications platform, which carries more than seven billion conversations a month. Sourcedelight.airead 2026-11-19 |
|
| Workflow Orchestration | Full |
|
Agent Steward takes ownership of complex, long running customer issues end to end instead of answering and closing, with agents that read intent, hold context, act and escalate only when needed. Behavior comes from an actionbook the customer writes, holding rules such as refund thresholds and escalation logic, which Openbook renders as a visual flow through complex branching. Above the single case, Zero-Touch Improvement runs a closed loop in which observability catches an issue, control tests the fix and oversight holds it to policy, moving step by step from human guided to autonomous. Sourcedelight.ai/autonomous-agentread 2026-09-05 |
|
| Knowledge Grounding & RAG | Full |
|
Agents draw on knowledge connected from Confluence, Notion, Google Drive and Salesforce, with Zendesk and Freshworks connected alongside, so the agent surfaces the right answer in real time from the customer's own content. Building an agent works like onboarding a team member, transferring knowledge through the supervisor dashboard, and the actionbook holds the authored rules the agent answers within. Trust OS flags factual errors and off policy responses automatically, and the review and grade surface scores knowledge accuracy. Sourcedelight.airead 2026-11-19 |
|
| Human Oversight & Guardrails | Full |
|
Trust OS has kill switch and escalation controls that pause, redirect or stop any AI flow at once, with one click escalation to a human agent and role based access control so the right people own the right agents. Operations teams set policy in the Actionbook Editor, changing refund thresholds, escalation logic and tone live with no code and no engineering ticket. Hallucination and unsafe content detection flags factual errors, off policy responses and safety violations before customers see them, and a review and grade surface lets people score responses and flag outliers. Oversight comes through authored policy, automatic detection and interruption rather than a sign off on each action before it is sent. Sourcedelight.ai/trust-osread 2026-09-05 |
|
| Security, Identity & Governance | Full |
|
Delight.ai holds SOC 2 Type 2 and ISO 27001 and adheres to GDPR, CCPA and HIPAA/HITECH, with a live trust center at trust.sendbird.com. Controls the customer sets include granular role based access control, which also scopes which people own which agents, full audit logs, retention settings that follow its own policy, encryption end to end and TLS in transit, and PII protection with data masking. Internally, Sendbird uses FIDO2 multi factor authentication with device trust, automated detection and response and automated compliance checks, and keeps a vulnerability disclosure address. Sourcedelight.ai/securityread 2026-09-05 |
|
| Observability & Auditability | Full |
|
Trust OS logs everything for tracing, and Openbook records every conversation, decision and tool call for review. Openbook also renders the agent's actionbook as a visual flow, so an operator can follow the logic behind a decision. Low CSAT Analysis sorts every conversation scored one or two into one of five failure types before anyone opens it, and Thinking Messages show a status for each tool call during the run, so the reasoning is visible as it happens as well as afterward. A separate Analytics product tracks conversations, common topics and return on investment. Sourcedelight.ai/trust-osread 2026-09-05 |
|
| Memory & State Persistence | Full |
|
The Agent Memory Platform, sold as its own product, gives agents a living memory and a business brain, joining what the agent remembers about each customer with what the business is trying to achieve. Agents absorb signals from every conversation and action into an evolving picture of the customer that goes beyond static CRM data, and context follows a customer between chat, SMS, email and voice so nobody repeats themselves. Memory carries across separate conversations as well as within one, rather than ending when a session closes. Sourcedelight.ai/agent-memory-platform-ampread 2026-09-05 |
|
| Deployment & Data Residency | Full |
|
Delight.ai agents run inside a Sendbird application, and a Sendbird application is created in a region the customer picks at setup, with regions in the United States, Europe and Asia Pacific, Frankfurt among them, and a documented path to migrate an application to another region. Regional hosting for data residency comes with retention settings the customer sets and PII protection with data masking. No on premises or self hosted deployment is offered. Sourcedelight.ai/securityread 2026-11-19 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
Capabilities each have their own page and are adopted as units, among them Agent Steward for ownership of long running issues, Agent Memory Platform, For You Conversations, Omnipresent, Voice AI Agent, Trust OS, Analytics, Desk and AI Agent Integrations, with industry pages for retail, travel, healthcare, financial services, logistics and more. There is no template gallery, agent library or marketplace to browse and put into service, and agents are built by onboarding them like a new team member through a supervisor dashboard, with behavior written in an actionbook. Sourcedelight.ai/builderread 2026-09-05 |
|
| Triggers & Channel Coverage | Full |
|
Channels are in app chat, website, email, SMS, WhatsApp, social messaging and voice, all carried on Sendbird's communications platform, with context following customers between channels instead of restarting. Omnipresent agents start conversations themselves, anticipating needs before an issue forms and reaching back out when a customer goes quiet or switches channel. Voice 2.0 adds multilingual outbound calling in real time, so work reaches the agent from the customer and from the agent's own follow up. Sourcedelight.ai/omnipresentread 2026-09-05 |
