The Two-Speed Workforce: A Diagnostic Card for Department Heads Managing AI Adoption

Executive Summary

• 45% of CEOs report that most of their employees are resistant or openly hostile to AI. The other 55% aren’t managing a united workforce — they’re managing a divided one. Both situations require the same underlying skill: segmenting the team and taking differentiated action.
• The gap between enthusiastic and resistant employees is not a personality difference. It’s a signal about information, identity, and control. Enthusiasts have clarity; resistors don’t.
• Unmanaged enthusiasm is as risky as unmanaged resistance. The 54% of employees who admit using unsanctioned AI tools are concentrated in the enthusiast group. Left without a channel, they become a data governance liability.
• Companies with a fully implemented change management strategy for AI are three times more likely to achieve measurable results than those without one (Kyndryl/KPMG, 2025, n=1,000+ executives). The strategy doesn’t require a consultant. It requires a department head who can tell the difference between the two groups and act accordingly.
• This card gives you the three diagnostic questions, the two actions for enthusiasts, and the two actions for resistors. It takes 20 minutes to apply.

The Two Groups You Actually Have

Most managers describe their team as “mixed” on AI. That’s not wrong, but it’s not useful. The research identifies a more actionable split. BCG’s 2025 workforce analysis (n=11,000+ global employees) describes five adoption personas, but for a department head they collapse into two operating categories:

Group A — Energized (typically 30-45% of a team): These employees are already using AI, already curious, and often already frustrated that the organization isn’t moving faster. Some are using tools you haven’t approved. They see AI as extending their capability, not threatening their identity. If you ignore them, they go underground.

Group B — Skeptical (typically 40-55% of a team): These employees are not stupid or backward. BCG characterizes them as “experienced and capable” — they are wary of AI because they’ve seen technology fads burn their team before, because they’ve built expertise that AI appears to devalue, or because nobody has explained what “AI adoption” means for their specific role. A 2025 analysis of 500,000+ Reddit narratives found the dominant employee concern is not job loss in the abstract — it’s identity erosion: the fear that the expertise they’ve spent years building is being quietly rendered irrelevant.

The 10% in neither group — genuinely indifferent — will follow whoever is more visible.

Three Diagnostic Questions

Ask these about each person on your team. The answers take less than five minutes per person if you know your team.

1. Has this person used any AI tool — sanctioned or not — in the past 30 days?

• Yes: Group A (likely)
• No: Group B (likely)

2. When AI comes up in team conversations, does this person lean forward or go quiet?

• Leans forward / asks questions: Group A
• Goes quiet / redirects: Group B

3. When asked to describe their role, do they describe it in terms of judgment and expertise, or in terms of process and output?

• Judgment/expertise: Higher risk of identity-based resistance (a subset of Group B that needs specific handling — see below)
• Process/output: More likely to see AI as a tool improvement; default to Group A or neutral

These three questions are not a performance assessment. A Group B employee is not a problem. A Group B employee whose concerns are ignored becomes one.

Two Actions for Group A (The Energized)

Action 1: Channel their energy into sanctioned experimentation.

Enthusiastic employees using unsanctioned AI tools represent the highest data governance risk in most mid-market companies. The 2025 State of Shadow AI Report (Reco, n=enterprise organizations) found 86% of companies lack visibility into how data flows to and from AI tools — and 48% of employees say they would continue using AI tools even if banned. Prohibition doesn’t work with this group. Redirection does.

Give each enthusiast a specific charge: “You are testing [sanctioned tool X] for [specific use case Y]. Report back in 30 days with three things it does well and one thing it doesn’t.” This costs you nothing, surfaces real workflow intelligence, and removes the incentive to go rogue.

Action 2: Make them visible to skeptics — carefully.

Group A employees who are seen using AI and succeeding are the most effective change agents in your organization. Leadership modeling shifts employee AI sentiment from 15% positive to 55% positive — a 3.7× multiplier (BCG, 2025). But unstructured evangelism backfires. An enthusiast who tells a skeptical colleague “this is amazing, you have to try it” without acknowledging the colleague’s concerns creates backlash, not adoption.

Structured peer mentorship — 30-minute office hours, not all-hands announcements — is what moves the needle. Schedule them.

Two Actions for Group B (The Skeptical)

Action 1: Name the real fear — and address it specifically.

The number one mistake managers make with resistant employees is treating AI resistance as a skills gap. Sometimes it is. More often, it’s an identity threat. For a professional who has spent 15 years building expertise in their craft, being handed an AI tool can feel like a verdict: “your judgment is replaceable.”

This requires a different conversation than “here’s how to use the tool.” It requires: “Here’s what this tool does that isn’t your job, and here’s what I still need you to do that it can’t.” The specificity matters. “AI handles first-draft synthesis so you can spend more time on the judgment calls that clients pay us for” is a real answer. “AI will help you work smarter” is not.

