The first time an AI system made a recommendation I disagreed with, I felt something I didn’t expect: ego.
It was a resource allocation call on a complex project. The model had crunched variables no person could hold in their head at once, and its recommendation was better than mine. Something in me didn’t want to let a model make that call. So, I made it myself.
I told this story to Michael Jabbour, AI innovation officer at Microsoft, expecting him to talk about the technology. Instead, he talked about authority. His point was that every new artificial intelligence tool changes who gets to decide, rather than what gets decided.
Since then, I keep coming back to one question: If intelligence doesn’t need a human body, what exactly makes someone a leader?
I’ve talked with executives, technologists, and researchers about this, including Jabbour and Avantika Sharma, global head of healthcare at Brillio, an enterprise AI company. The pattern I keep hearing is the same, and it’s that leadership is moving from directing people to designing systems. The job now is defining how humans and machines think together.
That sounds philosophical, but in practice it comes down to very specific leadership choices. Here are the shifts I’ve had to make in my own leadership:
1. STOP MANAGING TASKS AND START DECISION SYSTEMS
Early in my career, I thought leadership meant being the smartest person in the room. I prepared harder than everyone else and made the decision before others. But that doesn’t work once intelligent systems become part of the workflow. AI surfaces patterns and flags risks faster than humans can, so we won’t beat it on speed. But we can control how those insights get used.
Sharma showed me one version of what that looks like in a regulated industry like healthcare. Her approach is driven by risk, and she draws a hard line. Compliance, data governance, transparency, and operational reliability aren’t up for debate, but she steps back on how a solution gets designed. Her job is to frame the problem, agree on outcomes, and set the guardrails.
That distinction has changed how I lead. These days, I spend less time obsessing over who does what and more time asking: What are the guardrails? What outcomes matter most? Where does judgment have to remain human?
If you’re mapping this for your own team, start by looking at the decisions you make repeatedly. Which ones are structured, repeatable, and data-heavy? Which ones carry ethical, reputational, or human consequences? Let intelligent systems inform the first group, and keep humans accountable for the second. This way, leadership becomes less about supervision and more about architecture.
2. RETHINK THE ORG CHART
Most org charts put intelligence at the top and let authority flow down from there. That logic fails when an AI system is generating insights that shape strategy and coordinate work across functions faster than the executive team.
I brought up this tension to Jabbour. He pushed me on whether we’re clinging to hierarchy out of habit more than logic. If intelligence is distributed across humans and machines, positional authority stops being the whole story.
That raises a different set of leadership questions. Who sets the rules? Who audits the system? Who decides when to override it? And when something goes wrong, who ultimately owns the decision?
In practice, that looks like building clear paths for escalation and intervention. The reporting lines on the chart might not change, but what sits behind them does.
3. SEPARATE SPEED FROM RESPONSIBILITY
AI moves faster than any team you’ll ever build, and that speed is easy to mistake for authority. Sharma is explicit about keeping the two apart. A system can prioritize and spot patterns faster than a person, but accountability for judgment and validation still belongs to a human.
I learned this firsthand. My team once let an automated recommendation engine drive a sequence of operational decisions without anyone clearly owning edge-case validation. It worked fine until it hit a scenario it had never seen, and then nobody was sure who owned the call.
Now, before I turn on any AI-supported process, I name two roles out loud. First is who watches the outputs. Second is who owns what happens after someone acts on them. Sometimes it’s one person, but it’s never the algorithm. The line I return to now is, “The system recommends; I decide.”
4. ASK WHAT THE SYSTEM IS OPTIMIZING FOR
We still talk about AI like a piece of technology you deploy and upgrade. In reality, it learns from what you feed it and adapts over time, which raises the question I keep sitting with: If intelligence just means processing information and making decisions from it, does it have to be biological? Drop that assumption, and intelligence becomes a property that systems can have, too.
Once you think of AI that way, leadership starts to look more like stewardship. You are no longer simply adopting a tool, but are instead introducing a system that encodes assumptions about what matters, what counts as success, and what risk is acceptable.
Before I bring any new system into a workflow, I ask what it’s optimizing for and what it might be ignoring in the process. Human nuance is what usually gets dropped first, mostly because it’s harder to measure than efficiency. If you get handed a new AI dashboard or assistant, learn the interface, but also ask what the tool rewards and whether it pushes you toward speed at the expense of depth, quality, or relationships.
5. DEFINE WHAT HAS TO STAY HUMAN
As machine judgment gets embedded everywhere, it’s easy to assume every decision can be split between human and machine. However, some conversations can’t be automated, like feedback that affects someone’s career or a decision that touches a patient’s care. In those moments, leadership means showing up in person. A message routed through a system doesn’t carry the same weight.
The rule I use with my team now is that we automate analysis, not accountability. If you’re stepping into leadership, write your own version of that line down and put it somewhere you and your team will see it again and again.
The challenge in leadership used to be knowing more than everyone else in the room. Now it’s knowing exactly what you’re willing to hand off and holding the line on what you won’t.
