Background
In preperation for the holiday season, bol wants to support their users in finding gifts for their loved ones. In comes the AI Gift Finder/Cadeauhulp. This conversational AI chatting experience, helps users with finding inspirational ideas for a gift, or a direct product that may fit their needs.
Challenge
At this point in time, DEPT® had already done concepting for the Gift Finder. Though bol felt there were still design opportunities to enhance the user experience before releasing it to the public. My role here was to seek the answers to the questions; learning about how users behave with a conversational chat, learning about the value of output, learning about how intuitive the navigation feels, and more.
Seeking answers to these questions had been done so, by setting up and running the pilot for the Gift Finder. Consisting of a mixed methodology approach with usability tests and retrospective interviews in a lab environment. Learning about what users do, think, and say regarding finding gifts with the Gift Finder.
Suggestions and chip behavior
Users struggled in getting enough variety in products per output, variety in product categories, as well as the product amount. With the limit being set to 4 products per output, a behavioral pattern occured where users often relied on repeated use of the quick suggestion chips. Using it in order to quickly gain more product suggestions, as well as quickly navigate oneself through the conversation.
Reading input fields
When chatting with the Gift Finder, users generally showed 2 prompting styles. Either keyword focussed, or writing longer story-esque prompts. Even though users can finish their tasks quite well in both cases, the users who wrote longer prompts were frustrated that the chat's input field didn't allow for reading back a prompt before sending it. It was also the case that when users send their prompts in the chat, the position stayed in the middle of a chat field, and became disorienting.
Losing a chat
When navigating between a product detail page and a chat, users on some occasions double tapped the back button. The back button for both product detail pages, as well as the chat are on the same position, this made users exit the chat and lose their history. In other occassions, users also had performance issues. This either made a chat wipe, or made users want to restart their chat.
FAB's and mental modals
Entering the Gift Finder, goes via the floating action button on the home page. In some cases, users expected the entry point for the Gift Finder to be within a search, a gift page, or a card on the home page. For them, this aligns more with their mental modal of how they see and imagine the layout of the bol app.
Fun gift finding
To improve the user experience with the Gift Finder, some key improvements were made regarding the LLM behavior to increase variety, as well as gaining an increased output limit. Chatting with the Gift Finder now ensures clarity while staying informative, by making sure users go to the most up-to-date chat message when entering a prompt. As well as now using an input area, instead of a field. Allowing users to always know what they are prompting. Regarding future improvements, some experiments such as an additional entry point have been added to the backlog.
As of October 2025, bol's first conversational AI on the consumer journey is live. With 15-20k daily users, and over 64% of these users saying it makes "finding gifts more fun".
