Designing a banking experience where conversation became the interface.

A first step toward AI-powered banking.

The bank was already an IBM client and was exploring a broader vision for the future of its digital experience: using artificial intelligence to support customers across multiple channels.

Maya began as the first step toward that vision.

The initial product focused on a mobile conversational experience where customers could interact with an AI-powered assistant to get information and complete selected banking tasks. Starting with a single channel also created an opportunity to learn how customers interacted with the technology and build a foundation that could potentially expand to other banking channels over time.

Start with a conversation. Learn. Then expand.

The long-term vision went beyond creating a single chatbot. The mobile experience was intended as an initial environment for exploring how AI could support banking interactions, understanding how users communicated with the system, and learning how the technology could evolve before expanding the experience to additional channels.

Maya began as the first step toward a broader AI-powered banking vision.

From answering questions to taking action.

Maya was designed to evolve in two phases.

The first focused on helping customers understand banking concepts and products.

The second introduced selected transactional capabilities, moving the experience from informational conversations toward actions involving the customer's own financial accounts.

The shift from answering questions to supporting transactions fundamentally changed the design challenge.

Designing the experience between people and AI.

I worked as a UX Designer as part of a multidisciplinary team of four designers.

We collaborated with business analysts, developers, and IBM Watson AI specialists to connect user needs, banking requirements, conversational flows, and the capabilities of the underlying technology.

My role was not to build the AI model, but to understand the technology well enough to design meaningful interactions around what it could and could not do.

Understanding the problem before designing the conversation.

Before defining how Maya should interact, the team explored users, their banking needs, and the moments where a conversational experience could provide meaningful value.

Because conversational AI was still a relatively unfamiliar design space, the work required exploration beyond traditional interface design.

Rather than asking where AI could be added, the goal was to identify situations where conversation could provide a simpler path to information or action.

Learning enough about the technology to design with it.

Designing Maya required close collaboration between the UX team and IBM Watson specialists.

As designers, we needed to understand how the technology interpreted language and how different ways of expressing the same goal could affect the experience.

We were not building the underlying AI model, but understanding its capabilities helped us design conversations that were realistic, useful, and aligned with what the system could support.

A task is not automatically a conversation.

Traditional banking flows provided a starting point, but moving those tasks into a chat required a different way of thinking.

A user could begin with incomplete information, provide details in an unexpected order, ask for clarification, or express the same request in multiple ways.

Each conversational flow needed to consider what Maya already knew, what information was missing, and how to guide the user forward without making the interaction feel unnecessarily rigid.

Translating a traditional banking task into a conversation that could adapt to the information available at each step.

Personality becomes real through interaction.

Tone and personality were not intended to exist only as abstract principles. They needed to be reflected in the actual moments users experienced throughout the conversation.

The experience needed to remain consistent across information, action, and uncertainty.

Keeping the conversation at the center.

Maya was designed primarily as a conversational experience, but text was not always the most efficient way to communicate information.

Cards and visual elements were introduced selectively when they could help users understand information or take the next step. The goal was to use enough interface to support the conversation without turning Maya into a traditional banking dashboard.

The goal was not to eliminate UI. It was to use it only when it improved the conversation.

A conversation can feel casual. Banking cannot.

The challenge changed as Maya moved from informational conversations toward financial transactions.

Actions involving accounts and money required users to clearly understand what was happening before a transaction was completed.

Conversational language needed to remain approachable, but the experience also needed to communicate important information clearly and avoid ambiguity around sensitive actions.

1. Step 01

   The user requests a transfer in their own words.

2. Step 02

   Maya presents the transaction details clearly before anything happens.

3. Step 03

   The user explicitly confirms before the transaction proceeds.

My role was not to define the bank's security architecture. I focused on how banking and security requirements shaped the interaction and how sensitive actions could be communicated clearly through the experience.

What Maya could help customers do.

Maya brought together informational and transactional interactions in a mobile conversational experience.

Designing the behavior before polishing the interface.

Because the complexity of Maya lived primarily in the interaction rather than the visual interface, early wireframes helped explore how conversations, supporting UI, and user actions could work together before refining the final experience.

Exploring a new way to interact with banking services.