AI & Design
Design Manager - Team Lead
Led a cross-functional team of data scientists and UX designers to integrate AI into customer experiences for Docaposte
Missions as lead:
- Advocated for user-centered AI, balancing technical feasibility with business value
- Ensured seamless collaboration between data science and design teams
2024
What I’ve learned:
AI is more than just chatbots
—It’s about solving real user problems with the right AI for the right job.
Docaposte’s challenge in a context where AI is rising everywhere
“How can artificial intelligence improve
the customer experience across our solutions?”
The answer?
“Don’t we already have a cross-functional UX team ensuring coherence across our BUs?”
✅ Yes.
What if we infused AI into that same transversal mindset?
Let’s build an Innovation Team: Design + AI.
Facile. 😌
(And let’s ask Steph to lead it)
One team, shared by all, built to serve everyone.
Funded by the executive committee, ensuring a budget independent of business units—effectively making it a free resource for BUs.
Staffing
Behind the scenes: 2 Data Scientists = peer-to-peer sparring
On projects: 1 UX + 1 Data Scientist, working side by side
For prototyping: 1 dedicated UI designer across all initiatives
Our offer
Ideation
Ideation to deliver innovative solutions and API-driven Proof of Concepts (POCs) to ensure reusability across projects.
Team missions: Field observation, Audit, Workshops, Target vision
POC
Prioritizing POCs based on business value and feasibility Diagnostic data : No data, no solution
Missions: UX Research, Workshop facilitation
Assessing the project’s data value potential
Ensuring the proper functioning of the AI solution
Providing a first, lightweight front-end experience
The challenge
Industrialization
Not covered by the executive committee’s budget, this has become a major transformation challenge for the team.
The key issue: avoiding the trap of working on POCs that never materialize—risking that the team’s expertise remains theoretical rather than delivering real impact.
Use case : LaPosteGPT
Generative AI for La Poste Group employees
Project Objectives:
Define the target vision for a unified platform as the go-to entry point for Generative AI tools.
Design and prototype the first version (V1) of La Poste GPT as well as the target vision for the platform.
Success Criteria:
A clear vision: The platform becomes an essential daily reflex.
Perceived value: The tool is indispensable in users’ workflows.
Methodology
Step 1 - Personae
With so many different employee roles, we grouped them into broad families to simplify the approach.
Questions informed by business insights into user needs:
Who are the target users?
All employees (personal assistant)?
Different teams with specific needs?
Both experienced and new users?
Is there existing user feedback or data available?
What are the users' expectations?
What current user problems are we trying to solve with this project?
“What if the future of AI in business isn't generalist, but specialized— and deeply trained?”
— Us at some point
But what if the real game-changer isn't just specialization, but how we centralize or integrate AI tools?
“Centralized platforms vs seamless integrations—what’s the best path to empower your teams and workflows?”
— Also us at another point
Step 2: The business goal in 2 workshops
Value Proposition Canvas – A format I love, especially once an MVP exists.
Prioritize one key persona: hybrid profiles who switch between client-facing tasks and preparation work.
La Poste GPT is the Group’s self-service generative AI platform
— modular, hybrid, built by and for La Poste employees.
Semantic Alignment Workshop
– We aligned all stakeholders word by word, mapping out a unified vision + the ideal user journey.
Just accessing the damn thing.
– How do I even get to La Poste GPT? Is it an app? A URL? Does it work on my device? Are there restrictions?
We built the entire journey around one central question: “Do you have time?”
→ If yes: how do you use it?
→ If no: what’s stopping you?
→ And in both cases: are you satisfied?
Step 3: User research
Strong demand on computer only : all the other tools from the Group are only accessible on desktop
Strong need for their personal HR demands
These insights helped design our top features:
Usage Library
Rich, structured answers: Tables / Graphs / Offer comparison / Customer insights
for those who interact with clients directly: that they can have the direct link to subsctription links
Deliverables
Guiding Principles of the Target Vision
Usage Library to guide users from the home page
Categorized, ready-to-use actions (or create your own!)
Pin your favorite use cases for quick access
Customization available for any user
Fully responsive experience
Voice input enables hands-free interaction. The vision? A truly conversational assistant you can even use on the move.
One Entry Point
A seamless, centralized experience—no more tool overload
Embedded in existing business tools
One single access across the company
Ask one question → AI picks the right sources
Optional: filter sources manually
Flag outdated or untrusted data
By default, answers use all trusted sources
Beyond GenAI
AI isn’t just about chatbots—
It’s about solving real user problems with the right AI for the right job.
— Our GenAI fatigue
Pure AI approach
You don’t necessarily need design on all AI projects
Use case:
The finance team needed a smarter way to predict customer payments and optimize cash management at scale.
—just a seamless algorithm embedded directly into existing BI tools. The “interface” is the prediction itself.