Overview
Company > World-leading B2B retailer
Customers > Their eCommerce team
My Role > Product Owner (led research + delivery)
My Team > 1 UX Designer, 2 Software Engineers
Problem > Teams wanted to use AI to work faster and better, but many didn’t know how to use it well, and their experiments with public models risked leaking sensitive data and trade secrets.
Solution > I built a secure internal AI workspace that gave teams a structured way to explore, learn, and apply AI to their everyday work.
Impact > Adopted by 200+ users and established funding for a dedicated AI roadmap.
At a glance…
Spark Companion is a secure, connected, and collaborative AI workspace
SCROLL DOWN TO FIND OUT WHY AND HOW I BUILT IT
Who is Sonepar?
Sonepar is the world's leading electrical distributor, each year selling $35B of goods like cables, solar panels, and lightbulbs. They operate through 90 different brands in 40 countries.
In 2020, they launched Spark to bring all of these customers to a unified eCommerce platform. Now, 1/3rd of their sales are digital.
Spark is racing to bring more customers onboard…
/w=3840,quality=90,fit=scale-down)
So, in 2023, the team building Spark wondered…
“How can we use AI at work to speed up the delivery of the Spark platform?”
The need for safe AI
Some team members were already working faster with public AI tools…
…but they were accidentally sharing confidential data with the public.
AI tools like ChatGPT work best when provided with accurate data, but when they’re given access to internal documents and trade secrets, their other users are too.
Although AI gains were important, security was paramount.
So, Sonepar put together an AI task force for Spark.
👷🏻♂️ 1 PM
(that’s me)
👨🏻🎨 1 Designer
👩🏽💻👨🏼💻 2 Engineers
With only three months of runway… we got to work fast.
Building a secure workspace
With security as our top priority, we partnered with Microsoft Azure to get access to GPT-4, ensuring none of the data our users inputted into the AI would be used in retraining the model. Therefore, we eliminated the risk of sensitive data leaking out of the system.
Within 30 days, V1 of Spark Companion went live…
We designed a chat interface that would be familiar for users of ChatGPT
We invited 20 beta testers from the Spark delivery team to V1
This let us bring safe AI to power users and assess AI literacy across the wider team
We mapped where Spark Companion’s V1 did not meet user expectations of accuracy, response style, or efficiency
Adopting AI at work required a cultural shift, not just new technology
We determined that a successful Spark Companion must…
- Surface information across our internal documentation
- Encourage people to share tips and learn from others
- Be embedded into existing collaborative workflows
Making AI simple and approachable
In 2 months, Spark Companion went live for the whole Spark delivery team
On top of security-by-design, the full release featured…
1. Integrated Internal Knowledge
gives Spark Companion insider knowledge and makes it easier to prompt for the right information
2. Shared Chats
encourage teams to share learnings and integrate Spark Companion into collaborative workflows
3. Starter Prompts
to get everyone started using AI from proven examples of effective prompts
Turning usage into direction
As more teams used Spark Companion, our chats became a live map of how people actually wanted to use AI: the patterns, the frustrations, and the real work they were trying to speed up.
The results
200+ users signed up for Spark Companion in the first 30 days
Usage revealed signals where AI could scale across the 40K+ employees of Sonepar
Secured 12 months of funding to establish a permanent AI capability for Sonepar
This laid the foundation for my next big win…
Read about the Sales Automation that is saving Sonepar $400k/year
Where to next?
Get in touch
If you have any questions about this project, reach out and let’s set up a call.
- Email me at: edwardfraser@berkeley.edu
- Connect on LinkedIn: linkedin.com/in/edward-fraser/