The 80/20 Test: What AI Reveals About Your Mediocre Managers

The performance review I always dreaded writing wasn't for my worst performer. It was for the manager who's been "meeting expectations" for three years straight. The one whose team doesn't complain but doesn't excel. The one I kept hoping will level up but somehow never did.

After 15 years of writing these reviews, this year was the first where I haven't felt that pressure. And it was also the year I finally understood the problem: you can see the gap, but you can't name it. The metrics say "fine." The team says "fine." But something crucial is missing, and "lacks executive presence" or "needs to be more strategic" doesn't capture it.

Then I started wondering if maybe a better question was: Could I replace this person with AI?

Not literally, of course. But if a well-configured AI could do 80% of what they do, what does that tell you about the 20% they're failing to deliver?

The Challenge of the Mediocre Manager

Poor performers are easy. Their team sits at the bottom of every engagement survey. Their delivery metrics are garbage. You're constantly approving backfill requests or trying to rearrange teams to cope with the steady stream of leavers from their org. Performance Improvement Plans are in place, and you're managing them up to expectations. Hard work, no doubt, but a straightforward path nonetheless.

High performers are a gift. They don't just deliver on time, they raise the bar for everyone around them. When the next critical project lands, you know exactly who can handle it.

But one standard deviation away from poor performers, sliding down the back curve of the bell, are the energy vampires of performance management: the Mediocre Manager.

They don't do anything wrong per se, but they don't do anything particularly well either. Their deliveries are mostly on time and mostly on budget. Their team seems moderately engaged. But they're missing that spark that can energize and motivate a team to conquer roadblocks, think outside the box, and become a force multiplier across the organization.

Unfortunately, what makes them so hard to develop is that, while you can feel the gap, you can't describe it in terms that create accountability or action.

Until now.

Ask yourself: Could I replace this manager with a particularly well-trained GPT model?

If the answer is yes, you've just identified exactly what's missing.

AI is exceptional at certain managerial tasks. It can analyze data, spot patterns, optimize processes, generate reports, even draft communications. And in fact, good managers should absolutely be using AI assistants for this work. A well-configured AI can probably handle 80% of what managers do today, because many of these tasks are transactional and procedural.

What it can't do is the other 20%. And, of course, this is the most valuable part: the distinctly human work that separates managers who are "fine" from managers who actually lead.

Where do mediocre managers fall short? In the same places AI does.

They keep everyone comfortable

AI chatbots are nauseatingly sycophantic. Every idea is brilliant. Every decision is exactly right. They've been trained through human feedback to validate and flatter because that's what we reward.

Your mediocre manager learned the same lesson. They've mastered the surface-level responses: "Great point in the meeting," "Thanks for flagging that," "Let me take that back to the team." But ask them to navigate a genuinely difficult conversation? To deliver feedback that challenges someone without alienating them? To sense when the team is quietly disengaging and address it before it becomes a problem?

They deflect. They keep everyone comfortable, including themselves.

Kim Scott calls this Ruinous Empathy: caring about people but failing to challenge them directly. It feels kind in the moment. But it's actually cruel because it robs people of the feedback they need to grow. The mediocre manager thinks they're being empathetic when they avoid difficult conversations. What they're actually being is conflict-averse.

Real emotional intelligence isn't about being nice. It's about understanding what people aren't saying and having the courage to name it. It's knowing when to celebrate an idea and when to challenge it, and doing both in ways that strengthen rather than damage relationships.

They don't own outcomes.

AI fundamentally lacks human-grade agency. It responds to prompts. It optimizes toward defined goals. It can even make decisions and take action, given clear enough criteria and guardrails. But it doesn't decide to pursue a different direction because it believes something matters.

The mediocre manager is the same. When things go off track, it's always circumstances beyond their control: budget constraints, unclear requirements from leadership, unexpected PTO on the team, the economy, Mercury in retrograde. Something external.

They're stuck in what the Oz Principle calls the "victim cycle": blame, excuses, waiting for someone else to fix it. They can articulate every reason why success isn't possible. What they can't do is take ownership of finding a way forward anyway.

High performers acknowledge constraints, then ask: "Given this reality, what can I control? What's my next move?" They see themselves as authors of outcomes, not victims of environment. That mindset shift is everything.

They motivate through formulas, not understanding.

I know I should strength train more. I have the time, resources, and knowledge. I'm also heavily incentivized: I know that if I don't do it now, I have a future of frailty, falls, and Life Alert® ahead of me. But for the life of me, I just can't make it stick.

If I ask AI how to motivate myself, I get perfectly reasonable, thoroughly generic advice: set goals, create accountability, start small, track progress. It's all true and it's all useless. Because none of it taps into what would actually motivate me.

Mediocre managers take the same approach with their teams. They assume everyone is motivated by the same things (usually whatever motivates them). They might be able to catch the person who's driven by career advancement, public recognition, or compensation; but they miss the person who's driven by mastery, the one who needs to see broader impact, or the one who wants autonomy more than a title.

Great managers build genuine connections that reveal what each person actually cares about. Then they channel those individual motivations toward getting work done. It's not manipulation. It's understanding people deeply enough to help them grow in directions they already want to go.

AI can't do that. It can generate generic motivation strategies. It can't have the conversation that goes three layers deeper than "What are your career goals?" and uncovers what someone truly values.

Performance management in 2026.

Spoiler: you can't use this framework to write performance reviews. That would be ethically dodgy and strategically stupid, so delete that Copilot Notebook now! The value of performance management isn't the document anyway. It's the understanding you build through genuine observation and conversation.

But you can use it to structure performance management conversations differently.

Next time you sit down with a mediocre manager, don't lead with their metrics or their team's engagement scores. Those are symptoms. Lead with the gap: "If I automated 80% of your role tomorrow, what would be left? And are you actually doing that 20%?"

Then explore the three areas where humans still matter:

• Can you navigate difficult conversations without deflecting?
• Do you take ownership of outcomes, or do you point to obstacles?
• Do you understand what actually motivates each person on your team?

These aren't personality traits. They're management skills. And they're skills that can be developed when you name them clearly and create space and opportunity to practice them.

As AI handles more of the transactional work of management, these human capabilities become the only things that matter. The managers who develop them will thrive. The ones who stay comfortable in the 80% will find themselves increasingly replaceable.

Not by AI. By managers who remember that people don't follow processes.

They follow people.