Sendbird new AI agent features roundup

AI agent development & production modes - May 8, 2025

Sendbird now offers clear separation between two environments for every AI agent:

• Development: A safe workspace to build, test, and adjust your entire AI experience, including knowledge bases, Actionbooks, canned responses, integrations, and more, without touching live interactions.

• Production: A stable, read-only environment for your customers. Fully protected from unexpected or unapproved changes.

This unlocks precise control over how you operate at scale:

• Shared workspace assets: Easily manage common resources such as communication style, brand safeguards, knowledge content, Actionbooks, and integrations across multiple agents.

• Tailored deployments: Independently update agents to precisely target your audience segments or business units (e.g., buyer vs. seller), ensuring only the intended agents are affected.

• Clear visibility: Changelogs and deployment histories clearly show what changed, who published, and when—perfect for compliance and operational transparency.

Move faster while staying in control

With this update, you don’t have to choose between velocity and control.

Now, every agent has a clear boundary between staging and production. Your teams can build and test freely, while your live experiences stay consistent and protected.

Here’s what this unlocks:

• Faster iteration, safer launches. Teams can test changes freely without fear of breaking production—no bottlenecks, no fire drills.

• Targeted rollouts at scale. Whether you’re managing 2 agents or 20, across regions, products, or user types, you can deploy exactly where needed—by product line, region, user segment, or platform surface—without unintended side effects.

• Build trust, internally and externally. Changelogs and deployment history bring full accountability to AI operations—so you always know what’s live, what’s changed, and who made it happen. Perfect for compliance, troubleshooting, or exec visibility.

Deployment and production mode is live now for all Sendbird AI Agent customers. Contact us to learn more.

AI agents scorecards - May 1, 2025

In every support operation, human agents are managed — agents aren’t just hired and forgotten. You coach them, measure their quality, give feedback, and help them improve. This is how great support organizations stay great. 

However, most teams treat AI support agents differently. They’re launched once and left to operate on autopilot. There is no regular oversight, conversation reviews, feedback loops, or growth. That gap creates silent risks — bad conversations, lost customers, missed opportunities. Good AI isn’t just self-correcting. Good AI is managed. Just like your best teams are.

At Sendbird, we believe building a great AI agent doesn’t end at launch. It’s about ongoing oversight, continuous learning, and operational excellence. This is why we created AI agent scorecards that give you the power to:

• Review conversations one by one, right inside the dashboard.

• Score across four key quality categories: Tone, clarity and readability, relevance, and solution effectiveness.

• Tag common issues with preset reasons for faster, structured reviews.

• Leave coaching notes to capture context, patterns, and feedback.

And it doesn’t stop at review. With enough scorecards, your feedback becomes fuel — helping AI support agents improve over time. The more you evaluate, the better your AI agents become. Your AI agents can now be assessed and coached with the same rigor as your human agents — and grow from the feedback.

To turn your AI support agents into managed, accountable team members. 
👉 Contact us. 

Activity trails - Apr 30, 2025

Every AI response to a customer shapes trust. However, teams are left guessing when the reasoning behind that response is hidden. With Sendbird’s AI agent activity trails, you get complete visibility into decisions—what context was used, which tools were triggered, and what knowledge was applied. It’s no longer a black box. Now, you can trust not just what the AI says—but how it got there.

What's new

With activity trails, now you can:

• See the full context behind every response, including goals, communication style, guidelines, and user attributes.

• Understand which knowledge and tools were used, and trace every answer back to the exact articles, Actionbooks, or APIs involved.

• Follow the decision process, step-by-step, from the first welcome message to tool activations to human handoffs, with full visibility into each step.

With clarity comes confidence

Activity trails show you the exact context, knowledge, and actions behind every AI response. You’re not guessing how it got there—you see the full decision path. That transparency means fewer surprises, faster debugging, and better control. You don’t just see what your AI says—you know why it said it. That’s the foundation for absolute confidence in every interaction. Activity trails are now live for all Sendbird AI agent customers.

👉 Contact us to learn more.

End-to-end AI agent testing - Apr 28, 2025

It’s 2 a.m., and a subscription cancellation request pings your bot. One wrong API call and the customer wakes up to a double charge. By 7 a.m., your inbox is a crime scene—chargeback alerts, angry tweets, and finance in full-stop mode. Most AI testing stops at the surface level—just checking if the bot can answer a single FAQ. But real-world support conversations aren’t one-and-done. They involve follow-up questions, tool-triggered actions, shifting context, and unexpected edge cases. 

