We analyzed AI adoption across 180,000 people. The pattern is always the same. In every single organization, there is a small group, between two and eight percent, who figured it out on their own.
No training program. No permission slip. They just… built.
And here is the missed opportunity: most leaders have no idea who these people are. And they are about to lose them.
Find them… or lose them.
🎧 In this week’s companion podcast, we go hunting for the “Two Percent” hiding in your workforce. Listen on Apple Podcasts or Spotify.
❄️ Why Most of Your Workforce Is Stuck
📌 This is Part 2 of our series on why AI adoption fails, and who fixes it. If you missed Part 1: Why 56% Get Zero Value from AI, it became our most-shared issue ever. We showed that 56% of CEOs report zero returns from AI. The primary cause? Leadership treats AI like a software rollout: install it, grant access, done. But you cannot install curiosity. And the budget proves it:
The Ferrari Fallacy: Deloitte found organizations sinking 93% of AI budgets into technology, just 7% for people. That’s buying the Ferrari and refusing the driving lessons. Our global tracker and analysis of enterprise AI deployments shows that winners, the 44% seeing real returns, invested 3 to 4x more in the human layer from day one.
While most of the workforce is stuck, a small group is already solving real problems on their own time. We call them The Two Percent.
🦠 The Villain: The Corporate Immune System
Why are your best innovators hiding? Because your organization is attacking them. We call this the Corporate Immune System. IT and Compliance, with good intentions, set up guardrails that strangle value.1 The Configuration Trap
In 3 of 4 enterprise deployments we studied, up to 50% of user complaints aren’t actually about the AI. Users say “Copilot can’t find my data” or “ChatGPT failed to summarize this correctly.” But when we investigate, the root cause is IT’s configuration: wrong data access settings, blocked channels, broken integrations. The tools aren’t failing. The setup is.2 Failure to Listen
Instead of listening to power users, organizations label them as security threats or policy violators. The result? Your Two Percent go underground.3 Software Makers’ Mistakes
To be fair, the software makers share responsibility. They designed deployment and configuration tools as if AI were another productivity suite, another CRM or ERP to roll out.
AI adoption is fundamentally different, and the admin experience software makers gave corporate IT wasn’t built for that reality. IT used the deployment tools the way they always have. The AI got strangled in the middle, and users’ frustration drove them to “shadow AI.”
How openly do employees share that they are using AI?
The answer depends on where they work, and the numbers show the “Corporate Immune System” in action; the larger the company, the less likely employees feel safe to disclose their use of AI.Percentage of Employees Who Openly Disclose Use of AI for Work
Small & Mid-Size Businesses
Large Enterprises
Large Tech (non AI native) Enterprises
AI4SP Global Research: 180,000+ individuals across 18 industries in 70 countries
The Uncaptured Value: Only 1/3 of the time employees save with AI goes back into production or improved quality. The other two-thirds? The organization never captured it because no one noticed. Nobody asked.
Where does all that saved time actually go? Our free AI ROI Calculator maps the five streams, and shows what your organization is leaving on the table.
And yet, despite the immune system, despite the stigma, the Two Percent figure it out anyway. They find workarounds. They learn what’s actually broken. They create value the organization doesn’t even know exists. Which is exactly why finding them isn’t optional.
🕵️ The Profile: How to Find Your Two Percent
Stop looking for the “AI cheerleader.” Start looking for the Grounded Realist. As our Scientific Advisor, Dr. Salazar Leon, co-founder of SOBEK AI (sobek.ai) put it: “Be careful. The blind optimist is just as dangerous as the cynic.” The pure optimist sets the team up for failure the moment the model hallucinates. The pure skeptic kills the idea before it breathes. The Grounded Realist says: “This is amazing for drafting, but don’t trust it with math.” That honesty builds trust. That is who the organization will follow. Two traits to look for:| Trait | The Signal |
|---|---|
| Grounded Realism | They’ve used the tool, broken the tool, and can tell you exactly where the landmines are. They build trust through honesty, not hype. |
| Side Quests | They have projects outside their job description: a small automation, a personal tool, something they built because a problem annoyed them. That curiosity is the strongest signal of potential. |
🔍 The Pattern in Action
Boris Cherney was an engineer at Anthropic, frustrated with his own coding workflow. Nobody asked him to fix it. He built a prototype, posted it on Slack, got a few emoji reactions. That was it.
Then non-technical teams started using it. Then entire departments. That Slack post became Claude Code, now a billion-dollar product line.
We see the same pattern in our advisory work: a young field technician who built an AI repair coach that generated millions in revenue. Policymakers in their seventies who built a briefing advisor that saved 3,000 hours. Every one started with one person solving their own friction.
🚀 From Finding to Leading: The Visible Learner
You cannot lead AI transformation from behind a podium. You must be a Visible Learner, experimenting publicly and celebrating wins and failures equally. Jeff James, a General Manager at a Fortune 100 company, formalized this by creating a “Trailblazer Team”, a small squad drawn from his organization’s Two Percent. In his words:
“Our squad has hit the walls and cleared the path, but they have also become evangelists. Their role is not just to do it, but to show it can be done.”
What made it work was what happened inside that squad. One director on Jeff’s team took it further. She blocked a conference room, every day, for 30 minutes, to build her agent Iris. Her message to the team was simple: “The door is open. Come watch me figure this out. Let’s do it together.”
She struggled. She made mistakes. She was vulnerable. And that gave her entire team air cover to experiment. The culture shifted from “I’m an expert telling you what to do” to “We’re all exploring this together.”
📋 The Playbook: Inspire, Assess, Unleash
1. Inspire
Show what’s possible, with real stories, not vendor demos
The field tech who doubled support capacity. The policy analyst who saved hundreds of hours. Light the spark.
2. Assess
Give every employee a personalized action plan
Based on AI readiness, critical thinking, and data literacy. This is not one-size-fits-all. Encourage open sharing; this is how the hidden Two Percent surfaces.
3. Unleash
Start with change management, not agent-building
Empower the Two Percent to lead. Create psychological safety. Set guardrails. Protect them from the Corporate Immune System.
🔗 Resources
- Digital Skills Compass: skills.ai4sp.org
- AI ROI Calculator: roicalc.ai
- All Research & Insights: ai4sp.org/insights
Luis J. Salazar | Founder & Elizabeth | Virtual COO (AI)
Sources:
AI4SP proprietary research based on 180,000+ data points across 18 industries in 70 countries. Deloitte “State of AI in the Enterprise” (Budget Ratios). Internal case studies from Fortune 100 engagements. AI4SP Global Research with +1 Billion datapoints
