From Friction to Flywheel

How one reluctant factory turned its AI rollout from a morale drain into a momentum machine

When Dunelm Machining first wired computer-vision cameras above its milling lines, the future sounded frictionless: instant quality checks, zero scrap, slimmer lead-times.

Two months later motion-sickness bands were popular on the night shift. Operators complained the algorithm "nudged" them to speed up without warning; maintenance crews joked about working for a camera, not a company. Production uptime fell three points, and the union threatened a grievance.

---

The Predictable Pattern of AI Resistance

Dunelm's story isn't a glitch. The Future of Work & Leadership in the AI Era lists three predictable trip-wires that stall automation projects.¹

Structural drag — legacy KPIs reward utilisation, not experimentation, and traditional management structures create bottlenecks. According to the report, companies that eliminate middle management layers "in the name of efficiency" often leave remaining managers with double the responsibilities but little support, creating structural barriers to change.

Psychological drag — workers equate AI with job loss, managers equate it with loss of control. A Gallup study cited in the report found that while a majority of employees are optimistic about AI, 41% are apprehensive and need support to embrace it.

Cultural drag — teams lack rituals for surfacing risk early, so fear festers. The report emphasizes that without psychological safety, employees feel unsafe voicing concerns, which is itself built on trust and respectful communication.

Left alone these drags reinforce one another, turning any AI initiative into a vicious cycle.

---

Breaking the Cycle: The Flywheel Approach

The turnaround began with an uncomfortable all-hands on a rainy Thursday. Rather than defend the rollout, COO Leila Grant drew a circle on a whiteboard and wrote three verbs:

Align → Enable → Pilot → Scale.

"This is the flywheel that should have powered the cameras," she said. "We started at Scale and skipped Align. Let's spin in the right order this time."

---

Align

The first fix wasn't technical. Grant rewrote the success metric from "inspection throughput" to "scrap dollars saved and re-invested in crew training." Friction eased the moment people saw themselves in the upside. The report calls this shared value framing—connecting the technology's win to the worker's win.¹

This aligns with what the report identifies as a critical element of future-ready leadership: narrative and purpose-driven influence. Leaders use the power of narrative to "guide their organizations through change and to bind people together around a common purpose" because "humans make sense of complex reality through stories."

---

Enable

Next, two operators were seconded for a week to shadow data scientists, learning how the model flagged defects. Skepticism turned into shop-floor myth-busting: "The camera isn't spying—it's counting burrs." According to the Orion research, peer translators cut AI resistance by up to 30 percent compared with HR-led briefings.

This approach exemplifies what the report identifies as collective leadership, where "leadership roles are shared and rotated, and where teams self-organize to some extent around challenges." By empowering floor operators to become translators, Dunelm distributed leadership throughout the organization.

---

Pilot

Instead of rehitting the whole line, the team ran a micro-pilot on Station 4, infamous for vibration quirks. They paired each alert with a mandatory two-minute operator note: false positive, true catch, unclear. After 500 parts the model's precision climbed from 71 to 93 percent—and operators published a top-ten defect photo log that became onboarding gold.

This demonstrates the systems thinking approach the report advocates as a core competency for modern leaders—the ability to understand how changes ripple through "organizational culture, employee behavior, customer experience, and even society at large." The pilot created a feedback loop that improved both the AI system and human understanding.

---

Scale

Only when the scrap-savings ledger could fund new cutting tools did Dunelm roll the cameras to the remaining bays. The flywheel now powered itself: savings bought goodwill, goodwill surfaced smarter suggestions, smarter suggestions saved more.

Within six months scrap costs dropped 22 percent, and voluntary overtime climbed for the first time since installation. The lens through which workers viewed the cameras had flipped—from surveillance to co-worker that never blinked.

---

Why the Flywheel Sticks: Building Trust, Capability, and Evidence

Dunelm's experience echoes a pattern the report highlights: friction collapses when trust, capability, and evidence reinforce one another in that order. Skip the first step and momentum never forms. Skip the last and stories never turn into metrics the CFO believes.