|
| Model Flexibility & Routing | Not documented |
|
There is no word on which models run the agent or who chooses them, and no model selection, routing setting or bring your own key path. The product is sold as a managed outcome, with resolution rates, a governance layer that detects hallucinations, and Zero-Touch Improvement that tunes the agent on its own, so the model sits entirely on the provider's side and a buyer has no lever over it. Sourcedelight.ai/ai-agentread 2026-09-05 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
AI agent documentation lives at sendbird.com/docs/ai-agent, linked from the delight.ai developer menu alongside a community forum and a server status page, with a GitHub organization at github.com/sendbird. The underlying communications platform ships SDKs for iOS, Android, JavaScript, React Native and Unity with chat, voice and video APIs, which embed agents natively in a customer's own app instead of as a widget. Trust OS exposes the actionbook as an artifact operations teams edit directly. There is no MCP server or agent manifest. Sourcesendbird.com/docs/ai-agent/guide/v1/overviewread 2026-09-05 |
|
| Testing, Debugging & Optimization | Full |
|
Every change is tested before it reaches a customer, with staging, test suites built from real conversations, gradual rollout and A/B data. Staging environments mirror production with real integrations and policies. Test Upgrade builds test cases from real conversation IDs, groups them with labels, shares sets across environments and returns consistent pass or fail results. Gradual rollout ships to ten percent of traffic first, A/B testing runs two versions side by side on resolution rate and CSAT before a change is committed, and a review and grade surface gives every response a quality score. Sourcedelight.ai/trust-osread 2026-09-05 |
|
| Browser & Computer Use | Not documented |
|
The agent acts through integrations with help desks and CRMs, through tool calls that Trust OS logs one by one, and through the communications APIs and SDKs that embed it in the customer's own app. It talks to people over in app chat, web, email, SMS, WhatsApp, social and voice, and there is no browser control, hosted browser session, virtual desktop or interface automation for the agent to drive. Sourcedelight.ai/ai-agentread 2026-09-05 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Delight.ai launched Delight Customer Agent, which packages shopping, booking, support and retention into one agent that shares a single customer history across chat, voice, SMS, email and WhatsApp. The same agent that helps a customer choose a product handles a later support issue or renewal, acting within company approvals and oversight.
Bears on: Agent capability
View sourceDelight.ai introduces Intelligent Voice Handoff, which keeps a caller connected while the agent reaches and briefs the person taking over. Answer detection distinguishes people from voicemail, while recovery handles busy lines and unanswered calls by returning to the caller, leaving a message, or creating a follow up case. Supported integrations can attach the summary, transcript, and sentiment to a ticket before the transfer.
Bears on: Human approval / guardrails
View sourceSendbird launched the Delight Agent MCP, allowing users to connect AI assistants like Claude, ChatGPT, Cursor, and Codex directly to live Delight.ai data. The release includes access to over 35 operational tools for monitoring agent performance, searching transcripts, and updating agent configurations.
Bears on: MCP / tool calling / API
View sourcePricing
No published rate for the AI agent platform, which is quoted through sales. The separate Sendbird Chat SDK is billed per monthly active user with public list pricing.
Sendbird Chat bills on monthly active users and message volume, with published tiers setting each plan's MAU band.
What is public
Delight.ai, Sendbird's AI agent platform, has no public rate, tier or billing unit and is quoted through sales, with an ROI calculator and an RFP template in place of rates.
Billing mechanics
Agent pricing inquiries go to Talk to an expert.
Cost watchouts
Premium channels such as voice and SMS carry their own usage costs on top of the base messaging subscription.
Variable cost rationale
Sendbird Chat spend grows with audience size and interaction volume, while what the agent layer adds is set in each quote.
Additional watchouts
Sendbird's Chat MAU pricing is not the agent price, since they are separate products on separate domains. Trust OS, Agent Steward, the Agent Memory Platform and Voice are marketed as distinct capabilities rather than one bundle, and no tier is stated for Trust OS 2.0.
Sales call required
Yes, required for paid access
Free / trial
None for the agent platform, and Sendbird Chat has a free trial up to 1,000 MAU
Missing data
Any rate, unit, cap or volume table for the Delight.ai agent platform.
Related vendors
- Decagon — AI agent platform purpose-built for customer support, with deep…
- 5.Y — 5
- Ada — Agentic customer experience platform whose AI agents autonomously…
- Aisera — Enterprise agentic AI platform that orchestrates specialized agents…
- Amazon Connect Customer — Amazon Connect Customer, called Amazon Connect until April 2026, is…
- ASAPP — Generative AI for enterprise contact centers that plugs into…
Alternatives to Delight.ai
Other products that do the work Delight.ai does, and what each one is.
- SinchDelight.ai vs Sinch →
- InfobipDelight.ai vs Infobip →
- MavenoidAI support agent for companiesDelight.ai vs Mavenoid →
- TwilioDelight.ai vs Twilio →
- Bland AIEnterprise voice AI platform for high-volume and regulated phone workflows
- ASAPP