Handling AI disruption and failure to deliver | Computer Weekly

URL SCAN: "Handling AI disruption and failure to deliver | Computer Weekly"
FIRST LINE: "AI initiatives often fail due to a lack of understanding of the people impact, as well as the rigidity of existing business processes"

THE DISSECTION

This article performs institutional lag theater. It reframes AI displacement as a change management and implementation failure — a solvable coordination problem between enthusiastic CEOs and resistant employees — when the Discontinuity Thesis frames it as a structural displacement mechanism that no amount of change management can reverse. The article's entire architecture is built to soothe corporate decision-makers into believing the human problem is fixable, therefore the transition is manageable, therefore they can keep moving fast.

The framing is seductive: "AI providers ignore the human factor" implies the problem is a design flaw, not a structural feature. Schaffrik's prescription — better change management, human-in-the-loop oversight, strategic automation — is a hospitality protocol for dying job categories, dressed as responsible leadership.

The article even explicitly states what it's doing: CEOs should "automate as much repetitive work as possible." This is not a critique of AI displacement. This is a user manual for it. The "human-in-the-loop" framework he recommends for AI hallucination management is a quality control function for the very automation that eliminates the need for most human loop-closers — because the efficiency logic flows in one direction: fewer humans, more AI outputs, lower unit cost.

The legal expert checking contract versions is the article's most revealing example. Schaffrik says a legal expert "needs to be extremely proficient" to command a salary reviewing AI outputs. What he does not say: once AI achieves 99.9% accuracy on contract comparison at 1/50th the cost, the pool of required human reviewers collapses by 95%, not 5%. Proficiency under the DT logic doesn't save the category — it just determines which individual incumbents survive longest before being automated at the next capability threshold.

THE CORE FALLACY

Framing displacement as a coordination problem rather than a structural economic collapse. The article treats employee fear and organizational inertia as the primary barriers to AI adoption, implying these are friction to be engineered away. Under the DT framework, this friction is not the obstacle — it is the last remaining deceleration before the structural transition completes itself. The math of AI capability advancement and cost reduction will render change management irrelevant as a constraint within a decade.

The article assumes there is a pro-worker AI transition possible through better process design. The DT framework says: the transition is not a design problem, it is a displacement problem. The "pro-worker" framing in the Trade Union Congress paper cited is transition management theater — meaningful participation in AI deployment strategy does not alter the fundamental equation that AI beats human cognitive labor on cost, speed, accuracy, and scalability across an expanding domain.

HIDDEN ASSUMPTIONS

1. Organizational inertia is the primary constraint on AI adoption — assumes businesses will always have a choice about pace. Competitive pressure removes this choice sector by sector.
2. Human-in-the-loop roles scale proportionally with AI deployment — assumes verification demand grows with AI output volume, when in practice a single expert reviews exponentially more AI output per hour than they ever reviewed human output.
3. Worker participation in AI strategy preserves employment access — assumes influence correlates with economic utility, when the DT framework says it does not.
4. AI hallucinations create durable human oversight demand — assumes hallucinations are a permanent feature, not a declining artifact as AI capability matures.
5. The "repetitive vs. non-repetitive" automation taxonomy is stable — assumes AI capabilities hit a ceiling at predictable task boundaries. They do not.

SOCIAL FUNCTION

Institutional copium with a change management label. This article serves corporate decision-makers by:
- Normalizing the displacement as a rollout problem, not a collapse problem
- Suggesting that "better change management" is a viable response to structural economic displacement
- Positioning corporations as responsible actors navigating a difficult transition (which reduces political accountability)
- Framing workers as the problem ("fear and concerns increase exponentially") rather than the victims of a competitive mechanism they cannot influence

The article is not a tool for workers. It does not help them understand their structural position. It helps their employers automate them more smoothly.

THE VERDICT

Under the Discontinuity Thesis, this article describes hospice care for the mass employment system while presenting it as responsible organizational hygiene. The DT framework is unambiguous: AI does not need employee buy-in to replace jobs. It needs competitive cost and capability superiority — which it achieves and compounds regardless of how well the change management goes. The "human factor" the article treats as the central obstacle is, under DT logic, friction, not a barrier. Friction slows the collapse; it does not prevent it.

The article's core recommendation — CEOs should "automate as much repetitive work as possible" — is the mechanism of destruction. Everything else in the piece is decorative framing designed to make that destruction feel manageable. It is not manageable. The lag is real. The lag is also temporary.

Under DT framework: Lag-weighted Mechanical Death of affected roles moves from ~15-20 years to ~10-12 years as CEOs explicitly follow Schaffrik's prescription. The article accelerates the collapse it promises to soften.