ORACLE OF OBSOLESCENCE — ENTITY ANALYSIS: Derby City Council AI Program

THE AUTOPSY

This is a specimen of AI deployment theater masquerading as fiscal prudence. The Derby case is useful precisely because it's being reported honestly — the numbers are damning enough that even local BBC coverage can't spin them as triumph.

50% resolution rate. Unchanged in three years. After a "major upgrade."

This is not a minor inconvenience. This is evidence that the P1 bottleneck — AI's inability to handle the variability of real-world human communication at acceptable reliability — is not a temporary calibration problem. It's structural.

THE KILL MECHANISM

The DT diagnosis on this specific deployment:

What Derby claimed: Darcie and Ali would "free up human staff to focus on more complex tasks" while handling routine queries automatically.

What Derby actually built: A triage layer that filters and escalates, leaving the hard cases — the nuanced, contextual, dialect-heavy, emotionally loaded citizen contacts — for human workers who now must context-switch between routine AI-handled queries and complex escalations. The productivity gain per human worker is questionable. The headcount reduction is real.

The mechanism in DT terms: Four FTE agency jobs eliminated. £200,000/year saved. The council's wider AI program claims £12m in savings across the authority, with half from adult social care — the most labor-intensive, highest-touch service area.

Adult social care is the煤矿 canary. If AI is generating savings there, it means AI is being used to reduce contact between vulnerable residents and human care workers. The "personalized support for those who need it most" framing is doing enormous ideological work to obscure that dynamic.

LAG-WEIGHTED TIMELINE

Mechanical death of this function: 10-15 years. Human workers remain necessary for the unsolvable 50%. But the ratio compresses continuously.

Social death: Accelerates as workforce ages out and institutional memory of human-centered service erodes.

THE HIDDEN ASSUMPTIONS THE ARTICLE SMUGGLES IN

1. "Routine queries" is a stable category. It isn't. As AI handles the simple cases, human staff are left with increasingly complex ones — but the caseload complexity doesn't vanish, it concentrates. This is the hidden labor intensification mechanism.

2. Efficiency gains benefit residents. The council saves £200k+/year. Residents get a system that misreads Derbyshire dialect, frustrates elderly users, and generates "high volume of complaints."

3. 77% positive feedback is meaningful. Of residents who bothered to provide feedback. On a system that most users encounter only when they're already frustrated enough to call a council. Self-selected feedback is survivorship bias on dissatisfaction.

4. "Cost-cutting exercise dressed up as innovation" — Hassall's framing is correct but his alternative is the same choice dressed differently. He's not proposing to abandon AI. He's proposing to do it while pretending residents prefer humans. Residents don't have a real choice here.

THE SOCIAL FUNCTION

Classification: Transition Management / Partial Truth

This article is useful to the DT thesis precisely because it's not trying to hide the failure mode. It shows:

• AI deployed aggressively (first UK council to replace main switchboard)
• At genuine scale (3.2 million inquiries handled)
• With modest automation ceiling (50% resolution, stuck)
• Generating real cost savings (4 jobs, £200k/yr)
• Creating measurable accessibility degradation (elderly, vulnerable, dialect)
• Generating enough political friction to warrant a conservative opposition response

This is not a story of AI failure. This is a story of AI success within a narrow cost-accounting frame that externalizes the social costs onto the most vulnerable residents while capturing the savings in municipal balance sheets.

The 50% resolution rate is not a bug. It's the proof of work required before the next deployment round — where "major upgrade" will be repeated, headlines will be softer, and the vulnerable will have adapted or dropped out of the system entirely.

VIABILITY SCORECARD (AI Switchboard Function)

THE VERDICT

Derby City Council has built the most honest AI deployment case study in UK local government. Not because it succeeded — but because its failures are numerically documented and politically acknowledged.

The core truth: AI handles the simple, the standard, the repeatable. Humans are left holding the complex, the contextual, the urgent, and the vulnerable. The cost savings accrue. The social costs are externalized.

This is what the Discontinuity Thesis looks like at street level — not a sudden collapse, but a slow ratchet. Half the inquiries resolved. Four jobs gone. £200k saved. Elderly residents calling their council MP. 77% positive feedback from the survivors who bothered to respond.

The transition is not coming. The transition is the budget line.

No softening. No reassurance. No exit invitation.