That’s where the real risk lives—not in the first reply, but in everything that comes after. Sendbird’s end-to-end (E2E) testing changes that. Now, you can simulate the entire customer journey—every message, action, and decision—before you ship. Most importantly, no two customers are ever the same. Sendbird’s E2E Testing ensures your AI agent adapts — to every customer.

What's new

With E2E AI Agent tests, Sendbird AI now lets you:

• Test full conversations: Upload real customer < > agent dialogues to replay and test the entire back-and-forth.

• Simulate real-world flows: Trigger tool calls, hit live or mocked APIs, and mirror the complexity of real production environments.

• Catch breakdowns before customers do: Automatically detect where the AI drifts from the expected human behavior and spotlight exactly where it goes off track.

• Trace issues back to the source and improve: Every failure links you to the exact Actionbook step or knowledge base article, so you know exactly what to fix.

• Growing your customer testing library: Our E2E test isn’t a one-time check. Customers can continuously accumulate real-world scenarios, building a dynamic, ever-expanding library of critical tests over time.

• Schedule regular health checks: Run your whole test suite daily, weekly, or monthly to catch issues early and maintain AI performance over time.

Why it matters

Confidence comes from testing every message, not hoping for the best. With Sendbird AI agent E2E test, you can now:

• Launch with confidence. Know your AI agent can handle full conversations, not just scripted FAQs.

• Protect customer trust. Every message your AI sends either builds loyalty—or risks losing it.

• Move faster without compromise. Cover real-world complexity without slowing down launches.

E2E AI agent testing is live now for all Sendbird AI Agent customers. 

Want to trust your AI agents, launch with confidence, and sleep easy? 

👉 Get in touch.

AI Agent + Help Desk - Apr 2, 2025

Today's support is stitched together, not streamlined. AI tools, ticketing systems, and messaging platforms may be technically integrated but functionally isolated. Data moves between systems but not in real-time or in full context. Customer history is stored but buried across tabs and tools. Agents see the tickets but not the whole conversation that led up to them. Support workflows are automated, but only within each tool, not across them. Connected tools aren’t the same as a connected experience. We’re changing that.

Meet Sendbird Desk + Sendbird AI agent

Sendbird Desk is our first-party support software, built from the ground up with in-app messaging, ticketing, and agent tools in a single platform. Now, with AI agents tightly integrated, Desk becomes the first truly unified support suite – where automation and human agents work together in one thread, with full context and nothing lost in translation. There is no middleware, no hacks, and no context lost in translation. From first touch to final resolution, it’s one seamless experience—for agents, ops, and customers alike.

What makes Sendbird Desk + Sendbird AI agent different?

Most AI agent support tools integrate around your help desk, not into it. When an AI agent connects to third-party ticketing platforms via limited APIs, that restricts what AI can access – agent states, ticket status, full history – and limits how smoothly it can escalate or resolve issues. That means shallow context, clunky escalations, and brittle handoffs that leave your customers frustrated. With Sendbird Desk + AI Agent, everything lives in one ecosystem. That means:

• Shared channel, no context lost: AI and human agents operate within the same Sendbird channel, so there is no need to pass transcripts during escalation.

• Deeper visibility for human agents: Agents can see the entire AI-user conversation instantly, with zero lag or repeated work.

• One product, one partner: No third-party dependencies. One vendor, one contract, and one integration to maintain.

• Consistent analytics shared CSAT: Measure and optimize performance across AI and human support with shared analytics and CSAT tracking.

Lean by design and built for clarity

Legacy platforms are bloated, stacked with unused features, complex interfaces, and hidden costs. They slow teams down and drive up the cost. Sendbird Desk + AI Agent is lean by design. With messaging, ticketing, and AI in a single, unified platform, every feature is intentional, with no redundancy or excess. The result is faster onboarding, minimal maintenance, and a better agent experience. So, your team spends less time managing complexity and more time delivering results.

The future of support isn’t stitched together. It’s unified. With Sendbird AI support agents native to Sendbird Desk, you can streamline operations, elevate customer experience, and confidently scale support from a single platform. 

To learn more, 👉 Contact us.

Proactive AI agents - Mar 28, 2025

Your systems know what’s happening—long before your users complain. 

