What Happens When You Skip Steps

Substitution (Step 1)

The Temptation

“Why waste time on simple replacements? Let’s build something that transforms our workflow from day one.”

What Goes Wrong

1. Nobody Trusts the Technology

People haven’t experienced AI working reliably on simple tasks, so they won’t trust it with complex, critical processes.

Example: A government agency tried to implement automated case prioritization (Step 3) without first letting caseworkers use AI for simple research tasks (Step 1). Result: Universal resistance. “How can we trust AI to prioritize cases when we’ve never seen it work correctly?”

2. Too Many Variables to Diagnose Problems

When complex systems fail (and they will during implementation), you can’t tell if the issue is:

• The AI model
• The prompt design
• The data integration
• User misunderstanding
• Process design flaws
• Security configurations

Example: A municipality tried to automate permit approvals immediately. When it gave incorrect recommendations, they couldn’t determine if the problem was the underlying data, the decision logic, or the AI model itself. The project stalled for months.

3. Fear Blocks Exploration

People need the psychological safety of “I can’t break this” before they’ll experiment and learn. Starting with high-stakes transformation creates anxiety, not innovation.

The Missing Foundation

Perceived Ease of Use isn’t established. People think: “This is complicated, risky, and scary.”

Without Step 1, you lack:

• ✗ User confidence
• ✗ Clear proof of value
• ✗ Low-risk learning environment
• ✗ Quick wins to build momentum

How to Recognize This Mistake

Warning signs you’re skipping Step 1:

• Your first use case involves automated decisions affecting people
• You’re integrating AI into mission-critical systems from day one
• The use case requires multiple system integrations to work
• Users are asking “but what if it gets it wrong?” with fear, not curiosity
• You can’t explain the value in one simple sentence

Augmentation (Step 2)

The Temptation

“We’ve proven AI works with simple searches. Now let’s automate entire processes.”

What Goes Wrong

1. No Evidence AI Outperforms Humans

You’ve shown AI can match human performance, but not exceed it. People aren’t motivated to adopt something that’s just “as good as” what they already do.

Example: A finance department proved AI could find contract clauses (Step 1), then immediately tried to automate approval workflows (Step 3). Staff asked: “Why should we trust the system to make decisions when we haven’t seen it do anything we couldn’t do ourselves—just faster?”

2. Missing the “Aha Moment”

Step 2 is where people discover AI’s true potential. It finds patterns they miss, processes volumes they can’t handle, and maintains consistency humans struggle with. Skip this, and AI remains “a fast search tool,” not “a genuine capability.”

3. Insufficient Organizational Buy-In

Transformation (Step 3) requires changing roles, processes, and often job descriptions. Without Step 2’s proof that AI genuinely improves outcomes, you won’t get the organizational support needed for those changes.

Example: A procurement team tried to redesign their vendor evaluation process around AI before proving AI could analyze contracts better than humans. Middle management blocked it: “We don’t have evidence this is actually better than our current process.”

The Missing Foundation

Perceived Usefulness only reaches “this is easier.” It hasn’t reached “this is better.”

Without Step 2, you lack:

• ✗ Proof of superior performance
• ✗ Compelling case for change
• ✗ Understanding of AI’s unique strengths
• ✗ Examples that inspire broader adoption

How to Recognize This Mistake

Warning signs you’re skipping Step 2:

• You can’t point to specific tasks where AI outperformed humans
• The proposed transformation is based on efficiency, not effectiveness
• You’re arguing for change based on “best practices,” not demonstrated results
• Users are compliant but not enthusiastic
• Leadership is asking “why can’t people just do this?”

Modification (Step 3)

The Temptation

“We don’t need to automate workflows. Let’s use AI to create entirely new services.”

What Goes Wrong

1. You’re Designing with Pre-AI Thinking

True innovation (Step 4) requires understanding how AI changes the fundamental nature of work. Without experiencing workflow transformation, you’re just adding AI to existing processes, not reimagining them.

Example: An HR department wanted to create an “AI-powered talent marketplace” connecting employees to opportunities. But they’d never transformed a single HR process with AI. They designed it like a traditional job board with AI features bolted on—missing the true potential because they didn’t understand AI-native workflows.

2. No Organizational Experience with AI at Scale

Creating new capabilities requires deep expertise in how AI performs under real conditions, with real users, in real processes. You don’t have that expertise if you’ve only run pilot projects.

3. Can’t Identify What’s Actually Possible

The best innovation opportunities emerge from seeing AI handle transformed workflows. You notice: “Wait, if AI handles this automatically, we could now do X, which was never possible before.”

Example: Katrineholm’s Step 4 innovations (procurement optimization, budget forecasting) only became visible after they’d experienced automated invoice monitoring (Step 3). The insights emerged from the transformed process.

The Missing Foundation

Organizational readiness for AI-native thinking hasn’t developed.

Without Step 3, you lack:

• ✗ Experience with automated, AI-driven workflows
• ✗ Understanding of AI’s operational capabilities
• ✗ Culture of trusting AI for critical processes
• ✗ Technical expertise in production AI systems

How to Recognize This Mistake

Warning signs you’re skipping Step 3:

• Your “innovation” is just AI added to current processes
• No one in the organization has redesigned their job around AI
• You can’t point to autonomous AI-driven workflows
• The proposed innovation could theoretically work without AI (just slower)
• You’re designing based on vendor demos, not lived experience

The Cumulative Nature of the Steps

Each step unlocks critical capabilities needed for the next:

Step Transition	Unlocks	Comment
Step 1 → Step 2	Willingness to experiment	People who’ve experienced easy wins become curious about what else is possible. Without Step 1, they’re too afraid to try Step 2’s more complex applications.
Step 2 → Step 3	Organizational mandate for change	When AI demonstrably outperforms humans, you gain the political capital and business case needed to redesign workflows. Without Step 2, you can’t justify the disruption Step 3 requires.
Step 3 → Step 4	Understanding of AI-native possibilities	Experiencing transformed workflows reveals opportunities invisible from the outside. Without Step 3, you’re guessing at innovation instead of designing based on evidence.

Common Questions

”Can’t we run multiple steps in parallel?”

Yes—for different use cases. You might have:

• Use Case A at Step 1 (new department starting)
• Use Case B at Step 2 (building on early success)
• Use Case C at Step 3 (original pilot transformed)

What doesn’t work: Trying to do multiple steps for the same use case simultaneously.

”We’re technically sophisticated. Can’t we move faster?”

Technical capability helps execute each step faster, but doesn’t let you skip the organizational learning. Even technically advanced teams need to build trust, prove value, and establish AI-native workflows.

---

Remember: The fastest way to transformation is to commit fully to each step. Trying to skip ahead doesn’t save time—it wastes it.