Part 3: Keeping Humans at the Center

The Difference Between Augmentation and Replacement

There's a moment in every AI implementation where you face a choice: Do we use this to amplify what humans do well, or do we use it to replace them entirely?

Most organizations don't even realize they're making this choice. They drift toward replacement by default, following vendor promises and cost-cutting instincts. But there's another way—one that recognizes a fundamental truth: The magic happens when humans and AI work together, not when one replaces the other.

This week, let's explore how to keep humans at the center of your AI strategy, not just in philosophy but in practice.

The Automated Strike Zone Disaster

In 2019, the Atlantic League started testing an automated ball-strike system. A computer called balls and strikes, relaying the call to an umpire through an earpiece. The umpire would then make the call. Seemed perfect—computer accuracy with human presence.

Here's what actually happened:

For the first 50 pitches, umpires were engaged, ready to override obvious errors. By pitch 100, they were mentally checked out. By pitch 200, they were essentially unconscious robots repeating whatever the computer said.

Then the system would glitch. The umpire, completely disengaged after hundreds of automatic calls, couldn't snap back into decision mode. Accuracy plummeted. Games devolved into arguments.

The lesson? Humans can't maintain readiness without engagement.

Leagues are now using a challenge system. Humans make every call. Computers check disputed ones. Engagement stays high. Accuracy improves. Everyone wins.

The Automation Paradox

This baseball story illustrates a broader truth: The more you automate, the less capable humans become at handling exceptions—which are exactly the situations where you need human judgment most.

This paradox appears everywhere:

• Pilots who can't fly when autopilot fails

• Drivers who can't navigate without GPS

• Customer service reps who can't help when the script doesn't apply

The solution isn't less automation. It's smarter automation that keeps humans engaged, capable, and ready.

Designing for Human + AI Collaboration

The 70/30 Rule

We've found that the sweet spot for most tasks is about 70% AI, 30% human. Enough automation to eliminate drudgery, enough human involvement to maintain engagement and quality.

Examples from our implementation:

Email Writing: 70% AI, 30% Human

• AI drafts based on context and templates

• Human reviews, adds personal touches, ensures accuracy

• Result: 50% time savings, better quality than either alone

Troubleshooting: 60% AI, 40% Human

• AI suggests diagnostic steps and solutions

• Human applies context, handles edge cases, manages relationships

• Result: 30% faster resolution, higher customer satisfaction

Documentation: 80% AI, 20% Human

• AI captures details and formats reports

• Human verifies accuracy and adds insights

• Result: 40% time savings, more consistent quality

The Amplification Framework

Instead of asking "What can AI do?" ask "What are humans uniquely good at?" Then use AI to handle everything else.

Humans excel at:

• Building relationships

• Understanding context

• Making ethical judgments

• Creative problem-solving

• Handling exceptions

• Reading emotions

• Adapting to unexpected situations

AI excels at:

• Processing large amounts of data

• Following consistent patterns

• Remembering everything

• Never getting tired

• Calculating probabilities

• Finding patterns

• Generating first drafts

Design your implementations to combine these strengths, not pit them against each other.

The Empowerment Approach

From Support to Superpower

Our technicians used to spend hours every day on documentation. They hated it. It drained their energy and job satisfaction. Now AI handles the first draft, and they spend much less of their time reviewing and adding insights.

What happened to those other hours of time?

• More time actually solving problems

• More time talking with clients

• More time learning new skills

• More time mentoring others

They didn't become less valuable. They became more valuable doing uniquely human work.

The Innovation Invitation

Every employee should be able to answer: "How could AI make your job better?"

Not "How could AI do your job?" but "How could AI make YOUR job better?"

We created two groups:

1. The AI Task Force: Responsible for policy, security, major implementations

2. The AI Explorers: Anyone interested in learning and experimenting

The Explorers have generated our best ideas. Why? They know the pain points. They know what would actually help.

The Bottom-Up Revolution

Traditional IT implementation: Top-down, standardized, mandatory Human-centered

AI implementation: Bottom-up, customized, voluntary

A customer service rep built a tone analyzer that helps him match customer emotional states. He wasn't asked to. He wasn't trained to. He just knew what would help him do better work.

That tool is now used across the department. But it started with one human solving their own problem.

