Copilot Deployment at Longchamp: From 10% to 98.6% Adoption in 6 Months

No budget, no sponsor, 10% usage. That was the starting point for the Copilot project at Longchamp. Six months later: 98.6% satisfaction, 3.4 hours saved per week per user, and exploding internal demand.

How did a luxury house transform a timid test into measurable success? The answer lies in one principle: applying its own codes to transform AI from a consumer product into a luxury product. Renaud Beaufils, Chief Data Officer at Longchamp, and Christelle Vertus, Customer Success Manager at Devoteam, reflect on adoption through practice, without fanfare or excessive promises. Just concrete training, a restricted community that created desire, and results that ultimately convinced management.

Here’s the behind-the-scenes story of a Copilot deployment that proves that when it comes to AI, method matters more than technology.

A Pragmatic Launch Context

Unlike the spectacular announcements sometimes shared on LinkedIn, the Copilot project at Longchamp wasn’t born from immediate management conviction. Instead, the company seized a Microsoft funding opportunity to test the technology with minimal initial budget and no strong sponsorship at launch.

The business context was clear: a fast-growing company with a recruitment deficit and overloaded teams. Objectives centred on ROI and efficiency gains. As Renaud Beaufils explains: “In luxury, the main challenge remains operational efficiency rather than pure innovation.”

The starting point presented challenges: 50 licenses available, 40 assigned, but a usage rate of around 10%. A classic scenario where the tool is distributed without the necessary support for adoption.

Renaud Beaufils and Christelle Vertus during the Devoteam Microsoft AI Experience event

An Action-Oriented Training Strategy

Faced with no dedicated budget and weak sponsorship, the team adopted a direct and pragmatic approach. Rather than multiplying theoretical awareness sessions, priority was given to immediate hands-on practice.

“We didn’t try to convince people or explain the technology at length. We went straight to training,” explains Renaud Beaufils. The goal was to let users discover the tool themselves, develop their own use cases, and then support them with additional content.

Prompting training formed the foundation of this approach, a choice that proved pivotal going forward. Christelle Vertus emphasises the importance of this practical approach:

“Users need to test concretely and see how the solution responds to their specific needs.”

The adoption plan wasn’t fixed but constantly evolved based on field feedback and emerging needs. Team workshops identified priority use cases and business issues to address, followed by POCs and feedback sessions.

The Power of Exclusivity and Communities

A key success factor often underestimated: creating a sense of belonging. “Creating a relatively restricted group at the start generated a desirability effect among other employees,” observes Renaud Beaufils. This selective approach sparked interest and the desire to join the program.

Community recruitment happened progressively and iteratively. Managers were consulted to identify profiles that were curious and inclined to challenge their habits. But the team also implemented an adjustment process based on actual usage data, evolving the pilot group composition to maintain high engagement levels.

A community on Teams and SharePoint was then created to distribute content, share use cases, and capitalise on knowledge of each team’s business issues. These Champion or Ambassador communities can evolve over time, enriched by new teams expressing their desire to get involved.

Features That Transform Daily Work

Copilot adoption varied by business profile. Retail teams, accustomed to managing large amounts of data with limited tools, showed high adoption rates. Internal training teams, project management, and human resources also identified high-value use cases.

For project management, benefits are tangible, including the generation of meeting minutes, RFP preparation, and project plan development. “Automatic minutes compensate for the lack of communication that existed before, particularly when no one formally took notes,” explains Renaud Beaufils.

Presentation creation also illustrates productivity gains. Using structured prompts that precisely define needs, target audience, and key messages, it’s possible to generate professional-quality presentations in a fraction of the time usually required.

Translation represents another major use case for Longchamp. This feature, which translates complete presentations with approximately 95% accuracy, proves particularly relevant for a company that must train teams in its international factories. Real-time translation in Teams also addresses concrete multilingual collaboration needs.

However, some profiles didn’t adopt the tool as anticipated. IT teams, particularly developers, showed more limited adoption, partly due to a mistrust of generated code quality and confidence in their own expertise.

