Schneider Electric operates across energy management and industrial automation, supplying products and software that help customers run homes, buildings, data centers, and factories more efficiently. That positioning shapes the premise of this three-part CDO Magazine interview series: if AI is meant to improve efficiency, how does a company actually make it real in day-to-day operations, and what lessons hold up beyond the hype?

In this final part, Philippe Rambach, Chief AI Officer at Schneider Electric, speaks with Dr. Julian Schirmer of OAO, about the “what.” The conversation distills what works, what does not, and the key learnings as Schneider pushes AI into production.

In Part 1 of the series, Rambach explored AI’s role in organizational strategy, and in Part 2, the conversation explored how AI is operationalized inside the enterprise.

AI transformation is first a transformation

Rambach begins with a lesson he believes many organizations overlook once AI enters the room. “Everybody speaks about the AI transformation. But AI transformation is first an IT transformation, and it’s first a transformation,” he says.

He argues that the hardest work is familiar work. “Actually, 80–90% is change management, training the people deploying software, which we know is difficult,” he says.

In Rambach’s view, AI has unique challenges, but teams can lose time by treating everything as entirely new. For the people expected to use it, AI often shows up as something simpler: a feature inside the tools and workflows they already rely on.

What does not work: Shipping “Another Tool” instead of a feature

From there, Rambach moves to a concrete adoption mistake and the adjustment Schneider makes. His team is focused on ensuring AI solutions are actually used, not admired from a distance. “We’re focused on ensuring that our AI solutions are adopted and used to make sure that they don’t stay like, ‘It’s great, but nobody uses it’,” he says.

The learning is blunt. Rambach notes that integrating an AI feature in a tool adds to its usage. He illustrates it with an internal example. Schneider builds a tool for marketing teams to help them comply with European regulations related to greenwashing. “We say, ‘You just upload your job, and then you get the answer.’ Guess what? They don’t choose it,” he says.

Adoption changes only when the capability moves into the daily system of record. “Now that you are embedding it (AI) into the tool they use daily to manage our documentation, they use it,” Rambach explains.

Training is the anchor for adoption and trust

Responding to Schirmer’s query about how Schneider navigates change management, fear, and uncertainty, Rambach does not claim a perfect playbook. “I don’t have the perfect answer for that,” he says. But he identifies the most important lever: Training.

His reasoning starts with a simple observation. “People hear about AI every day, every minute, but they don’t really know what they talk about,” Rambach says.

He describes the range of mindsets he encounters, from fatigue and skepticism to panic and extreme expectations, and argues that training must calibrate understanding. “We need to make sure that they understand what it does and what it does not,” he says, “sometimes lowering expectations, sometimes increasing expectations.”

Four training audiences, one companywide mandate

Rambach then describes how Schneider structures training at scale. “We have split our 140,000 employees into four categories,” he says.

The first group is everyone, and Schneider treats it with the same seriousness as compliance training. Rambach explains that, like the annual corporate training system, this is also tracked, escalated through reminders, and tied to consequences. “We have added the same level of scrutiny; a mandatory training of 140,000 employees, including production line workers, on AI,” Rambach says.

The second group is senior leadership. “The top management of 500 or 1000,” needs to understand the value proposition, the “why,” and the limitations well enough to explain them further.

The third group is AI experts, who need continuous updates as the field moves.

The fourth group is the most operationally important and the most difficult to identify: the people who own transformation and delivery, such as product owners, IT owners, and process owners. Rambach wants this group to understand practical limitations and decision-making tradeoffs, not the mathematics of model weights.

He uses accuracy to make the point. “If we work very hard together, we may reach 85-90% accuracy. Is that good enough, or do we need a hundred percent?” he says. Then he draws a line: “If you need a hundred percent, let’s not waste time. We’ll not reach a hundred percent.”

Best practice: Resist the shiny object and start with value

When Schirmer asks what other companies should learn from Schneider, Rambach returns to the discipline that runs through the entire series. “It’s a constant fight to some extent, starting from the business value and the use case at scale,” he says.

The pressure to begin with technology is relentless, he explains, because people are constantly receiving messages about new capabilities. “There is a very strong temptation for everybody to start in technology,” Rambach says, describing the pull of the “latest technology.”

He does not deny that exploration matters. But he brings the focus back to outcomes. “In the end, we need to create business value,” Rambach says, calling it an ongoing balancing act.

What makes a Chief AI Officer: Domain and company first, AI as a learnable layer

Schirmer then shifts to Rambach’s personal story and asks for advice for aspiring AI leaders. Rambach starts with the moment Schneider decided to accelerate AI and called him to lead it. 

The organization sought someone who knows the business because they wanted AI “at scale.” They wanted someone who understands how change moves through the organization and is trusted by the company, and advised him to learn about AI.

From that, Rambach distills what he believes the role truly requires: “You need to have business knowledge of the domain. Of course, you need to have company knowledge and some AI knowledge.” For him, AI competence is essential, but it is also the piece that can be built over time.

The daily challenge: Downsizing expectations

In the final exchange, Rambach sheds light on his biggest day-to-day challenge as Chief AI Officer. “Normally, you spend all of your time explaining how what you do is incredible and try to convince people to put more money in it,” he says. In AI leadership, he feels the opposite is often required.

“To some extent we spend a lot of time downsizing expectations,” Rambach says, describing a world where people are “bombarded with information” that exaggerates what AI can do. The job, as he sees it, is helping people land at the right level of understanding, so the company builds what is achievable and valuable.

He closes with a familiar scenario that captures the risk of hype. The organization was debating a long-standing problem, and someone turned to him with confidence that a new AI technique would solve it. “I’m sure that it should be an agent,” Rambach recounts, and his reaction was simple: “I’m not sure.”

CDO Magazine appreciates Philippe Rambach for sharing his insights with our global community.