Improving a Chatbot Cloud Engineers Actually Trust
Improving an internal ING chatbot by redesigning its conversation flow and experience to help cloud engineers find answers faster and reduce reliance on manual support.
In-house
Services
UX Design
Conversational Design
User Research
Usability Testing
Industry
FinTech / Banking
Cloud Infrastructure
Project Overview
The project focused on developing an MVP of a Google Dialogflow chatbot for internal infrastructure support at ING. The department’s vision was enablement: helping engineers work independently by providing quick access to support services such as raising incidents, connecting with a live agent, or finding knowledge base information.
The team was developing the IPC (ING Private Cloud) Virtual Assistant, a chatbot integrated into the IPC portal. However, the existing MVP lacked clarity and usability. Engineers struggled to understand what the chatbot could do, how to interact with it, and how to complete tasks such as raising incidents. As a result, many users avoided the chatbot and preferred contacting human experts instead.
Problem Statement
The product existed, but the chatbot was at an early stage, and the content flow needed improvement.
Users generally face issues with chatbots. They are unclear about what is needed to raise an incident, leading to incomplete details and delays. The chatbot's limited guidance and lack of visual cues leave users stuck. Many find it impersonal and unhelpful, preferring human interaction for complex issues.
Research Process
Understood how employees perceive chatbots by conducting semi-structured user interviews using the Love Letter & Breakup Letter method.
Identified usability problems in the chatbot and touchpoint portal by performing a heuristic evaluation and UX review.
Learned how internal systems and services work by performing desk research on ING infrastructure, cloud services, and chatbot technologies.
Defined the chatbot’s role and user needs by creating an artefact persona for the chatbot and writing user stories based on insights.
Reviewed large-scale chatbot feedback data from previous chatbot initiatives across departments.
View More
Key Insights
Low trust in chatbots
Fast and accurate answers
Non-intrusive interaction
Clear guidance for the chatbot
Design Process
Several directions were explored, focusing on how the assistant could be a better support for engineers. Choosing the right direction was important because the chatbot had to fit the existing ecosystem while still improving the user experience.
View More
Directions Explored
Through an "Out-of-the-bot" workshop, utilising Future and Dark Side methods was explored:
A "Non-Ad Assistant" that stays hidden until needed.
"INGPT" featuring humanised conversational AI
A "Chat Saver" that remembers conversation history.
These explorations led to the “Name the Bot” team session, where Cleo (Cloud Leo) was defined as the chatbot’s identity for the IPC Virtual Assistant.
Constraints and Tensions
Solution
Designed Cleo (Cloud Leo), a virtual assistant integrated into the IPC portal to help engineers find information, understand system guidance, and raise incidents more efficiently. The solution improves the clarity of the conversation flow, focusing not only on usability but also on desirability.
Cleo you can trust
Introduced an illustration combining the ING lion with a cloud to create a recognisable identity for the assistant. This helped position the chatbot as a supportive tool within the IPC environment while keeping the interface non-intrusive and aligned with ING standards.
Usable and guided interaction
Improved the chat experience by allowing both button selection and free text input, supporting typo recognition and going back to previous content. Content was simplified into short answers, bullet points, and linked actions to help users move faster between guidance and ticket creation.
Supporting onboarding and understanding
Created a dedicated Tips page explaining how to use the chatbot and what users can expect from it. Added summarised content, clear navigation options, and direct links to relevant knowledge base pages to reduce confusion and improve adoption.
Impact
Testing and internal feedback showed that the redesigned chatbot made the interaction clearer and more usable for employees.
Users preferred direct conversation over browsing
Users perceived Cleo as friendly and supportive
Interaction with buttons improved task clarity
Clear UI aligned with ING design standards
Auto-filled ticket flow simplified incident creation
My mentor highlighted strong UX thinking, proactive collaboration, and clear contribution to improving the usability and structure of the chatbot experience.