Sketchin

The virtual assistant that improves customer service through AI

We supported Nexi’s Design Studio in designing a virtual assistant powered by generative artificial intelligence. The goal is to address Nexi users’ issues in quickly and efficiently.

Nexi Chatbot interface on tablet and smartphone.

We supported Nexi’s Design Studio in designing a virtual assistant powered by generative artificial intelligence. The goal is to address Nexi users’ issues in quickly and efficiently.

Industry

Fintech & Insurtech

Year

2024

Client

Nexi

Services

An analytical approach to choosing where to apply AI.

Developing an AI-powered chatbot is just the endpoint of a broader design process where, alongside Nexi’s Design Studio, we analyzed different business areas that could benefit from the impact of generative AI on end users. After analyzing KPI and ROI, one of the most interesting areas turned out to be user support.

It was then decided to enhance the user experience by shifting from a generalist support system to a tailor-made one that would address the specific needs of users.

Demonstration of Nexi Chatbot on a laptop.

The virtual assistant project had different project deliverables, very different from each other but essential in creating a coherent and credible experience for the user. Starting with a Tone of Voice appropriate to the Nexi context, followed by a UI Library and Experience Guidelines, all elements were designed to work in sync.

Average monthly conversations handled by the virtual assistant
35k
Average monthly intents generated during conversations with users
63k
Intents handled by generative AI
35%

A chatbot that speaks the same language as the customers

The voice of a chatbot defines its personality, while the tone sets its mood towards users. The use of an implicit first-person singular made the bot friendly yet not overly formal.

To create a detailed experience guideline, we also define how the chatbot should interact with users: message length, use of formal "you", the values it expresses, the use of genderless pronouns, and, of course, a UI that allows it to be easily recognizable and user-friendly.

Composition of various Nexi Chatbot UI elements.

Managing the main contact reasons. Done.

Nexi has a set of known contact reasons that account for a large portion of current conversations. To address these queries, deterministic flows have been defined to support users in achieving their set goals.

Once the artificial intelligence maps the intent to the correct flow, it directs users into guided conversations where they can find the solution to their problem.

Demonstration of Nexi Chatbot on a smartphone.

The largest AI project managed with CoPilot in Italy in 2024.

The largest Microsoft AI-based project implemented in Italy in 2024 to provide a completely new user experience used CoPilot as the core engine for managing the AI.

Implemented as a technology-as-a-service, CoPilot was used both to map user requests to predefined intents and to handle requests that do not have a preconfigured flow. For these, Microsoft's AI can scan the entire Nexi knowledge base to find the best response, which is then formulated in a conversational style.

Demonstration of Nexi Chatbot on a smartphone.
Deterministic flow charts.

Intensive testing to leave nothing to chance.

Unlike a traditional interface project, a virtual assistant is much more complex to test and validate.
It's not enough to simply check if it aligns with the design; it needs to be stressed in use, and the responses must be analyzed to ensure they are in line with the project goals.

The testing focused on:

  • Adherence of the UI interface to the designed patterns.
  • The textual response of the assistant being consistent with the intended tone of voice, both in form and manner.
  • Correctness of deterministic flows, for which hundreds of user stories were created to allow the internal testing team to validate their consistency.
Demonstration of a payment with support from Nexi Chatbot.

Nexi's virtual assistant is not only for assisting users but also for operators, who can use it to close user cases in Nexi's systems. It includes specific phrases in the console for human handovers and different operator chat queues based on the reason for contact through the chatbot.

Project numbers

weekly iterations, from research to design
45
Team members, including designers and AI specialists
5
Stakeholders involved in the project
35+
Deterministic flows created for the main contact reasons
107
Trigger phrases analyzed and cataloged to map to specific intents
348
Project releases to cover the entire Nexi customer base
Three

Get in touch

Learn more about our projects and what we can do for your business.