Contact centers are actively deploying artificial intelligence in routing, summarization, agent assist and real-time transcription, but one step in the customer journey has remained stubbornly difficult to modernize: the payment.
For many organizations, taking a payment still means a channel switch, a script change, or a transfer, which interrupts the flow and introduces risk. And as agentic AI begins to move from proof of concept into production, that gap is becoming more significant.
SequenceShift has built a direct response to this problem. Paytext, the company's new link management solution, is designed to bring payment capability into the agentic era, without breaking compliance or the customer experience.
Related Articles:
- Building a Compliant Phone-Payment Flow in Amazon Connect with SequenceShift
- The “Day 2” Migration Problem: Why Legacy Compliance Fails on Amazon Connect
- Insurance CX Is Losing Customers at the Worst Possible Moment: Payment
What Is Paytext and How Does It Work?
In an interview with CX Today, Dan Bloy, Regional Director at SequenceShift, described Paytext as a link management platform built for a world where multiple actors, not just human agents, need to initiate payment journeys.
The solution, which runs in Amazon Connect Customer, "enables different actors to collect payments using digital channels. The different actors could be a large language model, a scripted IVR journey, a human agent, or a completely autonomous API-driven end-to-end system.”
“All of those actors can create a link and then use one of the customer's existing channels to send that link to them, ensuring that they maintain their brand identity and then get real-time status of the link progression."
That is key, as most payment link systems are closed ecosystems, Bloy noted. A link is sent, and the business waits, with no visibility into whether the customer opened it, where they dropped off, or what went wrong.
“What you end up getting is customers not opening the link, or customers opening the link and then getting stuck.” Paytext changes that.
“We solve that problem by providing real-time status to both the human agent and also the autonomous systems."
An agent can send the customer a payment link and support them in real-time, with a clear view of when the customer has opened the linked and completed their payment information, helping them if required. The agent can see the result in real time and continue with the interaction.
The SequenceShift platform provides out-of-the-box integration with payment gateways, support for payments and tokenization with a self-service setup and consumption-based pricing.
Agentic AI Meets Deterministic Payments
The challenge with agentic AI in customer service is that it is best suited to open, conversational territory, where it can reason and adapt. But payment capture is not open, as it requires specific inputs, in a specific order, with full compliance.
Paytext is designed to bridge these two modes, Bloy explained:
"You can go from the human-driven payment journey to a self-service journey using typical scripted or deterministic AI. We have also designed Paytext so that you can expose its capabilities to a large language model using MCP servers, exposing the functions to the large language model and giving it the capabilities to be able to create those links and see the real-time status of those links."
As contact centers build out agentic journeys, the point of payment is where the experience most often collapses. Customers who have been handled by a smooth, natural AI conversation suddenly find themselves transferred to a different system or channel. Bloy noted:
"As customers start to lean into more agentic customer experience journeys, having a payment solution that does not support that means they are going to have a disconnected customer experience.”
“They will try and solve everything using the agentic AI, and then at the point of payment they are going to have to move into a different modality, a different channel, and the customer is going to be asking what is happening here. We have gone from this natural agentic reasoning capability to a different experience."
Context Is Not Lost at Handoff
One of the practical concerns with any multimodal payment journey is continuity. If an autonomous agent initiates a payment link and the customer then calls in for help, what does the human agent know?