Business user is the new architect of customer experience. No code required

Picture a summer thunderstorm in July that grounds flights across a major hub. Hundreds of travelers are stranded overnight. At most hotels, the calls start coming in. People wait on hold to extend a stay, move a checkout, or rebook a room they can no longer reach in time.

Lily doesn’t call. The hotel texts her first.

Her airline just flagged the cancellation, and the hotel already knows as it had her flight details: Lily is a platinum-tier guest with a 6 AM departure that no longer exists. Within seconds she gets a message. “Hi Lily, we see your flight has been cancelled. We’ve held your room for another night, same suite, no action needed unless you would like to change. Since you’re with us a little longer, here’s a complimentary 30-minute spa session tomorrow morning with your usual massage therapist. Tap here to talk to an agent and pick a time that works.”

Lily taps, and a call connects instantly. As the voice walks her through available slots, the times appear on her screen in real time. She says “10 AM works,” taps to confirm, and her updated checkout, new booking confirmation, and spa appointment land on her phone before she has even closed the airline cancelation notification. The conversation and the screen moved together.

No hold, no menu tree, and no explaining her situation at a time of distress. The hotel knew who she was, what happened, what she would want, and it handled everything before she had to ask.

Here’s the part that matters: nobody on the IT team built that interaction. A Customer Experience (CX) operations lead from the business team designed it on a Tuesday afternoon, tested it, and shipped it live the same day.

However, there’s engineering underneath it all. Someone built the airline data feed integration. Someone connected the loyalty system. Someone wired up room management and spa booking APIs. That infrastructure is real, and it’s hard. What’s different is that once that foundation exists, the experience itself (the logic of “storm detected, guest stranded, extend room, offer perk, confirm”) was designed by someone who understands the customer, not someone who understands a coding language.

This distinction is what the rest of this post is about.

The constraint that defined customer experience for thirty years

Every CX leader knows what customer obsession is supposed to look like: treating every interaction as if it were the only one you had to handle right now. Know who the person is, what just happened to them, what they need next. Handle it before they have to ask.

Nobody disagreed with the principle. The problem was the math.

Nearly every customer experience operation runs on the same constraint: the queue.

How many customers can you serve? How well? How fast?

When Andrei Papancea built the conversational AI system at a leading bank, he was the bottleneck. The business team knew exactly what to build. But they had to file a ticket, wait for engineering to prioritize it, and hope the team interpreted their intent correctly. That’s not an AI problem. That’s an organizational problem. And it’s the same constraint that has defined customer experience for thirty years.

CX leaders right now are being asked the same question by their CEOs: what are you doing with AI? Most have an answer. Fewer have a strategy that puts the right people in control of delivering it.

“Obsess over every customer and every interaction as if it were the only one you had to handle right now. For the first time, that’s no longer an aspiration, it’s possible.”

Agentic AI removes the ceiling on how many customers you can serve and how well you can serve each one. In production, at enterprise scale, across every channel. But the unlock isn’t that the AI got smarter. It’s that the people closest to the customer can now design and ship the experience themselves, without waiting in a queue of their own.

This shift is worth paying attention to.

What agentic AI changes and three things enterprises have to get right

Let’s be honest about where we are. Agentic AI doesn’t magically erase every constraint overnight. Models still hallucinate. Integrations still require engineering. Governance still requires rigor. If someone tells you otherwise, they’re selling you something.

What agentic AI does change is the ceiling. For the first time, the number of customers you can serve well simultaneously isn’t bound by the number of humans you can hire, train, and schedule. The aspiration to treat every interaction as if it were the only one you had to handle, that’s now architecturally possible. You can now obsess over every customer like they’re your only customer.

But most enterprises chasing this are still building for the queue underneath the surface. Call it queue thinking: How do we manage the volume? How do we staff for the peak? Those are legitimate operational questions, but the wrong frame for what’s now achievable.

Agentic thinking starts from a different premise: the ceiling doesn’t have to exist.

1. Serve the entire journey, not the parts you’ve tooled for

Most enterprise CX stacks weren’t built. They were assembled, one vendor at a time, over years. Voice lives in one system, chat in another, digital somewhere else, agent assistance in a fourth. They don’t share context, they don’t share data; and they definitely don’t share a view of the customer. Putting agentic AI on top of that fragmentation doesn’t remove the queue. It just gives you a faster queue.

