Capgemini Exec Says Successful Leaders Stop Layering AI, Simplify Processes First

Enterprises are rushing to deploy AI, but many of these initiatives are set up to fail. Instead of using artificial intelligence to reinvent their processes, leaders are layering it onto legacy workflows. Now, that approach is only amplifying existing dysfunction. By treating AI as a plugin for inefficiency rather than a design tool for transformation, the result is faster chaos, not tangible value.

For an expert's perspective, we spoke with Dinand Tinholt, Global Industry Lead of CPRD Insights & Data at Capgemini. Drawing on decades of experience in data-driven transformations, Tinholt explained how other leaders can escape this cycle. For him, the industry’s obsession with superficial metrics and legacy processes is the root cause of its struggles with AI.

"Business intelligence and BI reports are a bit like adult coloring books for executives," Tinholt said. "They're nice, peaceful, and very therapeutic, but they often stop short of driving action." He believes the path to a lasting ROI requires a radical shift in mindset.

• Decisions over data: The solution is moving away from creating prettier reports and toward driving more intelligent decisions, Tinholt explained. "I don't care about business intelligence. I care about decision intelligence."

The rush to implement AI has led many organizations down a familiar, flawed path, according to Tinholt. Rather than rethinking broken systems, most are simply trying to automate them.

• Automating dysfunction: Meanwhile, this approach is little more than a shinier version of old automation techniques that fail to unlock AI’s true potential, Tinholt said. "It's just taking the existing process and saying, 'Can I replace a human there? Or can I get a bit more efficiency?' That's kind of like RPA 2.0, and it's not being creative," he stated.

But the real path to transformation begins with simplification, Tinholt said. Before deploying a single algorithm, leaders must first challenge the logic of their existing operations.

• Simplify the solution: "The first thing that you really need to do is look at what process you are doing. Can you standardize, can you simplify, and can you minimize the process first?" Tinholt asked. "We have this whole methodology: Eliminate, Standardize, Optimize, Automate, and Robotize."

• Design over decoration: Also important is a disciplined, foundational approach that prioritizes clarity over speed. "Don't treat it as a plugin," Tinholt advised. "Treat it as a design tool and take a systems thinking approach."

Unfortunately, internal bureaucracy remains one of the biggest hurdles to transformational work, Tinholt continued. Often, necessary functions like risk, compliance, and governance are weaponized to slow progress and maintain the status quo. "We do these initiatives and hide them behind 17 layers of governance. We need legal to sign off on this, and risk to sign off on that. Governance and risk are being used too much as an excuse not to do some things."

To counter this, Tinholt proposed a practical model for building trust incrementally. Rather than demanding a leap of faith, organizations should design systems that allow humans to validate AI-driven proposals before granting full autonomy.

• Trust but verify: By giving them the tools to verify the logic and build confidence over time, leaders can turn skeptics into partners, Tinholt said. "When we propose to have AI take over certain decisions, we're not going to have AI take over. We're having AI propose the decisions. Then the human can click 'implement.' You start adjusting your thresholds, giving autonomy to the systems. But it's a gradual process."

Because the cost of inaction is existential, strategic patience is also critical. Here, Tinholt encouraged leaders to engage in a simple but powerful thought experiment: imagine a well-funded, AI-first competitor entering your market tomorrow.

• Think like a startup: Such a mental model shifts the focus from incremental gains to the urgent need for fundamental transformation, Tinholt explained. "What would a competitor like that look like? And how is that different from what you're doing? Don't just think those small baby steps, but think big and transformational. Because otherwise you will get wiped away sometime in the future."

Ultimately, the journey to becoming a data-powered enterprise requires resisting the temptation to automate dysfunction. "We don't necessarily just need faster processes," Tinholt concluded. "But we do need smarter ones, and we need different ones. Sometimes we need fewer ones. So it's a scalpel, not a sledgehammer. When you redesign rather than automate, you unlock new business models, new decision-flows and sustainable competitive advantage."

For him, the real power of AI is not in speeding up bad processes but in revealing the flawed logic that created them.