There’s a growing divide between leaders who avoid AI entirely and leaders trying to use it to replace judgment, feedback and decision-making outright. Both extremes miss the point. AI shouldn’t replace leadership. It should help leaders lead better, and the difference between those two things matters more than it might sound.
The Real Problem AI Can Help With
Leadership already comes with a heavy amount of context switching, decision pressure and cognitive load, and that load only grows as the number of people and systems someone is responsible for increases. Holding all of that consistently in your head is genuinely hard, and that’s the real problem worth solving, not the mechanics of leadership itself.
Used well, AI becomes an enhancement layer on top of a leadership system that’s already strong, not a replacement for the judgment inside it. It can help identify patterns a leader might otherwise miss, reduce recency bias during reviews and coaching conversations, synthesize large amounts of information quickly, improve preparation for difficult conversations, spot operational friction across teams and create more consistency in how a leader shows up for the people they support. None of that replaces the human part of the job. AI doesn’t build trust, doesn’t understand nuance the way a person does and doesn’t own accountability for a decision. The leader still owns the judgment, the empathy and the relationships. AI just helps them operate with more awareness inside a system that’s gotten more complex than any one person can hold perfectly in memory.
Why Context Is the Hardest Part of Leadership
One of the hardest parts of leadership is holding context consistently over time, not just for projects and systems, but for people. As teams grow, leaders are expected to track an enormous amount of information across delivery, performance, communication, team dynamics, career growth and shifting priorities, and most of it changes constantly. A good leader still needs to know their people directly: build trust, understand nuance and make the hard calls themselves. What AI can do is help synthesize the patterns that are genuinely difficult to hold onto over long stretches of time: recurring themes across 1:1s, repeated friction points, growth trends and patterns that suggest someone may need additional support.
Most leadership mistakes don’t happen because people don’t care. They happen because leaders are overloaded and operating with incomplete context. Used well, AI can reduce some of that load so a leader has more energy left for the parts of the job that actually require them to be human.
Recency Bias Is Quietly Expensive
Recency bias damages more teams than most leaders realize, not because leaders are careless, but because people naturally overweight whatever happened most recently. A difficult sprint right before a performance review can suddenly become “how things have been going.” A strong recovery can erase months of real struggle. Someone who communicates well in the room can get remembered more positively than someone who’s been quietly delivering behind the scenes the whole time, and the longer the review period, the more this effect compounds.
This is one of the places AI can genuinely improve leadership, not by writing reviews or making decisions about people, but by helping a leader see patterns across time instead of reacting primarily to whatever is freshest in memory: consistent feedback themes, long-term growth trends, repeated delivery friction and the real improvements that might otherwise get forgotten in the moment. Good leadership still requires judgment and direct relationships. Stronger systems just help a leader apply those things more consistently instead of unevenly.
AI Is an Overlay, Not an Operating System
There’s a real push right now to use AI to automate more of leadership itself: writing reviews, generating feedback, summarizing people and even recommending decisions. The problem is that weak leadership systems don’t become strong just because AI gets layered on top of them. They usually become faster versions of the same underlying problems. If a leader avoids difficult conversations, AI won’t fix that. If ownership is unclear on a team, AI won’t fix that either. If feedback already lacks honesty or trust, adding a tool on top doesn’t change any of it.
Strong leadership still requires judgment, accountability, empathy, consistency and a direct understanding of the people and teams involved. What AI can do is strengthen leaders who already care about those things, by helping them organize information, surface patterns and prepare more consistently under pressure. AI should sit on top of a strong leadership operating system, the same way good engineering tooling improves a team’s output without replacing engineering judgment itself. Used well, it creates real leverage. Used poorly, it just accelerates whatever dysfunction was already there.
How This Shows Up in the Work I Do
Inside the Engineering Manager Mentorship Program, this shows up as a structured, deliberately bounded practice rather than a loose suggestion to “try using AI more.” It means using AI to reduce recency bias in review and 1:1 preparation and to surface real execution signals a manager might otherwise miss, without AI making performance decisions or being used to justify disciplinary or termination decisions. The manager stays accountable for the interpretation and the judgment every time. That boundary isn’t a formality. It’s the entire point.