Pierre-Yves Calloc’h, Global Chief Digital Officer, Pernod Ricard, speaks with Julian Schirmer, Co-Founder of OAO, in a video interview about managing AI hype, training, focusing on innovation through business value, transparent decision-making, scaling AI solutions, the importance of proof of value, and driving adoption through early engagement and investing in user interface.

Managing AI hype: From education to scalable impact

At the outset, Calloc’h reflects on the balance between enthusiasm and execution. For him, the hype around AI is not a distraction; rather, it is a catalyst for focus. His approach begins with education.

“The first thing we are doing is first educating, explaining to people what ChatGPT and other tools can do.” He notes that employees are urged to explore generative AI (GenAI) for personal productivity and creativity. Crucially, they are made aware of tool limitations and confidentiality risks.

The training does not stop at how to use the tools, says Calloc’h, as it dives into why certain things work or do not. He also points out that models trained on public internet data often underperform when applied to internal company data. Even when using advanced techniques like retrieval-augmented generation (RAG), results can still be influenced by public data sources.

Using a familiar analogy, he underscores the value of technical understanding: “When people know how a car works, they know it won’t run on water and it won’t fly.”

Focusing innovation through business value

“The second thing we are doing is we are encouraging people to provide ideas with a business case,” says Calloc’h. He stresses that scalable AI applications require not just ideas but investments in data infrastructure, model processing, and UI development, so they must create tangible value.

Moving forward, Calloc’h notes that use cases must be tailored to specific business contexts.

“A bank will invest in fraud detection. For us, we don’t process any payment by credit card, not a great use case.” Even between competitors, use cases can diverge significantly based on business models, he adds.

“If my direct competitor has only six brands, it’s because they have a different strategy. It’s not relevant for them.” The selection process filters ideas through business, financial, technical, and data feasibility lenses, says Calloc’h.

Ultimately, Calloc’h emphasizes transparency in decision-making. “You also need to explain to people how you are making those decisions in a transparent way so that it can be challenged,” he states.

 “The good thing about the technology is that at the end of the day, you need to deliver lines of code, facts, etc. And so, it becomes a non-emotional decision. And when you explain and do the education around it, it works much better,” Calloc’h says.

Scaling AI solutions: From MVP to multinational rollout

According to him, the path from MVP to full deployment is a technical and organizational journey grounded in tested value and structured planning.

“The first thing we make sure of is that we scale something that works.”

For Calloc’h, the foundation of any scalable initiative starts with a live test in a real market environment. The team uses the Minimum Viable Product (MVP) process to validate effectiveness early.

He shares the example of a tool developed for the Salesforce team to optimize which outlets they visit and what actions they take in-store. Before scaling, it was tested in two countries to ensure performance gains were real.

MVPs, by design, are often held together by manual processes and temporary workarounds, says Calloc’h. “Usually, you do those MVPs with a lot of quick things. So you bid with Excel spreadsheets. People are pedaling in the background to make sure that things are delivered every Monday.”

But such methods do not scale, he says. “There’s no way you can scale that to 20 countries because you would need an army to maintain it, and it won’t be as stable.”

Once the value is proven, the team shifts to industrializing the process:

• Data pipelines are harmonized across countries.
• Machine learning engineering ensures code works across multiple countries and multiple brands and runs efficiently — not in 22 hours, but “just in a few minutes.”
• Error detection becomes automated.

“You progressively develop all those elements so that it runs smoothly,” says Calloc’h. Ultimately, the ideal state is to develop data products that systems can consume directly. Further, the proof of value justifies resourcing. Once the system is validated, it becomes easier to secure buy-in, Calloc’h explains.

He further shares that scaling also means seamless integration into the company’s existing IT environment. Going back to the Salesforce example, the MVP was initially deployed as a web application viewable on mobile phones.

Later, it was embedded into the sales automation system, providing recommendations and planning on the same interface the sales teams were already using.

“This is the part that you do once you are past the proof of the value. And you can finance all of that because you have proven that it works and that when it scales, it’ll bring more value to the company,” says Calloc’h.

Driving adoption: Why early engagement and UX investment matter

Furthermore, he frames adoption as a design problem that begins on day one, well before any full rollout. Calloc’h insists on engaging a wide cross-section of the organization in the MVP design.

The more people are invited in early, the stronger the design and the easier the buy-in later. He uses a simple analogy: when people help choose the “film,” they are far more likely to enjoy it afterward. In practice, that means involving:

• All levels of the hierarchy — from marketing directors and brand managers to brand assistants
• IT and business stakeholders — not just eventual users but also respected “reference” figures who influence peers

Use credible champions and peer-to-peer proof

On top of that, Calloc’h favors advocates with frontline credibility. For example, the Salesforce optimization project was led by a former U.S. sales director, which changed how the message landed in other countries — practical, in their language, and relatable. He also keeps pilots visible and honest: “We are also using the pilot countries as reference.” Those references include what works and where the limits are, fueling authentic, colleague-to-colleague conversations rather than top-down mandates.

Speaking of this internal communication, Calloc’h states that a clear explanation of benefits and constraints helps teams understand why the product is built the way it is and what is expected of them. That transparency smooths the path to adoption and reduces resistance.

Thereafter, he emphasizes the importance of user interface design in driving adoption, noting a conscious decision to overinvest in user interface. He contrasts this approach with the more rigid strategies of the past, where countries had little choice but to accept predefined systems.

Instead, his team has opted for flexibility and local adaptation to encourage broader uptake. While using pre-built solutions helps accelerate deployment, Calloc’h stresses the need to retain some room for local tailoring.

“It’ll go faster if you take the things that have already been developed, but we kept the ability to adapt so that we can get that adoption.”

CDO Magazine appreciates Pierre-Yves Calloc’h for sharing his insights with our global community.