Genuine agentic self-service requires one system that understands the full customer journey: the context before the interaction started, what the customer needs right now, and how to resolve it completely without routing through a chain of disconnected tools. That requires unified architecture, not a point solution. And it’s one of the convictions that drove how Amazon Connect Customer has built this from day one.

2. Don’t trust black boxes. Build the structure that makes agentic safe

A system that reasons and acts autonomously is genuinely risky, especially without guardrails, in regulated environments, with real customer consequences.

For thirty years, the enterprise playbook was built around one word: assisted. AI-assisted. Agent-assisted. Human-assisted. The assumption underneath every system was that AI is a support layer and a human handles the hard part.

“The goal isn’t assistance. It’s resolution and customer delight.”

Reaching that goal safely requires determinism and agentic reasoning to coexist in the same governed flow. Steps that require precision (identity verification, compliance disclosures, payment processing) are deterministic. Whereas the steps that require judgment, open-ended understanding, complex reasoning across systems, are agentic. Both in the same conversation, under the same governance layer. That’s what makes agentic AI safe enough to put in front of your customers, not just the demos.

And visibility is what makes it durable. End-to-end observability means you see exactly what the AI said, why it said it, and where the experience can improve. That feedback loop, from live production back into design, is how agentic CX gets better every day rather than drifting from intent.

3. Put the power in business users’ hands. They are heroes of your CX design

The bottleneck in customer experience has never really been AI capability. It’s been the distance between the people who understand customers and the people who control the technology that serves them. Close that distance and the whole system changes.

The goal is to make our agentic AI capabilities more powerful and more approachable for the business users who want to control and continuously optimize their customer experience without relying on developers at every turn. When that happens, every business user becomes a hero, and every end customer becomes a VIP: needs anticipated and met before issues ever arise, proactively and personally.

Here is what we’re building toward: a business user who comes in every morning, sits down alongside AI teammates, and continuously delivers. They see what changed overnight. They adjust the experience. They test it. They ship it. Before lunch. That is not a pilot. That’s the new normal for business teams that are winning in delivering delightful customer experiences. And when that person walks in the door, they’re no longer waiting for permission to be great at their job. That’s the heroic role business teams have always deserved.

A quick diagnostic

If you want to know whether your organization is building for the queue or building beyond it, ask these questions.

• How long does it take to change a customer journey once it’s live? If the answer is weeks, you’re still in queue thinking.
• Who owns the design of your self-service experiences? If the answer is engineering, the people closest to the customer aren’t in control.
• What happens when the AI hits its limit and a human takes over? If the answer is the agent starts from scratch, the architecture wasn’t built for resolution.

None of these are technology questions. They’re organizational ones. And that’s the point.

There’s a fourth question worth asking, and it’s the one most often overlooked in the excitement around agentic AI: What happens to everything you build today when the model underneath it changes in six months?

Models change. What was state-of-the-art last quarter is a commodity today. If your CX architecture is tightly coupled to the model powering it right now, every advance in AI isn’t an upgrade. It’s a rebuild. That is not innovation. That’s technical debt on a schedule.

The foundation has to be durable enough that your investment compounds rather than depreciates every time the underlying AI shifts. Every experience your team designs, every workflow they ship, every integration they wire up should still be there, still working, still improving, regardless of which model is running beneath it. The work accumulates. It doesn’t reset.

AWS offers that natively integrated structure through Amazon Bedrock. This is what separates our infrastructure from a demo.

What this looks like in practice: Launching Agentic CX designer and Live Sync

The hard part of agentic CX was never the AI. It’s everything around it: the governance, the integrations, the security posture, the compliance controls. That’s where most efforts stall. Amazon Connect Customer already handles these production workloads at enterprise scale across thousands of organizations. The agentic capabilities launching today aren’t a separate product grafted on. They’re a natural extension of what’s already running. You don’t rip out what you’ve built to get here, you upgrade.

Earlier this year, AWS acquired NLX to deepen the foundation of Amazon Connect Customer’s agentic CX capabilities. NLX spent nearly a decade building and hardening agentic conversational AI at enterprise scale and the acquisition was an accelerator to empower business users to design, simulate, test, and deploy agentic self-service in days and weeks, not months.