The Skill Evolution Strategy

The Retraining Imperative

When AI eliminates tasks, you have two choices:

1. Eliminate the people doing those tasks

2. Train them for higher-value work

IKEA chose option 2. When AI handled basic customer service, they retrained 8,500 call center workers as design consultants. Now customers get instant answers to simple questions and rich, valuable conversations about design.

Our Retraining Approach

When AI automated routine troubleshooting:

• Junior technicians learned advanced diagnostics

• Senior technicians became AI trainers and innovators

• Everyone learned prompt engineering

• Documentation specialists became quality analysts

Nobody lost their job. Everybody gained skills.

The Learning Loop

Human-centered AI creates a virtuous cycle:

1. Humans identify problems

2. Humans + AI create solutions

3. Humans evaluate and refine

4. Humans learn and grow

5. Return to step 1 with more capability

Each iteration makes both humans and AI more valuable.

Practical Techniques for Human-Centered Implementation

The Human Override Principle

Every AI system needs a human override. Not buried in settings. Not requiring manager approval. Right there, obvious, immediate.

Why? Because the moment people feel trapped by AI decisions, trust evaporates.

The Explanation Requirement

AI shouldn't just give answers. It should explain its reasoning. This:

• Helps humans learn

• Builds confidence

• Enables error detection

• Maintains human judgment skills

  
  

Our troubleshooting AI doesn't just say "Replace the network card." It explains why it thinks that's the solution, what else it considered, and what to check if that doesn't work.

The Collaborative Interface

Design interfaces that position AI as a colleague, not a replacement:

Bad: "AI Response: [answer]"

Good: "AI Suggestion: [answer] - Does this match your experience?"

Bad: Automatic actions without confirmation

Good: "I'm ready to [action]. Should I proceed?"

Bad: Binary correct/incorrect

Good: Confidence levels and alternatives

The Dignity Principle

Never use AI in ways that diminish human dignity:

• Don't surveil without consent

• Don't automate performance reviews

• Don't replace human interaction in sensitive situations

• Don't pretend AI output is human-created

Measuring Human-Centered Success

Traditional metrics focus on efficiency. Human-centered metrics include:

Engagement Metrics:

• Are people using AI tools voluntarily?

• Are they creating their own use cases?

• Is adoption spreading organically?

Satisfaction Metrics:

• Job satisfaction scores

• Sense of empowerment

• Feeling valued

• Work-life balance

Growth Metrics:

• New skills developed

• Problems solved creatively

• Innovation rate

• Career advancement

Quality Metrics:

• Customer satisfaction

• Error rates

• Edge case handling

• Relationship strength

The efficiency gains are nice. The human gains are transformative.

The Resistance You'll Face

From Management: "Why not just replace them entirely? It's cheaper."

Response: Show them Klarna's reversal. Calculate the cost of lost knowledge, hiring, retraining. Factor in innovation loss. The math supports human-centered approaches.

From Employees: "This is just the first step toward replacement."

Response: Prove it's not through action. Make public commitments. Celebrate automations that employees create. Show career growth paths.

From Vendors: "Our AI can do everything humans can do."

Response: Ask about edge cases. Ask about relationship building. Ask about creative problem-solving. Watch them squirm.

The Hard Truths

1. It's More Complex: Human + AI systems are harder to design than pure automation

2. It's More Expensive Initially: Training and transition cost more than replacement

3. It's Slower to Implement: Building trust and capability takes time

4. It Requires Constant Evolution: Human needs change, systems must adapt

But it's also more:

• Sustainable

• Innovative

• Resilient

• Valuable

• Human

Your Human-Centered Checklist

Before implementing any AI system, ask:

• Does this amplify human capability or replace it?

• Will humans remain engaged and growing?

• Can humans override AI decisions easily?

• Are we measuring human outcomes, not just efficiency?

• Does this preserve human dignity and agency?

• Will this make jobs better or just faster?

The Path Forward

Keeping humans at the center isn't a technical challenge—it's a design philosophy. It requires constantly asking: "How does this serve the humans who use it?"

Next week, in Part 4, we'll explore how to ensure every AI implementation is purposeful—solving real problems that real people actually have, not just implementing technology for its own sake.

But remember: The goal isn't to use AI. The goal is to make work better for humans.

AI is just a tool to get there.

Keep the humans at the center, and the rest falls into place.