Agentic AI and the Future of Business Enablement

Last week, I was climbing boulders in the forests of Fontainebleau. This week, I’m back at my desk reflecting on something that also requires serious grip: agentic AI.

Takeaways from Google Cloud Summit Benelux 2025

Two weeks ago, I attended the Google Cloud Summit in Amsterdam. The message was clear: if you want to stay relevant, it’s time to start building a collaborative AI network of tasks, roles, and contexts. In other words: agents, agents, agents.

Google is betting big on this vision, with a growing ecosystem to support it:

• Model Context Protocol (MCP) for seamless integration with tools like Gmail, Drive, and third-party APIs
• Agent Development Toolkit for modular, task-oriented agent design
• Agentspace, an environment where multiple agents collaborate in real-time to power customer interactions

These announcements weren’t just technical showcases. The biggest shift was the emphasis on business value. Two years ago, the conversation was about infrastructure and scale. Today, it’s about enablement, simplicity, and live use cases that matter.

The gap between promise and reality

Despite the buzz, we’re still early. Most of the showcased applications looked promising, but also complex, resource-heavy, and likely expensive to deploy at scale.

In practice, many organizations are still dealing with fragmented IT systems, legacy tools, and incomplete data. Bringing your data together remains a challenge—let alone deploying autonomous agents across your stack.

So where do we start?

Large language models have already shown that data unification is essential. But the next step isn’t building agents for everything.

Instead, we should focus on proven ML applications where business value is clear. Think:

• Optimizing sales workflows
• Smarter marketing activation
• Reducing operational costs (returns, lost revenue, logistics)

Trying to apply a general-purpose LLM across your entire organization is often still a step too far.

The real shift: this is not just a tech theme

The most important insight I left with: agentic AI is an organizational challenge, not just a technical one.

If we want to move in this direction, we’ll need to start thinking differently:

• Modular working with clear handovers and responsibilities (think: playbooks, agent design patterns, testable RAG modules)
• Agents that collaborate like teams, with roles, context, and shifting priorities
• Sales through working demos, showing clients what actually moves the needle

I postponed writing this post—Fontainebleau was calling. But now that I’m back, one thing is clear:

Strategic leadership—not just smart tools—will determine who wins in the age of agentic AI.