The Attribution Wall Is Breaking: Why 70% on White-Collar Displacement May Still Be Too Conservative

Let's start with the resolution question, because it's legitimate: if Block's Jack Dorsey cut 4,000 jobs explicitly because AI 'could do the work more honestly,' and Meta is eliminating 8,000 positions while simultaneously committing $135B to AI infrastructure and installing employee-tracking software to train AI agents — why hasn't this forecast already resolved YES?

The honest answer is that our forecast target requires more specificity than we've historically applied to it. We're not forecasting that any company will attribute any layoffs to AI. We're forecasting a threshold event: a single-company action of 5,000+ eliminations with explicit, earnings-call-level AI attribution from a non-tech-native Fortune 50 firm. Block and Meta are significant, but Block is a fintech mid-cap and Meta is about as tech-native as companies get. The forecast is asking whether AI displacement becomes undeniable enough that a traditionally non-tech company — a retailer, insurer, bank, or manufacturer — publicly commits to that attribution at scale. That threshold has not been crossed. We're flagging this explicitly so readers can evaluate it themselves, not hiding behind definitional fog after using these events to move the probability.

Now to the Tech Insider data, which we cited as our strongest proximate indicator: 47-50% of 150,000 tech layoffs year-to-date explicitly attributed to AI efficiency gains in analyst and company communications. We want to be precise about what this proves. This is proximate evidence, not direct evidence. The attribution is applied by a third-party data aggregator classifying layoff announcements using its own taxonomy — which may be applied loosely, particularly when an earnings call mentions AI strategy in the same breath as restructuring. The 47-50% range itself signals methodological imprecision. We can't link directly to the underlying Tech Insider methodology, which limits verifiability. What this data point establishes is that AI-framed restructuring language is now common enough that aggregators are tracking it as a primary category — that's meaningful signal about executive communication norms, but it is not the same as verified causal evidence that AI systems literally replaced identified workers. We're watching whether independent payroll and job-category analysis corroborates the narrative.

The counterargument that keeps us honest: executive AI-attribution language has a strong investor-signaling incentive that is independent of causal reality. When Dorsey says AI is 'more honest' than humans, or when Oracle redirects freed headcount capital to AI data centers, these statements may reflect IR strategy as much as operational truth. Executives know that AI-framed restructuring trades at a higher multiple than 'we over-hired in 2021.' The question of whether AI is the actual displacement mechanism — versus AI being the strategic framing for inevitable correction — matters enormously for this forecast. Evidence that would help distinguish the two: job category analysis showing whether eliminated roles are actually being replaced by AI tooling (not just eliminated), and whether the companies are showing productivity gains in the affected functions rather than simply shrinking. The S&P 500 aggregate 400,000 headcount decline in 2025 (from Office Chai, citing a first-since-2016 reversal) is consistent with this thesis but cannot distinguish AI-driven displacement from post-pandemic correction, rising rates compressing headcount, or M&A restructuring. We're treating it as contextual, not confirmatory.

What would move us? Above 80%: a Fortune 50 non-tech-native firm announces 5,000+ cuts with specific AI deployment cited in earnings guidance as the enabling mechanism — not just capital reallocation language, but named systems replacing named functions. Below 55%: two consecutive quarters of net tech sector hiring recovery, or a major executive walks back AI-attribution language under analyst scrutiny, signaling the framing was IR theater. Our 70% reflects the Block and Meta evidence as strong proximate signals that attribution norms are shifting, offset by the investor-signaling confound and the unmet resolution threshold for our specific forecast target. We think 70% is right. We're not certain.