• An order is clearly delayed – your system knows it. The customer knows it, too. But they still have to dig through your help center, hit chatbot dead ends, and repeat their issue three times before someone helps. 
• A storm is about to ground flights – your system sees the alerts. But even your premium travelers must contact you first to explore their options.
• A long-time subscriber hasn’t logged in for 4 months. You do nothing until they hit “Cancel.”
• A driver’s license is about to expire. But your AI stays quiet – until the platform blocks the user.

The frustration? Avoidable. The data? Already in your hands. And yet, nothing happens. Most AI agents wait quietly until a customer raises their hand. By then, it’s too late. They’re already frustrated and ready to leave. What if support moved first? What if your AI reached out before your customer did?

What’s new

Proactive AI support via chat gives the AI agent the power to move first, reaching out before frustration builds, churn begins, and a ticket even exists. With Proactive Chat, you can now:

1. Detect key signals using real-time triggers based on user behavior, system events, or external APIs/webhooks.

2. Engage across channels: Activate proactive AI messaging on SMS, WhatsApp, and more – wherever your users are.

3. Deliver personalized, contextual messages. Use customer-specific data and historical context to tailor every message. Making it feel timely, relevant, and human.
   

Each proactive interaction is built on three core components:

1. Dynamic conditions. Define real-time triggers based on user attributes, activity, delivery status, or system events. Decide when, why, and how the AI should engage first.

2. Prior knowledge. Leverage existing context, such as VIP status, past purchases, or usage patterns, to personalize messaging and build trust instantly.

3. Channel orchestration. Choose where and how many channels to activate, transforming a single message into a coordinated multi-channel experience delivered simultaneously.

Why it matters

Reactive support puts your business on defense. Proactive AI chat flips the script, helping you intercept issues early, reduce ticket volume, and build customer trust before frustration sets in. With Proactive Chat, you can:

• Reduce inbound volume and cut operational costs

• Prevent churn by engaging high-risk users early

• Drive conversion by stepping in when the intent is high

• Turn support into a brand moment, not a damage control effort
  

Your team stays in complete control, defining triggers, copy, timing, and performance tracking in one place. Proactive AI chat is how modern support gets ahead of problems. 

It's now available on Sendbird AI agent for customer service. 

👉 Try our SMS AI agent demo or contact us to learn more.

WhatsApp AI agents - Mar 25, 2025

We’re bringing the best AI agent to the world’s most popular messaging app. Your customers already trust WhatsApp. With over 2 billion users globally, it’s not just the most popular messaging app—it’s a daily habit where conversations start and stick. Your support can live there, too, with Sendbird's AI agent for WhatsApp.

What’s new

You can now deploy your AI agent on WhatsApp to deliver fully automated, human-like customer support without customers ever leaving the app.

Why it matters

Customer expectations have changed. They want fast, effortless support – on channels they already use. WhatsApp delivers precisely that. By bringing Sendbird's support AI Agent to WhatsApp, you can meet customers where they are with real-time, human-like, and contextual conversations.

There’s no app to download, no login to remember, and no learning curve. Customers send a message like they would to a friend, and the AI agent handles the rest – from resolving issues to capturing feedback and closing the loop. It’s automation that feels intuitive, familiar, and personal - not transactional.

Behind the scenes, teams stay in complete control. Configure consent messages, set conversation-close timing, track CSAT, and monitor performance – all from one place. This is scalable support without the trade-offs: personal, fast, and easy to manage.

Modern, fast, familiar, and always-on support is now available on Sendbird's AI agent platform for customer service. 

To learn more, 👉 Contact us; we're here to help. 

AI agent Salesforce knowledge source - Mar 20, 2025

Customer service teams depend on fast, accurate information. But knowledge is often scattered – spread across multiple tools, buried in outdated uploads, or requiring constant manual updates. With Salesforce knowledge source integration, AI support agents can now pull answers directly from live Salesforce knowledge articles. No more outdated documents. No more manual syncing. Just real-time, authoritative responses from your company’s single source of truth.

What does this mean for customer service teams?

• Accurate, up-to-date responses: AI agents directly retrieve the latest policies, product details, and operating instructions from Salesforce knowledge.

• Zero manual effort: No need for teams to manually upload or update documents. Everything stays in sync automatically.

• Faster resolutions: AI agents provide consistent, context-rich answers, reducing escalations and improving customer satisfaction.

Salesforce joins the expanding list of supported knowledge sources for Sendbird's AI support agent, including Confluence, website URLs, direct file uploads, and Zendesk. It is available now on Sendbird AI agent for customer service. 

To get started, 👉 Get in touch. 

