G.A.R.I - AI Companion
G.A.R.I - AI Companion
G.A.R.I - AI Companion
How to turn a knowledge library for banking advisors into an AI companion to speed up processes and search information?
How to turn a knowledge library for banking advisors into an AI companion to speed up processes and search information?
How to turn a knowledge library for banking advisors into an AI companion to speed up processes and search information?
Year
Year
Year
2024
2024
2024
client
client
client
BNPP
BNPP
BNPP
industry
industry
industry
Banking
Banking
Banking
Service
Service
Service
Research
Research
Research
UX/UI
UX/UI
UX/UI
Prototyping
Prototyping
Prototyping
Design Thinking
Design Thinking
Design Thinking
Challenge
Challenge
Challenge
Design a proprietary AI, trained internally
Design a proprietary AI, trained internally
Design a proprietary AI, trained internally
Find the tone of voice and personality of the AI
Find the tone of voice and personality of the AI
Find the tone of voice and personality of the AI
Make this AI based companion accessible for everyone
Make this AI based companion accessible for everyone
Make this AI based companion accessible for everyone
Creating G.A.R.I
Creating G.A.R.I
Creating G.A.R.I
For demonstration purposes, G.A.R.I was initially based on a licenced free avatar, customized in order to represent the various customer segments. The choice was then made to start over from scratch to avoid licenses issues.
For demonstration purposes, G.A.R.I was initially based on a licenced free avatar, customized in order to represent the various customer segments. The choice was then made to start over from scratch to avoid licenses issues.
For demonstration purposes, G.A.R.I was initially based on a licenced free avatar, customized in order to represent the various customer segments. The choice was then made to start over from scratch to avoid licenses issues.
The idea was to make G.A.R.I as modular as possible, while keeping the same base, for quick iterations and to make future personalizations possible.
The idea was to make G.A.R.I as modular as possible, while keeping the same base, for quick iterations and to make future personalizations possible.
The idea was to make G.A.R.I as modular as possible, while keeping the same base, for quick iterations and to make future personalizations possible.
The base of the face and assets for accessorization were made using an AI tool, polished and put together on Photoshop.
The base of the face and assets for accessorization were made using an AI tool, polished and put together on Photoshop.
The base of the face and assets for accessorization were made using an AI tool, polished and put together on Photoshop.
G.A.R.I's faces
G.A.R.I's faces
G.A.R.I's faces
G.A.R.I's faces
Establishing G.A.R.I's personality
Establishing G.A.R.I's personality
Establishing G.A.R.I's personality
In order to define G.A.R.I’s tone of voice, his personnality and how he will answer specific requests, I led a workshop with the key stakeholders, inspired by Aaron Walter’s method for Persona Design.
In order to define G.A.R.I’s tone of voice, his personnality and how he will answer specific requests, I led a workshop with the key stakeholders, inspired by Aaron Walter’s method for Persona Design.
In order to define G.A.R.I’s tone of voice, his personnality and how he will answer specific requests, I led a workshop with the key stakeholders, inspired by Aaron Walter’s method for Persona Design.
This exercice was essential to lay down the foundation of the UX writing and the overall user experience while interacting with the AI.
This exercice was essential to lay down the foundation of the UX writing and the overall user experience while interacting with the AI.
This exercice was essential to lay down the foundation of the UX writing and the overall user experience while interacting with the AI.
Workshop's boards
Workshop's boards
Workshop's boards
Workshop's boards
Shaping the UX
Shaping the UX
Shaping the UX
Due to the wide variety of profiles amongst banking advisors, and in order to facilitate the quick adoption of the companion, G.A.R.I’s interface had to be easy to use and feel familiar.
Due to the wide variety of profiles amongst banking advisors, and in order to facilitate the quick adoption of the companion, G.A.R.I’s interface had to be easy to use and feel familiar.
Due to the wide variety of profiles amongst banking advisors, and in order to facilitate the quick adoption of the companion, G.A.R.I’s interface had to be easy to use and feel familiar.
After a benchmark of existing AI tools, and a priorization of the core features, the strategy was to design the structure of the experience to make it look like ChatGPT, due to its popularity amongst the users.
After a benchmark of existing AI tools, and a priorization of the core features, the strategy was to design the structure of the experience to make it look like ChatGPT, due to its popularity amongst the users.
After a benchmark of existing AI tools, and a priorization of the core features, the strategy was to design the structure of the experience to make it look like ChatGPT, due to its popularity amongst the users.
Within a week, an MVP version was presented to the stakeholders in high-fidelity wireframes, based on a real use case.
Within a week, an MVP version was presented to the stakeholders in high-fidelity wireframes, based on a real use case.
Within a week, an MVP version was presented to the stakeholders in high-fidelity wireframes, based on a real use case.
Finalizing the UI and create a feedback loop with beta users
Finalizing the UI and create a feedback loop with beta users
Finalizing the UI and create a feedback loop with beta users
After a few iterations, the final UI for the MVP was delivered to the dev team to be shipped to a sample of users.
After a few iterations, the final UI for the MVP was delivered to the dev team to be shipped to a sample of users.
After a few iterations, the final UI for the MVP was delivered to the dev team to be shipped to a sample of users.
A user testing protocol has been established to gather informations in the course of several months to improve the MVP before a broader release. This protocol includes a user journal based on a short form to answser after each use of G.A.R.I, and a monthly user interview to collect deeper feedbacks.
A user testing protocol has been established to gather informations in the course of several months to improve the MVP before a broader release. This protocol includes a user journal based on a short form to answser after each use of G.A.R.I, and a monthly user interview to collect deeper feedbacks.
A user testing protocol has been established to gather informations in the course of several months to improve the MVP before a broader release. This protocol includes a user journal based on a short form to answser after each use of G.A.R.I, and a monthly user interview to collect deeper feedbacks.
