Leader reviewing AI-generated data before making a final decision

Leadership & Resilience

How AI Is Redefining Leadership: Why Human Judgement Still Matters Most

Resilience Institute · Weekly Resilience Brief · 5 minute read

AI genuinely improves decision-making, until leaders start trusting it too much. Automation bias, the tendency to accept an AI’s output even when it’s wrong, is a well-documented, decades-old risk now showing up across medicine, aviation, and increasingly, leadership.
A 2009 airline crash, later linked partly to over-reliance on cockpit automation, remains one of the starkest illustrations of what happens when a skilled human stops actively questioning a system.
Resilient leaders don’t reject AI. They stay self-aware enough to keep genuinely questioning it, which is precisely the skill automation bias erodes if left unchecked.

Artificial intelligence has been genuinely transformative, reshaping industries and changing how work gets done. For leaders, it brings real opportunity alongside a complex, easy-to-miss risk. AI can’t replace the emotional intelligence, adaptability, and judgement that define good leadership, but it can quietly erode a leader’s willingness to actually use them.

Table of Contents

How AI Is Actually Reshaping Leadership

AI is changing leadership in three clear ways. It analyses large volumes of data quickly, supporting faster, better-informed decisions. It automates routine tasks, from scheduling to performance tracking, freeing leaders to focus on strategy. And it surfaces predictive insight, flagging trends and risks that might otherwise go unnoticed. Tools now generate detailed reports and forecast shifts in market behaviour well beyond what a person could track manually, giving leaders real room to focus on long-term vision rather than operational grind.

The first item on that list, better decisions, deserves a closer, more honest look than it usually gets.

The Risk Hiding Inside “Better Decisions”

Automation bias, the tendency to accept an automated system’s output while discounting contradictory information, was first documented in the research literature in 1999, well before modern AI, and it hasn’t gone anywhere since. A 2026 international AI safety report notes it “undermines competence by discouraging active reasoning and verification,” weakening both the accuracy of a person’s judgement and their willingness to act independently of the system in front of them.

The effect isn’t hypothetical. In February 2009, Continental Connection Flight 3407 crashed near Buffalo, New York, an accident later linked in part to increased cockpit automation contributing to a loss of situational awareness among the flight crew. More recent research finds the same pattern in far lower-stakes settings: a 2025 study found that participants who held more positive general attitudes toward AI performed worse at distinguishing real from AI-generated content when given AI guidance, precisely because their trust in the tool discouraged genuine, independent scrutiny. Similar patterns have been documented in medical diagnostics, where clinicians using AI-based decision support have been found to accept incorrect recommendations without sufficient critical evaluation.

None of this means AI-assisted decisions are worse than unassisted ones. The research consistently finds real benefits from good human-AI collaboration. It means the benefit depends entirely on the human half of that collaboration staying genuinely engaged, not quietly deferring the moment a confident-looking output appears on screen.

Why Human Skills Are the Actual Safeguard

AI excels at processing scale. It’s the human capacity to stay genuinely engaged, sceptical, and present that determines whether that processing actually gets used well.

Emotional intelligence underpins trust, collaboration, and a team’s willingness to raise a genuine concern about a system’s output rather than deferring silently. One Forbes contributor piece put it directly: empathy and adaptability are vital for leaders navigating the AI shift, a view broadly consistent with the wider research on human-AI collaboration, even if that specific framing is one perspective among several rather than a settled consensus.

Adaptability matters because AI is a tool, not a finished solution. Integrating new systems, adjusting workflows, and navigating the genuine ethical questions AI raises all require a leader who can pivot without losing their footing.

Creativity and vision stay firmly human. AI optimises within existing parameters; it doesn’t set new ones. Visionary leadership is still what turns a good forecast into an actual decision worth making.

Resilience as the Framework for Staying Sharp

Resilience is what keeps a leader’s judgement genuinely intact alongside AI, not despite it. Self-awareness lets a leader notice when they’re deferring to a system rather than actually evaluating it. Emotional regulation keeps decision-making composed under pressure rather than rushed toward whatever answer arrives fastest. And agility in the face of uncertainty turns a wrong AI output, or a wrong human call, into a genuine learning moment rather than a repeated pattern.

Building This Deliberately

Build a Habit of Second-Guessing the Output

Before accepting an AI-generated recommendation on anything consequential, ask directly: what would make this wrong, and have I actually checked for it? This is the specific habit automation bias research shows erodes fastest under time pressure and repeated exposure, and it’s exactly the one worth protecting deliberately.

Invest in Emotional Intelligence Directly

Strengthen your capacity to empathise, connect, and read a room, capacities that don’t erode with automation and directly support a team’s willingness to speak up when something looks off.

Stay Genuinely Curious About How the Tools Work

Understanding roughly how an AI system reaches its output, not the underlying code, but its general logic and known limitations, makes it meaningfully easier to spot when it’s confidently wrong rather than accepting a plausible-looking answer at face value.

Use a Resilience Assessment to Find Your Own Blind Spots

Tools like the Resilience Assessment reveal where self-awareness, emotional regulation, and adaptability are genuinely strong versus quietly under strain, useful information well before those factors get tested by a high-stakes, AI-assisted decision.

AI is a powerful ally, not a substitute for judgement. The leaders who thrive won’t be the ones who trust it most. They’ll be the ones who stay resilient enough to keep questioning it.

FAQ

Frequently Asked Questions

Find answers to some of the most common questions about this topic.

What is automation bias?

Automation bias is the well-documented tendency to over-rely on an automated or AI system's output while discounting contradictory information, first studied in the research literature in 1999. It has been observed across aviation, healthcare, and increasingly general AI-assisted decision-making, and tends to worsen under time pressure or with repeated, positive prior experience with a system.

Does this mean AI-assisted decisions are actually worse than human-only decisions?

No. The research consistently shows real benefits from good human-AI collaboration. The risk isn't the AI itself, it's the human side of that collaboration disengaging and accepting outputs without genuine scrutiny. Well-managed collaboration outperforms either humans or AI working alone in most studied contexts.

How is this different from other resilience-and-AI content on this site?

This piece focuses specifically on decision-making quality and the psychological risk of over-trusting AI outputs. A related piece on this site addresses the broader workforce impact of AI, including job displacement data and the emotional experience of organisational change. The two are complementary rather than duplicative.