AI agent Actionbooks version control and profile - Mar 18, 2025

Keeping AI agents aligned with your business isn’t a one-and-done task but an ongoing process. As your operations evolve, so do the instructions guiding your AI. However, updating them without visibility or testing is a recipe for inconsistency. Tracking changes manually is inefficient, and deploying untested updates is unpredictable. That’s why we’re introducing version control for Actionbooks & Profile, which allows you to track, test, and refine AI behavior confidently.

What’s different?

• Track every change. Easily see what’s been updated in your AI agent’s instructions over time.

• Test before going live. Try out new instructions in a safe environment before applying them to real customer interactions.

• Revert anytime. If an update doesn’t work as expected, roll back to a previous version with one click.

• Better control. Keep a clear record of all changes to ensure AI agents follow the proper guidelines.

For teams managing AI support agents, this means more control, fewer surprises, and AI consistently delivering the right experience. 

Now available on Sendbird AI agent for customer service. 

👉 Contact us to learn more.

Specialized AI agent teams - Mar 14, 2025

A single AI agent handling every conversation — sales, support, onboarding — for every customer and every situation? That’s not how real teams work. And it’s not how your business runs. With the Sendbird AI agent team, you can build a team of specialized AI agents, with each agent having its role, purpose, inputs, and outputs. Each agent comes with its knowledge, manual, and guardrails, purpose-built for the job it’s meant to do.

Why this matters

Even if a single AI agent follows different manuals and playbooks, it still pulls from the same shared knowledge base and works under the same global rules. Even worse, the insights from every conversation get lumped together, mixing sales signals, support trends, and churning risks into one messy pool of data. That creates problems across the board:

• Data governance risks. Sensitive customer data from support cases can leak into sales recommendations, creating compliance issues and damaging customer trust.
• Regulatory and compliance challenges. Different markets, products, and conversation types need their own data handling rules — rules a single agent can’t fully enforce.
• Blurry performance tracking. You can’t optimize AI for sales, support, and retention when all the data is blended — leaving you guessing what’s working, what’s failing, and how to improve.

With an AI agent team, each AI agent stays in its lane:

• Data stays clean. Each agent has a knowledge base and conversation history, so sales, support, and retention data never overlap.
• Access stays controlled. Each agent only accesses the data, systems, and tools it needs, reducing governance risks.
• Compliance stays intact. Agents follow region or product-specific policies, supporting local regulations and industry standards.
• Performance stays clear. With clean data and role-specific insights, you can track and optimize each agent’s conversations, outcomes, and impact.

Your human teams are built with specialized roles for a reason — your AI should work the same way. With the AI agent team, every agent is purpose-built, governed, and optimized for its specific job, working seamlessly together to power your entire customer journey.

👉 Now available on Sendbird AI agent for customer service.

Email AI agent - Mar 13, 2025

We’re bringing the best AI support agent to your customers’ inboxes. Email is the backbone of customer support for many businesses—and for good reason. However, traditional email support is slow, manual, and hard to scale. Customers wait. Tickets pile up. Resolution times stretch. With Sendbird AI agent for email, your team can deliver instant, intelligent, and fully automated support—directly through your existing email channels.

What’s new

You can now deploy the Sendbird AI agent to your support email address. It reads and understands incoming messages, replies with human-like responses, and resolves issues end-to-end without human intervention.

Why it matters

• Email remains essential. It’s one of the most widely used support channels – and often the first point of contact for customers.

• Expand your reach without expanding your team. Automate high volumes of email inquiries without sacrificing quality.

• Human-like support at scale. The AI understands context and tone, delivering fast, natural responses.

• Frictionless for customers. No apps, no portals – just email. Customers write in; the AI handles the rest.
  

For customers, it’s just email. For your team, it’s automation that scales. With Sendbird's AI agent for Email, your team can deliver always-on, enterprise-grade support through the world’s most familiar channel – without adding complexity or cost. 

To learn more, 👉 contact us.

Freshwork AI agent integration - Mar 11, 2025

Your AI agent can resolve many conversations, but some situations require a human touch. A clunky handoff frustrates customers, disrupts workflows, and leaves agents scrambling for context. With Sendbird's AI agent Freshworks integration, AI agents can escalate conversations directly to Freshchat, keeping all the conversations in one place. 

Why it matters

• Faster resolutions: AI hands off conversations instantly, so human agents can step in without delay.

• No disruptions: Customers don’t need to switch platforms or repeat themselves. Conversations stay within the same chat.