Launching today, the Agentic CX designer in Amazon Connect Customer is available in preview to empower business teams to build end-to-end conversational experiences visually. It blends agentic and deterministic steps without writing code. They test before deploying, and they iterate after. The full lifecycle, in one place, owned by the people closest to the customer.

For interactions that require both conversation and visual action simultaneously, Live Sync, also launching in preview today in Amazon Connect Customer, adds something no competitor offers. A patented technology that synchronizes voice with a digital on-screen interface in real time. That’s what Lily experienced in the example above. The conversation and the screen as one unified experience. True conversation isn’t just about isolated voice or text; it’s about shared context, like two people solving a problem while looking at the same document.

Saks Fifth Avenue reached production in 6 weeks, with a business analyst leading the design. It is not a function of AI sophistication. It is a function of who was empowered to design and iterate, and of an architecture durable enough to ship. Interaction quality at that speed holds: sub-1% error rate on read-backs and compliance disclosures, under two-second response time per conversational turn. In a voice interface, one wrong digit’s one lost customer. The architecture has to be fast and precise.

As Pasquale argued in Deflection is Dead. Resolution is King: the right measure isn’t containment, it’s resolution. A 90% CSAT alongside 67% automation isn’t a tradeoff. It’s proof that the queue ceased to be the constraint.

Come find uss at CCW: Hear directly from customers like United Airlines and Citizens Bank

The brands that will define customer experience over the next decade aren’t the ones that will have the most sophisticated AI. They will be the ones that gave their CX teams the power to act on what they already know. Amazon Connect Customer is built for this moment.

“The business user who understands the customer has always been the hero waiting to be unleashed. Now they have the tools.”

A business user who comes in every morning, works alongside AI teammates, sees what needs to change, and ships it before lunch. No ticket. No sprint. No waiting. That’s what heroic looks like in customer experience today. And it’s no longer a vision. It’s available now.

If you want to see this in action, come see us in person at Customer Contact Week in Las Vegas, June 22–25. Two sessions worth adding to your schedule:

• On Wednesday June 24 at 4:30 PM in the Academy Ballroom, Pasquale DeMaio, VP Amazon Connect Customer, joins Citizens Bank and United Airlines on the main stage to discuss how two of the most recognized brands in the world are rebuilding customer experience, and why they bet on Amazon Connect Customer. Moderated by Rebecca Jarvis of ABC News.
• On Thursday June 25 at 12:00 PM in Summit C, Caesars Forum, hear how United Airlines’ Customer Service and Technology teams partnered with Amazon Connect Customer to build a conversational AI and self-service strategy that drives high automation rates and push CSAT to new heights, especially during Irregular Operations (IRROPs). United became the first airline to deploy Amazon Connect Customer’s patented Live Sync technology, a bidirectional multimodal experience pairing guided voice with a real-time visual companion on customers’ mobile devices. With a goal of up to 50% voice containment on IRROPs days, learn how United designed seamless AI-to-agent handoffs with full context continuity, and how you can bring this next-gen technology to your customers. United Airlines will dive deep into how they built multimodal agentic automation at enterprise scale.

To learn more, visit the Amazon Connect Customer webpage, explore the documentation, or contact your AWS account team.

About the authors

James Huddleston is the Head of Product Marketing for Amazon Connect Customer at Amazon Web Services (AWS) based in San Francisco, California. He is focused on product positioning, go-to-market strategy, and helping customers realize value from AI-powered customer experience solutions. James is a father of two girls and enjoys spending time with his family, traveling, and going on adventures with his daughters.

Andrei Papancea is a leader on the Amazon Connect team, driving the future of agentic AI for customer experience following AWS’s acquisition of NLX, the company he founded and led as CEO. He also teaches graduate-level courses on AI Engineering and Cloud Computing at Columbia University and NYU.

Aartika Sardana Chandras is a Senior Technical Product Marketing Manager for Agentic Customer Experiences at Amazon Web Services (AWS), focused on Amazon Connect Customer. With over 15 years of experience in product marketing, she’s passionate about helping enterprises use AI to deliver delightful customer experiences and putting that power in the hands of the people closest to their customers.