Who is Cedric Wells and why he's worth listening to

Cedric Wells is no ivory-tower theorist. As the former director of innovation and emerging technologies at Gorilla Glue — the American manufacturer of adhesives and consumer goods — he deployed new technologies in an environment where IT isn't the core business but still has to work reliably. It's precisely these kinds of companies that make up the majority of the market today, grappling with the same questions as Czech engineering firms, e-shops, or manufacturing plants: how to leverage AI to the fullest without endangering customer data and system stability.

In an interview published on July 17, 2026, Wells built on themes that resonated this year at the AI Agent & Copilot Summit NA. His message is simple: the AI era doesn't just need better tools — it needs better leaders above all.

Vibe coding: from an X joke to corporate reality

The term vibe coding was unleashed on the world in February 2025 by OpenAI co-founder Andrej Karpathy — he used it to describe a style of programming where a person simply tells AI what they want in natural language and lets it generate code without checking every line. What began as hyperbole ended up as Collins Dictionary's word of the year in November 2025. Today, app prototypes are built this way in hours instead of weeks.

In the podcast, Wells points out the flip side: speed of code generation is not the same as speed of deployment to production. Between "it works on my laptop" and "it runs securely for thousands of customers" lies architecture, security auditing, quality control, and regulatory compliance. Companies that skip this step, he argues, are buying technical debt — and in the worst case, a data breach.

Three pillars that will decide it, according to Wells

Three concrete recommendations for IT leadership emerge from the interview:

1. Growth mindset is the most valuable skill. "Technology changes so fast that learning is no longer an occasional activity, but a permanent responsibility," says Wells. Technical expertise from 2024 is no longer enough today — those who stop learning are left behind.

2. AI narrows the gap between business and technology — but doesn't eliminate it. Thanks to AI, managers today can grasp complex technologies faster, but they still need enough technical context to ask the right questions. A leader who blindly trusts AI outputs is just as dangerous as one who ignores AI entirely.

3. Governance and security must pull in the same direction. AI projects require close collaboration among infrastructure, security teams, and data stewards. Without it, sensitive information leaks or unreliable AI systems that nobody understands become real risks.

Shadow AI: 78% of employees use unsanctioned tools

Wells's statement that "employees are doing it, whether you want them to or not" is backed by hard data. A survey by WalkMe (a subsidiary of SAP) of a thousand US employees found that 78% of people admit to using AI tools their employer hasn't approved. Nearly half (49%) kept their AI usage secret from those around them, and only 7.5% received thorough training from their company.

This phenomenon is called shadow AI — the covert use of artificial intelligence outside the purview of the IT department. A typical example: an accountant pastes a client contract into a public chatbot to get it summarized. The data could leak outside the company without anyone ever knowing. Wells therefore recommends proactive governance instead of blanket bans: give people approved and secure AI tools before they find their own way.

What it means for Czech companies

For the Czech and European market, the topic has an extra dimension: the Artificial Intelligence Act (AI Act). Obligations for providers of general-purpose AI models have been in effect since August 2025, and most of the remaining obligations — including rules for high-risk systems — kick in from August 2, 2026. A company that doesn't know what AI tools its people are using today won't be able to demonstrate regulatory compliance — and fines can reach up to 35 million euros or 7% of global turnover.

The good news: tools for governed vibe coding are available in the Czech Republic too. GitHub Copilot, Claude Code, and Gemini Code Assist offer enterprise plans with data protection (code is not used for model training) at prices ranging from roughly 19 to 40 dollars per developer per month. The difference between chaos and managed adoption, then, isn't about money but about leadership approach — exactly as Wells says.

A leader in the AI era: less ego, more questions

The strongest moment of the podcast isn't about technology but about people. Wells describes how a successful IT leader in 2026 must be able to admit they don't understand something and go learn it — even with AI's help. "It's about bridging the gap with your leadership skills and leveraging AI fully on the technical side," he summarizes. In an era where AI agents handle more and more routine work, human value shifts toward judgment, context, and responsibility. And those, unlike code, cannot be generated.