KPIs That Convinced Management

The major challenge was proving ROI without initially having sophisticated measurement tools. The team relied on three main KPIs: time saved, adoption rate, and user satisfaction.

The decisive KPI for unlocking budgets was time saved, measured by surveying the equipped population. “With 173 users having the paid license, we measured an average of 3.4 hours saved per week,” indicates Renaud Beaufils.

The adoption rate now reaches 98.6%, demonstrating real tool appropriation. Adoption rate tracking also identifies teams requiring additional support, while satisfaction by tool guides training efforts on new features.

These quantitative results were complemented by qualitative benefits:

• Document standardization
• Improved information structuring (like User Stories)
• Reduced missing information in deliverables

Obstacles Identified and Overcome

Several obstacles emerged during deployment, requiring adjustments to the support approach.

The first obstacle concerns expert profiles who believe their manual production remains superior to what AI can generate. This perception, particularly present among developers, requires a personalised approach to demonstrate added value on their specific use cases.

Occasional hallucinations also created credibility issues, although their frequency decreased with model evolution. These situations required transparent communication about technology limitations.

The tone of AI-generated messages presents a particular challenge in the luxury universe, where customer relationship quality is paramount. Some departments remain vigilant about the potentially artificial character of generated content. Christelle Vertus emphasises the importance of detailed prompting to solve this problem: taking time to build precise prompts with details on expected structure, desired tone, and required expertise level.

Finally, concerns about job replacement persist among some employees. These fears are addressed through individualised support, in small groups or dedicated team sessions. “We explain that the tool not only improves work quality but also develops a valuable market skill,” specifies Renaud Beaufils. About thirty people benefited from these reinforced support sessions.

The Future: Copilot Studio and Broader Deployment

Longchamp now approaches a new maturity phase. Early adopters have been won over, and the strategy evolves to reach follower profiles. Deploying free Copilot Chat now serves as a pool to identify engaged users who can then access paid licenses.

This “Leagues” approach maintains the exclusivity effect while optimising paid license allocation according to budget constraints. Usage data enables nominative identification of the most active users and prioritisation of their access to premium features.

In parallel, Longchamp launches into Copilot Studio after an experimentation phase. The objective is twofold: develop a business use case with clear ROI by relying on a partner, and simultaneously build internal governance enabling the emergence of Copilot Studio “Citizen prompters.”

Adopting new features remains a permanent challenge. As Renaud Beaufils highlights: “Users receive initial training but don’t necessarily follow evolutions. Yet truly transformative features regularly come out.” The ambassador community plays a crucial role in identifying, testing, and distributing these innovations via the collaborative platform.

Advice for Launching an AI Transformation

From this experience, several recommendations emerge for companies considering similar deployment:

Favour a Progressive Approach

“Starting with a restricted core of a few people proves more effective than massive deployment,” recommends Renaud Beaufils. Adoption works largely through mimicry: observing a colleague effectively using a tool naturally creates the desire to adopt it.

Create Desirability

The feeling of belonging to a selected group constitutes a powerful engagement lever, particularly in early project phases.

Optimise User Time

Offer high-value content, quickly exploitable. For the most reluctant profiles, a simple approach like replacing classic web search can be a good entry point.

Maintain Continuous Support

“You must regularly offer ready-to-use content that users can progressively appropriate,” emphasises Renaud Beaufils. Adoption doesn’t happen at first contact but requires several solicitations.

Accept Change Temporality

Allow necessary time for new practices to establish durably, without imposing too sustained a pace that could generate rejection.

Personalize Support

Small group or team sessions create a reassuring environment where users can express difficulties without judgment.

This experience demonstrates that beyond technology, it’s the human, pragmatic, and iterative approach that determines generative AI adoption success in business. A transformation that started without a dedicated budget and now convinces through measurable results and high appropriation level.

