Amazon Connect has grown well beyond cloud telephony, and Agentic CX Designer (ACXD) is one of the clearest signals of where it's heading. It's a design environment for building conversational AI applications — the kind that don't just answer a question, but guide a customer through a whole task and take action on their behalf.
Because it's new (introduced as a preview in 2026), there's very little independent explanation of it out there — mostly just AWS's own documentation. This guide fills that gap: what ACXD is, how it works with the rest of Amazon Connect, how it differs from the tools you already know, and when you'd actually use it.
Preview note: ACXD was introduced as a preview. Preview features evolve, so specific capabilities, availability, and pricing may change. This guide is grounded in AWS's current documentation and cites it throughout — but always confirm the latest details in the official AWS docs before building for production.
What Is Agentic CX Designer, in One Sentence?
Agentic CX Designer is a service within Amazon Connect Customer that gives teams a single workspace to build AI-powered conversational applications — experiences that can guide structured workflows, answer questions from trusted knowledge, connect to APIs, reason through multi-step tasks, and escalate to human agents. Content was rephrased for compliance with licensing restrictions. [Source: AWS ACXD docs]
One useful detail from AWS's own framing: you might instinctively call these things "bots," but ACXD deliberately calls them conversational applications, because they do far more than respond to simple messages. That naming shift captures the whole point — this is agentic, task-completing AI, not a scripted FAQ bot.
What Can You Build With ACXD?
ACXD is aimed at the common, high-volume self-service use cases — the ones every contact centre wants to automate well. Per AWS, typical applications include:
- Customer service automation
- Appointment scheduling
- Order lookup and updates
- Account support
- Knowledge base Q&A
- Troubleshooting flows
- Agentic task completion
- Human handoff and escalation
The key is that a single application can combine deterministic workflow logic with generative AI, integrations, knowledge bases, and guardrails. Content was rephrased for compliance with licensing restrictions. [Source: AWS ACXD docs] That blend — structured enough for business control, flexible enough for natural AI conversation — is what makes it different from both rigid IVR and unconstrained chatbots.
How ACXD Fits With Amazon Connect Customer Flows
This is the part that confuses newcomers, so let's make it clear. ACXD doesn't replace Amazon Connect's flows — it works alongside them, each with a distinct job.
| Connect Customer flows | Agentic CX Designer |
|---|---|
| Manage the customer entry point | Design and run the conversational AI experience |
| Handle contact flows, routing, queues, channel setup | Handle applications, flows, nodes, prompts, integrations, knowledge bases, guardrails, analytics |
| Route the customer into the selected ACXD application | Execute the application logic and return the next response or action |
| Support escalation into the contact centre | Decide when and how the conversation should escalate |
In a typical setup, the customer enters through a Connect Customer flow, which uses an Agentic CX block to invoke the chosen ACXD application and deployed environment. ACXD then runs the conversation — routing, responses, API calls, knowledge retrieval, generative behaviour, guardrails, and monitoring. Content was rephrased for compliance with licensing restrictions. [Source: AWS ACXD docs]
The mental model: Connect flows are the front door and hallway — they greet the customer and direct them. ACXD is the expert in the room they're shown into — it actually holds the conversation and gets the task done.
The Building Blocks: Workspaces, Applications, Flows, and Nodes
ACXD organises everything into a clear hierarchy. Understanding these four terms is most of understanding the product:
- Workspace — the top-level container. Administrators create workspaces in the Admin Hub and assign users to them. Think of it as your team's project space.
- Application — the "control centre" for a conversational experience your users interact with. This is what AWS says you might think of as your "bot." You attach flows, default behaviour, and guardrails to it.
- Flow — a set of logic, messages, nodes, variables, and routing for a specific intent or process. One flow might book a room; another answers policy questions, authenticates a user, or routes to a human. Flows can be reused across applications in the same workspace.
- Node — a single action within a flow. Each node does one thing: send a message, collect a user response, call a data request, generate AI output, route to another flow, transform data, escalate to a human agent, or end the session. Content was rephrased for compliance with licensing restrictions. [Source: AWS ACXD nodes docs]
You build flows visually on a Canvas, dragging and connecting nodes to map out each step. Supporting pieces include data requests (to fetch real-time info or trigger actions in other systems — e.g. check order status or send a confirmation), knowledge bases (for grounded Q&A), slots and variables (to capture and carry data), and guardrails (to keep the AI within safe, on-policy bounds).