This sequencing aligns with what the report calls "adaptive and humble leadership"—recognizing that "no leader can have all the answers" as "the technology and the business environment evolve too quickly," and success comes through "a willingness to learn, to admit what one doesn't know, and to empower others who know more."

---

The lesson scales beyond shop-floors. Operations directors can swap "camera" for "factory automation," customer experience leaders can swap it for "voice analytics." What matters is spinning the loop in sequence:

• Align: Create shared value and purpose

• Enable: Build capabilities through peer learning

• Pilot: Test and refine with feedback

• Scale: Expand only after proving value

---

A Quiet Word on Practice

Because Systems Thinking Needs Practice

Dunelm's turnaround wasn't just about following steps—it was about developing a new perspective.
If complexity overwhelms, it's not a lack of intelligence. It's a missing perspective.²

Inside Orion's Leadership Gym, the Practice Labs specialize in helping leaders work through real-world challenges like AI resistance. Leaders bring a live blocker—like Dunelm's camera implementation friction—and apply systems thinking to see beyond symptoms to patterns. In these 85-minute sessions, they map the interrelationships between technology, people, and processes, finding the leverage points where small changes create significant impact.

Most leaders feel complexity daily. Issues intertwine. Solutions create new problems. Actions have unintended consequences. But few have the tools to step back and see the whole system.

This is precisely what happened at Dunelm—a technical solution created human resistance because the system wasn't seen as a whole.

The Practice Labs transform how leaders understand problems by revealing these underlying structures. They don't just rehearse the flywheel; they help leaders see the reinforcing loops (how resistance breeds more resistance) and identify balancing loops (how shared value can flip the dynamic). The focus on Feedback and Causality is essential—examining how interconnections and feedback loops create organizational behaviors.

This approach moves organizations beyond Silos and Fragments where teams optimize their part while the whole suffers. Instead of HR, operations, and tech teams implementing AI in isolation, the system perspective brings them together to learn and align collectively.

---

The Next Time AI Hisses on Contact

The next time an AI project hisses on contact, remember Dunelm's U-turn.
Friction isn't a sign to brake—it's a sign the flywheel is spinning the wrong way.
Flip the order, and every turn gathers speed.

---

Key Systems Leadership Takeaways

• See patterns, not just problems — Rather than treating AI resistance as isolated complaints, map the system dynamics creating that resistance. Lead through patterns, not just problems.¹⁵

• Find leverage points — Identify where small changes create significant impact. For Dunelm, rewriting the success metric from "inspection throughput" to "scrap dollars saved and re-invested in crew training" was a high-leverage intervention that shifted the entire system.

• Work with feedback loops — Understand how resistance reinforces itself and design interventions that create positive momentum instead. Practice Labs help leaders understand and harness reinforcing and balancing loops.¹⁶

• Foster emergence through alignment — Enable collective learning by aligning everyone around shared value and purpose. When operators published their top-ten defect photo log, this was emergence and self-organization in action—an unplanned positive outcome arising from the new system conditions.¹⁷

• Build system health and resilience — Focus on cultivating system conditions that foster vitality and the capacity to adapt¹⁸ by investing in capabilities and continuous learning rather than just driving compliance.

This approach aligns with the report's conclusion that *"technology is only a tool – it's the people wielding it, with purpose and ingenuity, that truly move the world forward."*¹⁹

---

Footnotes

1. The Future of Work & Leadership in the AI Era, Orion Group, 2025, p. 12

2. Ibid., p. 11

3. Ibid., p. 10

4. Ibid., p. 14

5. Ibid., p. 16

6. Ibid., p. 22

7. Ibid., p. 15

8. Ibid., p. 5

9. Ibid., pp. 19–20

10. Ibid., p. 6
    11–18. Navigating Complexity with Systems Thinking, Orion Group Limited, 2025. https://www.oriongrouplimited.com/programs/navigating-complexity-with-systems-thinking

11. The Future of Work & Leadership in the AI Era, Orion Group, 2025, p. 20