Focus on Devoteam Support

In this interview, Renaud Beaufils shares his feedback on Copilot deployment and Devoteam’s support.

The Challenge

Can you describe the context of your collaboration with Devoteam on the Copilot project? What were your initial request and specific needs?

Initially, our request was relatively open. We had Copilot licenses and a support budget, aiming to measure return on investment. So we called on Devoteam’s expertise to structure our approach. The approach was truly collaborative: we integrated their methodological recommendations, which we adapted to our context. Most of the proposed framework came from Devoteam.

So that allowed you to structure the approach and establish clear milestones?

Exactly. We relied on proven methodologies, which we customised to our needs. We did adaptation work (reorganisation, adjustment), but started from a solid structured base.

Had you defined precise objectives from the start regarding target groups, first users, or priority use cases?

Our approach was truly exploratory, without rigid predefined objectives – and that’s still the case today. That’s actually what makes this project particularly interesting: we’re building progressively, without strict constraints. For us, it’s an innovation program, which is new. We’re giving our employees the opportunity to build their own vision of their work tools, while supporting them in this reflection.

Doesn’t this approach risk worrying some organisations, who might fear starting from a blank page without clear visibility on the destination?

The initial phase doesn’t require a considerable time investment, notably because Devoteam has a perfect mastery of deployment challenges. In project management terms, startup is quite smooth. However, support must be maintained long-term, with constant but reasonable effort. Personally, I rarely dedicate more than one day per week to the project, and today, about two hours per week suffice to maintain high adoption levels and integrate new users. Investment is therefore moderate, but must be regular.

The Solution

Has support evolved over time? Did you regularly organise checkpoint meetings to adjust the approach?

We established weekly follow-up meetings throughout the initial phase. These half-hour to one-hour sessions covered project management and adoption tracking. We adopted an agile approach, allowing us to continuously adjust our hypotheses and choices. This frequency proved largely sufficient to effectively pilot the program.

In terms of skills, what’s the value brought by a partner like Devoteam?

On a tool as open as Copilot, value mainly resides in feedback capitalised with other clients. Devoteam brings deep knowledge of possible use cases, classic issues and their solutions, as well as integration of different tools with each other. They also have a vision of upcoming evolutions.

One aspect I hadn’t anticipated was their detailed knowledge of the Microsoft ecosystem. Microsoft offers support funding, but without a partner like Devoteam, we wouldn’t have been aware of these opportunities. It creates a win-win dynamic. Devoteam helps us navigate Microsoft’s complex environment – licenses, programs, etc.

In your view, what’s the overall added value of this Devoteam support?

We benefited from structured and effective project management. The project proceeded without delay, with controlled tasks delivered on time. The project was fixed-price, and this budget framework was respected without overrun. I had constant visibility on progress and remaining resources. Add to that expertise on the tool and training, which correspond exactly to what you expect from this type of mission.

What did you particularly appreciate in this relationship?

What struck me was Christelle’s exceptional responsiveness, our contact. We set up a communication channel adapted to our usage – a Teams channel allowing real-time exchanges. Responsiveness was constant, and there was perfect consistency between what was announced and what was delivered. This predictability, combined with skill level, created immediate trust. Conducting a project under these conditions is truly comfortable.

The Benefits

Beyond measurable quantitative gains, do you identify more qualitative benefits related to Copilot deployment?

The major qualitative gain lies in structuring our data and company knowledge. Previously, a meeting without minutes represented a net loss of knowledge. Today, we effectively capitalise this knowledge without disproportionate effort; transcription and content exploitation have a very reasonable cost.

Concretely, we gain in internal knowledge quality. For example, if I don’t know a colleague, Copilot can generate a descriptive sheet that lets me quickly understand their role and responsibilities before engaging in a discussion.

We also see significant improvement in our deliverables. Documents and presentations are now more homogeneous and better structured. Copilot ensures consistency throughout a document, uniform formatting, and harmonised presentation conventions, which previously avoided time-consuming inconsistencies between sections or slides.