Book, move or cancel
Connects to your schedule system
In development. Early access is open.
Agentic CX for Amazon Connect
Describe the experience you want your customers to have. SelfCX builds it on Amazon Connect with AI agents, puts it through hundreds of test conversations, fixes what fails and waits for your approval.
Then it keeps improving: it finds where customers drop off, drafts the fix and waits for your OK.
Describe, review, test, send live, watch, improve. Repeat.
A manager describes what callers need. SelfCX does the building and the testing. A person makes the final call.
Concept demo. The clinic, the calls and every number are fictional. SelfCX is in development.
The harness
The harness is the core of SelfCX. Every version faces each layer below before anyone is asked to approve it. When a call fails, SelfCX finds the cause, proposes a fix, runs the tests again and shows you the proof.
Calling the practice version
Simulated test run, for illustration.
Dead ends, unreachable steps and missing hours are caught before a single call.
The same design always produces the same flow, so a release holds no surprises.
Scripted callers walk every path, including the ones nobody thinks to try.
Systems go down mid-call. The experience has to cope and the caller has to hear something sensible.
Callers interrupt, change their minds, go off topic and try to talk it into things.
Test calls run in Amazon Connect, including real audio through the real phone path.
Common use cases come ready to adapt. A template joins the library only after it passes the harness. Change the wording, hours and routing as much as you like, and the integrations stay connected.
Connects to your schedule system
Connects to your order system
Connects to your billing system
Connects to your on-call list
Connects to your ticket system
Connects to your call center
Library preview. Templates are being written now.
Design advisor
While you describe an experience, Design advisor suggests flows from the template library that cut friction, move more customers to self-service and improve the numbers you track. Each suggestion says what it should change. You choose what goes in.
Spots extra questions, long menus and dead ends in your draft, and offers a shorter path that has already passed the harness.
Suggests handling the requests that usually end up with an agent, like order status, balance questions or rescheduling.
Shows which numbers each suggestion is meant to move, such as abandonment, self-service rate and time to book, so you pick on purpose.
Design advisor suggests only templates that have passed the harness.
After go-live
Data-driven insights show where customers drop off, abandon or get stuck, and what each problem costs. SelfCX then builds the fix, tests it with test conversations and waits for your OK before anything changes.
Insights show where customers drop off, abandon or get stuck, step by step, and how many conversations each problem touches.
SelfCX spots reasons people still reach an agent that self-service could handle, then drafts the change.
The draft runs through the same test conversations as any release, with the expected effect on your numbers beside it.
You approve, it goes live, and the next round of data shows whether it worked. Nothing changes without your OK.
Preview. The clinic, the calls and every number are fictional.
Every fix runs through the same test conversations as a new release, and nothing goes live until you approve it.
Preview. Simulated calls, for illustration.
Early access is opening to a small group of contact center teams on Amazon Connect. Tell us about the experiences you want to build.
aidan@selfcx.com