BackGenAi Chatbot
How can Ai add value to the online shopping experience?
GenAI LLM UXDesign tech
Agile Customer Service
E-Commerce
Role & Responsibility
UX Designer (Futurice GmbH): Research, Agile Facilitation, UX Prototyping
Product / Output
Click-dummy, Ai feature
Year & Duration
2024, 3 months
Client / Partner
[NDA] E-Commerce Platform
challenge
How might we deploy a GenAi-enhanced customer-facing chatbot to add value for store owners, their customer service teams and for end customers?
outcome
MVP implementation of a chatbot feature connected to an LLM in the environment of our partner's e-commerce platform.
context
The platform offers shop owners a full package for their e-commerce needs. With the current developments in GenAi Futurice and the platform provider joined forces to implement a GenAi feature in a very lean way.
insights
Most of the shops handle customer inquiries via phone and e-mail with own employees. Only a view use a rule-based chatbot. A GenAi enhanced chatbot could significantly reduce the costs for customer service.
impact
Fast Implementation
Together with our partner, we were able to implement the first epic /product recommendation in six weeks. Working in an agile interdisciplinary tech team, we focussed on concrete implementation.
My visual prototyping helped the team to get aligned quickly. This allowed us to move forward when we got stuck.
PROJECT JOURNEY
From Use Cases to a Chatbot Interface
In a nutshell
As a team we worked in a 2 week agile sprint rhythm. We validated our use case selection and learned from the shop owners about their needs. While creating happy flows, we started to implement the basic architecture in parallel.
Phases/Artefacts
1. Use Case Mapping
2. Secondary Research
3. Customer Interviews
5. User Flows
6. Figma Click Dummies
research & MVP Definition
I joined the project after the first round of customer research. Taking into account the efforts for implementation and where Ai can add value, the decision was made to implement a customer-facing chatbot.
We collected insights from a survey with shop owners (110 responses) and 3 qualitative interviews. We extracted the most important use cases for the MVP:
/Product recommendation
/Order status
From the interviews we knew that the opportunity to escalate to a human agent is very important:
/Escalation to human agent
Tech implementation
I was working closely together with the developers in the team. They were starting to set up the basic infrastructure, data bases, connections and agents in parallel to the design work. I facilitated our meetings to ensure that the whole team is on the same page.
tech implementation ©Futurice & [NDA]
tech model ©Futurice & [NDA]
user flows
We had a week in which we went round in circles with user stories and made no progress. We discussed the behaviour of the chatbot and which criteria should guide the answers of the bot.
After this week I proactively went ahead and thought through different possibilities. For this I used basic user flows in Miro.
The reaction of my developer colleagues was remarkable: We were immediately on the same page.
chatbot flow ©Futurice & [NDA]
user flow ©Futurice & [NDA]
Figma clickdummies
We wanted to learn quickly and validate our insights from the first research phase. I translated the user flows from Miro into click dummies in Figma. These click dummies were used in the next round of interviews.
Figma click dummy ©Futurice & [NDA]
Figma click dummy ©Futurice & [NDA]
Figma click dummy ©Futurice & [NDA]
My takeaway
Power of prototypes
In the first few weeks, it was difficult to speak the same language in the team. I took on the role of facilitating our planning meetings, reviews and retrospectives.
Nevertheless, we had the problem that we were going round in circles. It was such a moment of clarity when I showed my colleagues the first user flow in Miro. It was a tangible thought piece. From then on, it was easy to discuss it.
I learnt a lot about AI and working in a tech-heavy team.
user flows ©Futurice & [NDA]
You are curious? Drop me a line
Contact
Mail
LinkedIn2017 – 2024, © Benedikt Bandtlow. Rights of images and concept of commercial works remain with the respective clients.