How leaders can use AI as a thinking partner, analytical engine, and capacity multiplier — without surrendering judgment
When people talk about artificial intelligence in the workplace, the conversation almost always starts with automation: what task can it eliminate, what process can it speed up, how many hours can it save. Those are fair questions, but they understate the opportunity. In my own work leading an operations team, the most value I’ve gotten from AI hasn’t come from handing it my work. It’s come from building it into how I think through problems, test assumptions, and turn ideas into execution. Used that way, AI stops being just a productivity tool and becomes operational leverage.
Build a Partner, Not Just a Tool
General-purpose AI is designed to be agreeable — quick to help, quick to affirm, quick to tell you your idea has merit. That’s fine for everyday use. It’s a liability in leadership, where the last thing you need is another voice confirming what you already believe.
When I’m evaluating a process or a decision, I don’t want AI validating my first read of the situation. I want it arguing with me: what am I assuming, what evidence contradicts my position, what would have to be true for my conclusion to be wrong. That has to be built into the interaction deliberately. “I think this process is causing the problem — analyze the data” produces a very different conversation than “here’s my hypothesis, treat it as unproven, tell me what supports it, what contradicts it, and what I’m missing.” The second version doesn’t just get better wording — it gets a different kind of thinking partner, closer to how a good advisor works: you listen, you push back, you combine it with what you already know, and you still own the decision.
Turn Expertise Into Operating Logic
AI gets far more useful once real subject-matter expertise is built into how you use it. On a portfolio with several thousand open accounts, it would be simple to ask AI to sort by largest balance. It would also miss most of what actually matters — aging distribution, payment terms, unapplied cash, whether risk is concentrated or spreading, how the picture has shifted since the last review. The AI can process all of that quickly, but it didn’t decide which variables mattered. That judgment has to come from someone who understands the business. Once it is built into the instructions, the framework can be used consistently across a team instead of five people reviewing the same portfolio five different ways. That’s not automating the decision. It’s standardizing the discipline that leads to it. And once that framework exists, AI can also help design and pressure-test the workflow itself instead of simply being used after the process is already built.
Compress the Investigation Cycle
The clearest gains show up when a problem isn’t understood yet. Say an aging report shifts by several hundred thousand dollars in a week, with no obvious cause. Tracing that manually means pulling multiple reports, comparing periods, and checking individual accounts one at a time. I don’t ask AI “what happened” — I ask it to quantify the movement by category, separate ordinary aging progression from real deterioration, flag whether unapplied payments or credits could be distorting the number, and tell me which parts of its explanation it’s least sure about. Then I push on the answer: what doesn’t fit, what should I verify in the source system, what else could produce the same result. AI compresses the investigation. It doesn’t get to decide whether the explanation actually makes operational sense — that’s still on me.
Capacity, Not Just Efficiency
Saving twenty minutes on a task is useful. Making room for analysis that would otherwise never happen at all is a bigger deal. A supervisor can run a real portfolio review every week instead of occasionally. A manager can test three competing explanations before deciding, instead of stopping at the first plausible one. A new process can be documented and refined while it’s being built rather than six months later, once everyone’s forgotten why it was designed that way. None of that required more headcount. It required using the capacity that already existed differently.
AI Can Scale Bad Thinking, Too
This part is worth being honest about. AI doesn’t just make good thinking faster — it makes bad thinking faster and more convincing. Feed it a biased premise, incomplete data, or criteria built to confirm what you already wanted to believe, and it will hand you a polished, confident justification for the wrong decision. That’s exactly why expertise matters more with AI in the loop, not less. Someone still has to know the operation well enough to catch an answer that sounds right but isn’t, still has to validate the evidence, and still has to own what happens next. AI should sharpen judgment. It shouldn’t replace accountability for it.
The Real Advantage
I don’t think the organizations that gain the most from AI will be the ones that automate the most tasks. It will be the ones that figure out how to combine real operational judgment with machine-scale speed — where people bring context, accountability, and an understanding of what actually matters, and AI brings the ability to process complexity that would otherwise eat up days of management time. Each is useful on its own. The real advantage is in the system that connects them. Used well, AI doesn’t think for the leader. It gives the leader more capacity to think, test, challenge, and build. That is the difference between using AI as a productivity tool and using it as operational leverage.