Build, Test, Deploy, Observe — the Full Lifecycle
One of ACXD's genuine strengths is that it brings the whole conversational AI lifecycle into a single workspace, rather than stitching together separate tools:
- Build — create applications, flows, nodes, prompts, slots, variables, integrations, knowledge bases, and guardrails.
- Test — validate routing, flow behaviour, variables, state, and tools before you ship.
- Deploy — package changes into a build (a deployable snapshot) and deploy it to the right environment (development or production).
- Observe & improve — monitor live behaviour and iterate.
If you want to automate that build-and-deploy cycle with proper version control and CI/CD — which matters once you're running real production applications — that's exactly what the ACXD SDK enables. We cover it in depth in Amazon Connect Agentic CX Design: How to Integrate with a CI/CD Pipeline.
How Is ACXD Different From What You Already Know?
ACXD vs. traditional Amazon Connect contact flows
Classic Connect contact flows are deterministic: you script every branch and the flow does exactly what you told it. ACXD applications blend that deterministic control with generative AI that can understand natural language and reason through tasks. Contact flows still handle the entry point and routing; ACXD handles the intelligent conversation. They're complementary, not competing.
ACXD vs. Amazon Lex
Amazon Lex is an intent-and-slot conversational bot service — you define intents, sample utterances, and slots. ACXD sits at a higher level: it's a full application design environment where generative AI, knowledge bases, integrations, data requests, and guardrails come together visually on a canvas, aimed at agentic, task-completing experiences rather than intent classification alone. For many new builds, ACXD is the more modern starting point — though Lex still has its place, and the two can coexist.
Coming soon on this blog: dedicated deep-dives on ACXD vs traditional contact flows and ACXD vs Amazon Lex — the two decisions teams wrestle with most when adopting it. This guide is the overview; those will be the detailed comparisons.
When Should You Use ACXD?
ACXD is a strong fit when:
- You're building self-service that completes tasks, not just answers questions
- You want natural-language understanding rather than rigid press-1 menus
- You need to combine structured workflow with generative AI, knowledge, and integrations in one place
- You want a single workspace for the whole build-test-deploy-monitor lifecycle
- You're already invested in Amazon Connect and want a native, first-party path to agentic experiences
It's less relevant if your needs are genuinely simple (a basic menu and routing may be all you require) or if you're not on Amazon Connect. And because it's in preview, weigh production-readiness carefully for mission-critical, regulated use cases — and pair any agentic deployment with a proper AI risk assessment.
Frequently Asked Questions
What is Amazon Connect Agentic CX Designer (ACXD)?
Agentic CX Designer (ACXD) is a service within Amazon Connect Customer for building AI-powered conversational applications. It gives teams a single no-code workspace to design experiences that can guide structured workflows, answer questions from trusted knowledge, call APIs, reason through multi-step tasks, and escalate to human agents.
How is ACXD different from Amazon Connect contact flows?
Connect Customer flows manage the customer entry point — routing, queues, and channel setup. ACXD provides the conversational AI application layer that runs the experience — applications, flows, nodes, prompts, integrations, knowledge bases, and guardrails. A Connect flow invokes an ACXD application via an Agentic CX block.
Is ACXD the same as Amazon Lex?
No. Amazon Lex is an intent-and-slot based conversational bot service. ACXD is a newer, higher-level design environment for agentic conversational applications that combine deterministic workflow logic with generative AI, knowledge bases, integrations, and guardrails, built visually on a canvas.
Is Amazon Connect Agentic CX Designer generally available?
ACXD was introduced as a preview. Because preview features evolve, specific capabilities, availability, and pricing can change — always confirm current details in the official AWS documentation before building production experiences.
Conclusion
Agentic CX Designer is Amazon Connect's answer to the shift from scripted IVR toward agentic, task-completing self-service. It gives you one no-code workspace to build conversational applications that understand natural language, follow structured logic, tap into knowledge and APIs, stay within guardrails, and hand off to humans when needed — then test, deploy, and monitor them in the same place. Connect flows still own the front door; ACXD is the intelligent experience behind it.
Ready to go deeper? Read how to put ACXD under CI/CD with the SDK, explore the shift from deterministic to agentic design, or map out your experience visually with the free IVR Design Tool.