The employees got there first: how we designed PardNEXT with Pardgroup
Nearly nine in 10 people at Pardgroup were using AI at work before the company had a program for it. We built PardNEXT with them on one rule: don’t sell AI to people who’ve already started. Give what they’re doing somewhere to go.

A display is designed in Milan, built in one of Pardgroup’s four production sites and installed in a store in London, Dubai or New York. By the time a shopper stops in front of it, the project has crossed several countries, companies and systems. Pardgroup builds the places where brands meet their customers, and its clients value the teams who do it.
In March 2026 the group’s shareholders and its CEO, Francesca Rapisardi, decided to change how that work gets done, with digital technology and AI, and asked us to design the program with them. They called it PardNEXT, and they started while results were good, when a company has room to decide for the long term.
Milano Finanza told the story on 28 October 2026 under the headline “PardNEXT, il programma di Pardgroup per utilizzare l’AI nel retail”. It led with one number. This is the work behind that number.
Start with the work, not the technology
We didn’t begin with tools. We began with what the tools would have to work on. We followed a project from the client’s brief to the installation in the store, asking three questions at every step: who decides, with which data, and where does that data live. More than 50 interviews in nine countries and 200 internal documents later, the picture was clear. Strong local teams. And underneath them, processes, definitions and systems that change every time you cross a border.
That matters more than it sounds. AI doesn’t fix inconsistency. It inherits it, and repeats it faster than any person could.
Nobody waited for the program
Then we asked the people. 77% of those invited answered our survey. Nearly nine in 10 already use AI tools at work. Nearly four in 10 use them every day. Most had started on their own, before any company program existed.
The usual plan for an AI program assumes people need convincing. Here, they’d convinced themselves. 85% say they’re ready to change how they work, and 78% expect the change to make their work better. What worries them isn’t their job. Doubts about whether AI output can be trusted come up more than three times as often as fear of being replaced, and curiosity nine times as often as anxiety.
Read that as a brief. These are people asking for better tools, reliable data, clear rules and investment in their skills.
The challenge isn’t to convince people to use AI. It’s to give direction to what they’re already doing.
Why you can’t ban your way out
There’s a less comfortable side to that number. When people use AI on personal accounts, internal documents and client data end up on platforms the company doesn’t control, and nobody checks the results. It’s called shadow AI, and the reflex is to forbid it. That doesn’t stop it. It only stops you seeing it.
The useful response is to build what no individual can build alone: data that can be trusted, rules about which uses to develop and under which controls, and processes redesigned around what AI makes possible. That step, from one person’s shortcut to the way a whole group works, is where transformations tend to stall. It’s also where the value is.
The part nobody sees
Most of PardNEXT is work nobody will see in a demo. It comes in two layers, and neither holds without the other.
The first is the organization. For each of 13 functions we wrote down which decisions belong to the group center and which to the countries. We drew a common blueprint for a country company, and built up at group level the functions that serve all of them.
The second is the AI, and it’s at least as much work. It starts with a data layer: one set of definitions across the group’s different systems, an owner for every data domain and controlled access, so a figure means the same thing in every office and to every AI solution that reads it. Then the processes. Redesigning a process for AI isn’t adding an assistant to step four. It means deciding again what the steps are, so we redrew each function’s key processes to show which steps AI carries out and where a person checks the result. And a roadmap brings the countries onto common platforms, because a solution built once should run everywhere.
None of it was designed in a room and handed over. Each plan was agreed with the person who runs that function, then discussed with the shareholders and the leadership team. A model people haven’t helped write is a model they’ll work around.
What it’s for
Buying licenses or adding AI to existing processes changes very little. The gain comes from redesigning the work, and from deciding what higher-value work people take on once the routine is handled.
The goal isn’t to do the same things with fewer people. It’s to do more with the people Pardgroup already has.
PardNEXT is now in implementation, and we’re carrying it out with Pardgroup as one team. The order matters. Data definitions, ownership and governance come first, so every AI solution draws on the same information under the same controls. Then the solutions are built and rolled out function by function and country by country, each measured against how the work was done before.
Key takeaways
- Ask before you assume resistance. At Pardgroup the people were ahead of the organization.
- Shadow AI is a governance question, not a discipline one. Bans move it out of sight; shared tools, data and rules bring it inside.
- AI inherits your operating model, so design the two together: decision rights and governance on one side, the data layer and redesigned processes on the other.
- Write the model with the people who’ll run it. Every function plan at Pardgroup carries its function head’s agreement.
Somewhere to go
Go back to that display on its way from Milan to a store. The people moving it along had already found their own ways to work faster. PardNEXT gives what they started one set of data, one set of rules, and processes built for the group Pardgroup intends to be.
The employees got there first. The organization is catching up, in the right order.
Disclosure: Marte works with Pardgroup on PardNEXT. Milano Finanza covered the program on 28 October 2026: read the article (in Italian).