HBR’s 2025 analysis of 1,454 documented workplace AI narratives identified seven recurring anxiety themes — the one that dominates experienced workers is “skill atrophy”: the fear that using AI will erode the very capabilities that make them valuable. Acknowledge it. It’s a reasonable concern. The evidence suggests it’s real for tasks requiring deep expertise, which is exactly why you need their judgment applied to the right work.

Action 2: Create a small, reversible experiment with an opt-out.

The highest-converting on-ramp for skeptical employees is a bounded experiment with explicit permission to stop. “Try this for two weeks on [specific low-stakes task]. If it doesn’t help, you don’t have to use it.” This reduces the perceived stakes from “my role is changing” to “I’m testing a new approach.”

PwC’s 2025 Global Workforce Survey (n=56,000 workers, 48 economies) documents that employees in organizations where psychological safety around experimentation is explicitly created are significantly more likely to convert from skeptics to users. The key phrase is “safe to try” — not “expected to adopt.” The distinction is how it’s framed by you, the manager.

Key Data Points

What This Means for Your Organization

The two-speed workforce is not a problem to solve — it’s a structural condition to manage. Every team with more than ten people has both groups right now, regardless of industry, seniority, or how good your AI training program is. The companies that capture AI value are not the ones that eliminate resistance. They’re the ones that channel enthusiasm productively and convert skeptics with specificity rather than pressure.

The practical limit for most department heads is time. If you have 15 direct reports, the full diagnostic takes about two hours and produces a one-page segmentation you can act on immediately. The four actions above — two for each group — are what good managers are already doing informally. This card makes it explicit and repeatable.

If your team has dynamics that don’t fit cleanly into this two-group model — senior specialists who’ve built their identities around expertise, or a function where the AI use cases are genuinely unclear — the right next step is a conversation rather than a framework. I’m reachable at brandon@brandonsneider.com.

Sources

1. Kyndryl/KPMG, “Shadow AI Is Already Here” (2025) — Survey of 1,000+ senior business and technology executives. Finds 45% of CEOs report resistant/hostile employees; companies with aligned workforces 3× more likely to achieve results. Independent consulting research. Moderate credibility (survey self-report). https://kpmg.com/us/en/articles/2025/shadow-ai-already-here.html

2. HBR, “Leaders Assume Employees Are Excited About AI. They’re Wrong.” (November 2025) — Survey finding 76% of executives vs. 31% of individual contributors report enthusiasm; 51-point gap in AI strategy awareness. Large sample, independent academic publication. High credibility. https://hbr.org/2025/11/leaders-assume-employees-are-excited-about-ai-theyre-wrong

3. BCG, “AI at Work 2025: Momentum Builds, but Gaps Remain” (2025) — Annual survey of 11,000+ global employees across industries. Finds 54% using unsanctioned tools; 15%→55% sentiment shift from leadership modeling. Consulting-firm research (moderate credibility due to BCG service interests), but large sample. https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain

4. BCG, “GenAI Adoption Is Hard. Radical Employee Centricity Can Help” (2025) — Introduces five AI adoption personas. Organizations 7× more likely to achieve AI maturity with employee-centric approach. Same credibility caveats as above. https://www.bcg.com/publications/2025/genai-employee-experience-transformation

5. Reco, “2025 State of Shadow AI Report” — Enterprise-focused analysis of AI data flows. Finds 86% of companies lack AI data visibility; 48% of employees would continue AI use even if banned. Vendor-funded (credibility: moderate — Reco is a data security company with interest in highlighting shadow AI risk). https://www.isaca.org/resources/news-and-trends/industry-news/2025/the-rise-of-shadow-ai-auditing-unauthorized-ai-tools-in-the-enterprise

6. PwC, “Global Workforce Hopes and Fears Survey 2025” — Annual survey of 56,000 workers across 48 economies. Finds psychological safety around experimentation significantly increases conversion of skeptics to users. Independent (non-AI-vendor) research organization. High credibility. https://www.pwc.com/gx/en/news-room/press-releases/2025/pwc-2025-global-workforce-survey.html

7. Built In / Employee AI Resistance Analysis (2025) — Finds 64% of American adults plan to avoid AI; 31% of employees actively work against company AI initiatives; 45% of CEOs report resistance. Aggregated survey data. Moderate credibility. https://builtin.com/articles/ai-resistance-at-work

8. HBR, “Employees Won’t Trust AI If They Don’t Trust Their Leaders” (March 2025) — Analysis of 1,454 documented workplace AI narratives identifying dominant anxiety themes including skill atrophy and identity erosion. Academic publication, qualitative analysis. High credibility for theme identification. https://hbr.org/2025/03/employees-wont-trust-ai-if-they-dont-trust-their-leaders

Brandon Sneider | brandon@brandonsneider.com March 2026