Why I Spent My Sunday Installing an AI Agent — And Why Every Leader Should

Sometimes even as a senior leader, you need to go deep.

I learned this lesson years ago from my former boss — the President of a multi-billion dollar retail group in Southeast Asia. He didn’t just want a Tableau demo. He wanted to understand the databases, the data models, the application itself — in enough detail to build his own dashboards and generate his own insights. When we discussed how the technology worked under the hood, he went into granular detail. This was a man running a billion-dollar business, and he considered that level of understanding essential, not optional.

That stuck with me. And it shaped how I approach technology leadership today.

A Sunday Afternoon, Two Screens, One Experiment

So here I am on a Sunday afternoon in Bangkok. After an excellent lunch with my daughter, I sat down with my Mac — French TV soap opera running on one screen, terminal open on the other. A proper geek day.

My goal: to go hands-on with agentic AI. Not read about it, not attend another vendor demo — actually install and configure an autonomous AI agent myself. I wanted to feel the friction, understand the setup, and see firsthand what’s possible today.

Over the past few months, I’ve been exploring various AI agent tools with my team at Xponential (XPO), rolling out MCP Servers and self-setup agents. The team also kept trying new open source tools around there, to experience it hands-on as well.

But how about user enabled platforms? I have tried Manus AI — the general-purpose agent platform that Meta acquired for over $2 billion in late 2025, signaling just how seriously big tech is taking this space.

But today was about getting personal with the technology. I chose OpenClaw — the open-source AI agent framework created by Austrian developer Peter Steinberger that took the tech world by storm, surpassing 145,000 GitHub stars before Steinberger joined OpenAI in February 2026 to build the next generation of personal agents.

What I Did in 40 Minutes

The actual work took about 15 minutes. The French soap interrupted sometimes. Here’s what I set up:

• Connected OpenClaw to Telegram — giving me a conversational interface to interact with my agent from anywhere, including my phone.
• Configured the AI backbone — I set up Google’s Gemini for search capabilities and Kimi K2, a powerful Chinese language model, as the reasoning engines powering the agents. Both on free tier.
• Had the agent learn its environment — I asked it to understand the computer it was running on, complete its own setup, and secure itself.
• Added skills — OpenClaw uses a modular “skills” system (essentially extensions that tell the agent how to perform specific tasks). I added skills for productivity, self-improvement, WordPress integration for this very website and automated email responses.
• Using my laptop camera Web Cam to take automated images and share them with me
• Social listening on various firms and keywords, sharing reports with me .
• Started exploring what it could build next — including having it develop software and help me plan what to automate.

While we use many Google Cloud related tools, we are always open to explore more. Hence s

ome of my team members at XPO had already been experimenting with OpenClaw, and we recently held a townhall where everyone shared their experiences. Today I was catching up — a couple of weeks late, but with the benefit of their learnings.

The Strategic Implication That Should Keep Enterprise Leaders Up at Night

Here’s the thing that should make every CTO, CIO, and strategy leader pause.

Everything I just described — every single person can do. Every startup founder. Every ambitious intern. Every small business owner with a laptop and a Sunday afternoon. For a fraction of what the corporate enterprise vendor ecosystem charges for comparable automation.

The workflows I can now automate and validate through a simple Telegram chat — social listening, content management, email triage, research, even software development — are significant. These are capabilities that traditionally required enterprise software licenses costing six or seven figures annually, plus implementation partners, plus ongoing support contracts.

Can you build the next SAP or Salesforce with tools like this? That question deserves an honest answer: if you know in sufficient detail what those platforms do — or more precisely, what you need them to do — then increasingly, yes, you can replicate the workflows that matter to your business. Not the entire platform. But the 20% of functionality that delivers 80% of your value.

The ability to automate workflows, connect systems, and have an AI agent that develops and iterates alongside you is no longer the domain of well-funded IT departments. It’s available to anyone willing to spend a Sunday afternoon getting their hands dirty.

A Call for Honest Reassessment

From a strategic perspective, this shift demands honest consideration before any organization embarks on a large technology initiative. The questions that need asking:

• Which enterprise software, services, and advisory engagements are truly required — and which are legacy assumptions?
• What can be prototyped, validated, or even fully delivered using agentic AI tools at a fraction of the traditional cost?
• How does your technology roadmap account for the fact that your competitors — or a startup you’ve never heard of — might be building equivalent capabilities in days rather than quarters?

The key differentiator is no longer access to technology. It’s the connectivity and skills — the ability to define what you need, connect the right tools, and iterate rapidly. That’s a leadership capability, not a vendor relationship.

What’s Next

This was just a Sunday experiment. But it reinforced something I’ve believed for a long time: the best technology leaders don’t just set direction — they understand the terrain. My former boss knew that. And on days like today, I’m reminded why that matters more than ever.

Expect some deeper whitepapers from me on this topic — exploring what agentic AI means for enterprise technology strategy, vendor selection, and the future of digital transformation.

Just some Sunday night thoughts from Bangkok.