How It Actually Started

My path into AI began like most: ChatGPT, custom GPTs, gradual workflow integration. Useful, but still manual. I was doing most of the thinking.

The shift came when I moved to Claude and started treating AI as an operating layer, not a chatbot. I migrated my GPTs into structured skills and connected the tools I actually use: email, calendars, project data and scheduled recurring tasks I'd been tracking manually.

The tools matter less than the approach. Whatever AI you use, the principle is the same: Stop asking questions and start giving it responsibilities.

Operational work that used to consume mental space (constant, not hard) started running in the background: status updates, data pulls, scheduling logistics, first-draft communications. Asana's research shows workers spend 60% of their day on coordination tasks, leaving only 27% for skill-based work and 13% for strategy. That freed up something unexpected: time to think about what actually needs me.

The Automation Trap

Here’s where most AI stories go wrong. They frame adoption as a progression: manual, assisted, automated and fully autonomous, implying more is always better. A Deloitte survey found that 85% of organizations increased their AI investment in the past 12 months, yet only one in five qualified as true AI ROI leaders. Most companies focus on how much to automate, not what to automate.

When you automate everything, you don't become more strategic. You become disconnected. The patterns in client communication and the early signals that something is off live in the details. Outsource them, and you outsource your instincts.

I've seen this at Vangwe, where we build fintech and payments platforms across Latin America, North America and beyond. The real value we provide isn't in the code. It's in understanding what a client needs before they say it, reading regulatory environments and knowing when a project is going sideways. That requires presence, not automation.

A Framework: Own, Delegate, Automate

Over the past year, I've settled on a simple mental model for deciding what AI should and shouldn't touch:

• Automate what is repetitive, low-context and time-consuming: scheduling, data formatting, status aggregation and routine first drafts. Every hour reclaimed here is an hour for work that requires real judgment.

• Delegate work that benefits from AI but still needs human review: research synthesis, financial modeling, competitive analysis and drafting proposals. AI accelerates these significantly, but the output needs someone who understands the stakes. AI gets you to 80%; the last 20% is where your experience matters.

• Own whatever defines your value as a leader: client relationships, strategic decisions, team culture and ethical calls. Anywhere there's trust is what matters. If AI is making these decisions, you've abdicated, not optimized.

The mistake I see founders and executives make is treating this as a spectrum where they should push everything toward full automation. The goal isn't to minimize your involvement. It's to be intentional about where your involvement matters.

What Changes When You Get This Right

When AI handles the operational layer well, something unexpected happens: You become more human.

I spend more time with clients now, not because I have less to do, but because time I used to spend on logistics is now available for interactions that build trust. I review my team's work more carefully. I think more about strategy.

The value isn't just efficiency; it's where your attention goes.

The Question Every Leader Should Ask

In fintech, we talk about the "human-in-the-loop": the idea that automated systems still need human oversight. Every leader adopting AI should ask a version of this: If AI is doing everything, what am I actually contributing?

Boston Consulting Group found that accelerating AI is a top-three priority for 65% of CEOs. But your board, clients and team aren't paying for someone who monitors workflows. They're paying for judgment, relationships and the ability to make difficult calls under uncertainty.

AI is good at processing information, identifying patterns and executing defined tasks. It is not good at understanding why a client's tone shifted in a meeting, knowing when to push back on a regulatory interpretation or deciding which direction to take when data doesn't give a clear answer. Those are human skills. More valuable now, not less, precisely because everything around them is being automated.

Getting Started Without Losing The Plot

Start with tasks that drain you operationally but don't require your unique judgment. Audit where your time goes for a week. APQC research shows the average knowledge worker loses about 10 hours per week to managing communications, searching for information and unproductive meetings, a quarter of the workweek. Build your AI around how you actually work, not what tools can theoretically do.

Then protect what matters: the 20% of your work where your presence is irreplaceable. The goal isn't to free up your calendar. It's to fill it with work that only you can do.

The Real Shift

The right question isn't what AI can do. It's what you should still own.

The leaders who get the most from AI aren't the ones who automate the most. They're the most intentional about what they refuse to automate. They use AI to remove noise so they can focus on signals. They delegate execution so they can own judgment.

That's not a technology strategy. It's a leadership philosophy. And it will separate the leaders who use AI well from those who simply use AI.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?