Designing Conversational Agents for Payment Collection

About —

I am a conversation designer at Interactions, a conversational AI company that provides Intelligent Virtual Assistants that enable businesses and consumers to engage in productive conversations across various industries, including electronics (Samsung), telecom (AT&T), food services (Chipotle), financial services (Citi), and hospitality (Hyatt) to name a few.

My Role —

I design critical voice experiences for four of my payment collections projects that are expected to take 1.4 million inbound calls monthly. I work closely with developers to create complex conversation flows. In addition, I identify and track business and performance metrics by implementing reporting tags in design.


Interactions LLC


Conversation Designer


2019 - Current

The Value —

Creating an engaging customer experience to increase Payments and Debt Recovery

With recent advances in Conversational AI, the payment collections industry has been transformed to be more automated. This has resulted in the ability to increase revenue recovery at a reduced operational cost.

If you are not familiar with automated debt collection services, here is a video demo showing how our intelligent Virtual Collection Agent (VCA) interacts with the caller. The VCA helps collection agencies lower agent costs and also allows customers to feel more comfortable due to the conversational and "judgment-free" experience.

A comprehensive calling experience starts with identifying and verifying the right party, followed by playing the legal disclaimer (Caller ID & Auth). The following steps are asking preferred payment methods and collecting payment details (Payment Collection). The last steps are offering payment arrangement options and subsequently processing payments (Payment Negotiation).

Design Process —

01. Understanding the Clients and their Customers

Who am I designing for and what is their primary goal are questions I ask myself before designing any conversation. In this case, the collection agencies are our client, but the key players are both the agencies and their customers, the callers. The top priority of collection agencies is to get the payment, and the primary goal of callers is to resolve the debt. That being said, the conversational experience should be simple and direct. Below are design examples demonstrating how I make iterations with user goals in mind:


Our records indicate that you have a balance of $300. You can make a payment in full, or you can settle your account today if you pay a one-time payment of $240. Which option would you like, full payment or settlement?


Okay. How much would you like to pay today?

Refined V2

Our records indicate that you have a balance of $300. You can settle your account today if you pay a one-time payment of $240. How much would you like to pay today?

*Amount Provided

This gives the caller the negative impression that s/he has to make a full payment, which is likely to result in hanging up.

This provides the same information, but reduces the conversational turns and gives the caller a clear prompt with flexible options if s/he is not paying the full amount or a settlement.

Design Process —

02. Designing How You'd Want to Hear it

Before diving into the design tool and building out the flows directly, I always start by drafting the call flow and writing sample dialogs. It helps me to think about the structure and logic of my design. Moreover, writing sample dialogs is an effective way to discover problems so that I can discuss and get clarifications from clients at an early stage.

Sample Dialog: Ask Payment Amount

How can I help you today?

I want to make a payment.

You have a balance due of $25. How much would you like to pay today?


Thank you.

After writing this sample dialog out, I immediately realized there are details missing: do we have a minimum amount accepted? If we do, will this value be passed from the API? A new conversation path is needed to handle cases when the user provided an amount that is lower than the minimum.

Collect Payment Details - ACH

Please tell me your bank’s routing number.


Sorry about that, I didn’t hear anything. What is your bank’s routing number?

Here is a better way:

Please tell me your bank’s routing number. If you need some time to look it up, just tell me. 

Collect Payment Method - Credit Card

Are you going to pay this with a credit card or by checking account?

Credit Card

Any credit card information collected from you here is secure and not accessible to anyone. Please
say or enter your 16 digit card number.

Here are more examples demonstrating how I iterate my design by taking a step ahead and considering a variety of scenarios.

Be aware of the various environments users could be at. No one will read their card number out loud in public.

The Challenge —

01. Simple on the Surface, but Conversational Flow Runs Deep

When we gather requirements from clients, the feature requests are always high level. Take Payment Collection as an example. Ideally, the dialog flow is simple and straight-forward as follows: read the balance, ask for the payment method, collect the payment details, confirm and process the payment. However, when it comes to design, questions burst out. Are we taking credit cards, debit cards, or checking accounts? Is there a processing fee? If the caller refuses to pay, should we offer a settlement? How many settlement options should we provide?

All of the questions above need to be considered in the flow. Understanding the business rules and presenting all the necessary information to the caller without overloading them is the main design focus.

The Challenge —

02. Incorporating Business Analytics into Design

One of the most common priorities across collection clients is to successfully take the payment. As a designer, I need to pay attention to business cases and success metrics throughout the design and build process.

Another big part of my job is to implement reporting tags in the design so we can track data like transaction success rate, transfer reason, abandon rate, etc. These data are also useful to drive design decision making in the future.

The Challenge —

03. Frequent Requirement Changes

Business needs are more dynamic than ever. One of the biggest challenges in my current workflow is that there are so many moving pieces at the same time. For instance, when iteration 1 is under development, iteration 2 might be just signed off and Iteration 3 is at the design phase. One benefit of this is that I now have extensive experience collaborating with developers and QA to build and make changes to the designs.

This agile working environment fosters my communication skills, both those required for communicating within the company and with external customers. Moreover, learning how to create and manage multiple design versions is another value skill I gained to deal with requirement changes when designing conversational experiences in a professional setting.

The Challenge —

04. Design with Data Protection in Mind

Although producing conversation flows might seem to be the dominant job responsibility of a conversation designer, knowledge related to data retention and data security is also required, especially when working with sensitive data such as social security numbers and payment card data. During the design process, I always familiarize myself with all the data protection policies and mark all the sensitive data so it will be encrypted, not recorded. Additionally, understanding what data can be stored helps facilitate the conversation when working with clients and developers to define API work.

Conclusion —

Working on industry projects has boosted my proficiency in designing with adaptive thinking as well as my proficiency in refining requirements from ambiguous to concretely defined. All of my projects are currently under development and awaiting go live. Although I can’t disclose the details of my projects, I am happy to chat about my learnings if you want to know more!