AI may be everywhere, but in digital analytics, it often feels more like a novelty than a working tool. For Gunnar Griese, that’s not a failure of the models themselves, it’s a failure of integration.
“LLMs are great at talking,” he says, “but they’re isolated from the tools we actually use. That’s exactly what MCP servers fix.”
On October 9, Gunnar will take the stage at the Digital Analytics Summit 2025 to explain how MCP servers (Model Context Protocol servers) are quietly turning AI into a powerful execution partner. One that doesn’t just suggest what to do, but actually does it.
What are MCP servers and why do they matter?
“While LLMs are great at generating text responses, they’ve been isolated from the tools we actually use in our work,” Gunnar explains. “MCP servers act as intermediaries that enable AI applications not only to discuss what should be done, but also to actually do it.”
In short: rather than having to manually copy-paste code or AI suggestions, analysts can now build workflows where AI “understands” the setup and implements changes within analytics platforms directly. It’s a shift from passive suggestion to intelligent action.
From assistant to execution partner
Gunnar has used tools like ChatGPT and Claude to solve coding problems and work faster, but, until recently, always stayed in charge of execution. That’s changing.
“With MCP servers, we can start outsourcing the laborious, time-consuming technical tasks to AI agents,” he says, “while maintaining human oversight for strategic decisions.”
The benefit? Analysts can spend less time configuring tags or debugging setups, and more time thinking critically about measurement strategy, business context and what the data actually means.
Automation meets human oversight
Will AI replace digital analysts? Not according to Gunnar. Instead, he sees a division of labor emerging: AI handles the mechanical implementation, while human experts take care of strategic guidance, interpretation and quality control.
“AI can execute tasks efficiently,” he says, “but it won’t replace the strategic thinking, business context understanding, and quality assurance that experienced analysts bring.”
This becomes even more crucial as privacy regulations tighten. Automation must go hand-in-hand with ethical data handling and clear oversight. This is a task that still firmly belongs to people.
So how do you start?
“Don’t overcomplicate it,” Gunnar advises. “Start with one repetitive, annoying task in your workflow. Something technical and time-consuming. That’s your entry point.”
Many building blocks are already available. Try the GTM MCP server for tag management, or the Playwright server for automating browser tasks. Hook these into an AI interface of your choice, and start experimenting.
“Don’t expect it to work perfectly the first time. My proof-of-concept needed multiple iterations. But that’s the point: you’re building something new. The goal isn’t to replace your expertise; it’s to let you use it where it matters most.”
The future is now
MCP servers might sound futuristic, but they’re already here, and the people experimenting now will be tomorrow’s leaders.
“We’re entering a new phase,” Gunnar says. “If you’re still doing everything manually while your competitors are automating with AI, you’re going to fall behind.”
Ready to see how it works?
🧠 Catch Gunnar Griese live at the Digital Analytics Summit 2025, on October 9 in B. Amsterdam.