• Complete context: AI-generated transcripts and summaries ensure agents have all the necessary information.

• Unified insights: Track AI resolution rates, handoff rates, and customer satisfaction - all in one place.

SMS AI agents - Mar 3, 2025

" Hey, I need Help!" That’s how most customer support conversations start—direct and straightforward. But here is the problem—most support systems aren’t built for real conversations. Instead, you get a ticket. A number in a queue. Maybe an email response - only hours later. Perhaps you get stuck in a chatbot loop, pressing buttons, hoping to talk to someone who can fix the problem. We think there’s a better way. So, we built an SMS AI agent. A fully automated AI customer service that texts like a human and solves your issue, end-to-end, over SMS. You send a message. That’s it. No apps. No waiting. There are no dead ends, just a real and fluid conversation that gets things done.

Why AI agents for SMS?

• Always accessible – The most universal support channel. No downloads. No internet is required.

• Preferred by modern customers – 82% of Millennials and 74% of Gen Z favor real-time, digital-first service.

• Feels natural – Customers reach out just like they would to humans by sending a simple text.
  

3. Set up intelligent automation. View channel-specific messages sent from the AI Agent and configure automated flows, such as conversation auto-close time and messages sent once an AI SMS conversation closes.

4. Monitor and optimize performance. Filter and analyze AI engagement on SMS, resolution rates, and CSAT scores.

With an SMS AI agent, businesses gain real-time support, actionable insights, and a frictionless way to engage with customers. Now available on the Sendbird AI agent platform for customer service. 

👉 Contact us to learn more.

Sprinklr AI agent integration - Feb 20, 2025

Support gets smarter and more seamless with the Sendbird Sprinklr AI agent integration. Now you can escalate conversations to human agents directly in Sprinklr, while also training your AI using your existing Sprinklr knowledge base.

With Sendbird AI agent’s integration with Sprinklr, you can:

• Auto-route conversations: Seamlessly transfer chats from AI to human agents in Sprinklr for smooth, real-time handoffs when a human touch is needed.

• Train AI with Sprinklr knowledge: Instantly enhance AI accuracy by syncing your Sprinklr knowledge base to deliver faster, context-aware answers to customers.

To learn more, 👉 Contact us.

Localized AI agents - Feb 17, 2025

Most AI agents claim to "support" 100+ languages. But they’re just translating the same answers across regions, like running global customer support through Google Translate. Real support isn’t just about language. It’s about context, culture, and operational know-how. Knowing what to say and how and why things work differently in each market. That’s why global enterprises don’t rely on a single, centralized team to handle customer questions in different languages. Instead, they build localized teams with deep expertise in their regions. That’s precisely how Sendbird’s localized AI agents work.

How are sendbird AI agents different?

• Independent localized AI agents, not just translations. Instead of a single AI bot translating responses, each AI agent operates independently—like a real-world regional team.
• Own knowledge base & internal data. Every AI agent pulls from region-specific sources, ensuring answers aren’t just translated but contextually correct.
• No cross-region interference. AI agents operate independently, eliminating mismatched responses across markets.

Built for enterprise operations, each AI agent can connect to different CRMs, helpdesks, and internal databases, mirroring how global businesses actually operate. For international brands, this means an AI agent team that feels as real as a local team—scalable, compliant, and truly built for the way enterprises operate.

How it works

1. Create up to 5 independent AI agents per language, region, or teams.

2. Set AI agent’s language. This determines the language used initially by the AI Agent to welcome the user.

3. Configure the welcome messages, suggested replies depending on the language set.

4. Manage AI agent’s playbooks, knowledge, settings, and data individually, making sure AI agents follow language or region-specific policies

To get started, 👉 Contact us.

Confluence AI Agent knowledge source - Apr 2, 2025

Your teams already capture critical processes, product details, and internal knowledge in Confluence. Now, your AI agents can tap directly into that content — staying up-to-date without manual uploads or syncing.
With Confluence connected, your AI agents:

• Pull answers directly from live Confluence pages, no manual uploads.
• Stay up-to-date with your latest policies, product info, and internal processes.
• Handle complex questions by tapping into detailed internal documentation.

Confluence is now part of our growing list of integrated supported knowledge sources, including website URLs, direct file uploads, text snippets, Salesforce, Sprinklr, and Zendesk knowledge bases. It is also now available on Sendbird AI agents for customer service. 

To learn more, 👉 Contact us.