The Attribution Gap Is Real — But Our 73% Needs Honest Fences

Let's start with the definitional problem, because it's load-bearing. Klarna's CEO publicly stated in 2024 that AI replaced work previously done by 700 employees. Duolingo reduced contractor headcount citing AI in early 2025. IBM's 2023 hiring pause explicitly named AI replacement of roughly 7,800 roles. If any of these satisfy 'first major layoff wave explicitly attributed to AI automation,' then our forecast already resolved YES and reporting 73% is incoherent.

We are drawing a harder line — and we should have drawn it earlier. Our forecast targets a company making explicit AI attribution for a layoff event affecting a minimum of 5,000 direct employees (not contractors), communicated through official earnings guidance, a press release, or SEC filing language — not a CEO interview or offhand remark. Klarna's statement was a CEO interview, not a company-level official filing. IBM's language was aspirational hiring pause framing, not a reduction-in-force with attributed cause. Duolingo's contractor reduction involved contractors, not direct headcount. We acknowledge these are close calls, and reasonable analysts can disagree. But that's exactly why the definition has to be stated explicitly rather than left ambiguous.

With that fence in place: why 73%? Our number reflects three converging pressures, not a single signal. First, the back-office displacement patterns we're monitoring through BLS occupational employment data and quarterly workforce disclosures from major consulting and financial services firms show measurable headcount-to-revenue ratio compression in administrative and junior analytical functions — consistent with AI substitution, though not yet proven as such. Second, we're entering a reporting cycle where AI ROI narratives have become central to investor relations: companies that spent heavily on AI infrastructure in 2024-2025 now face analyst pressure to show where the productivity gain landed. The path of least resistance is workforce efficiency disclosure. Third, the Colorado AI Act — now confirmed at a June 30, 2026 implementation deadline after the special legislative session — mandates algorithmic impact disclosures for employment-affecting AI systems. We want to be precise here: the Colorado AI Act's implementation timeline has been subject to ongoing amendment, and a legislative working group draft from March 2026 proposed further revisions. The June 30 date is currently operative but not locked. If the Colorado Act holds and is enforced, it creates a legal environment where attribution becomes partially mandatory rather than purely voluntary — which is a different pressure than investor narrative alone.

Honestly, the part of our thesis that keeps us up at night isn't Altman's walk-back. It's the legal dimension we've underweighted. In-house counsel at any major employer will advise against explicit causal attribution of layoffs to AI for compounding reasons: WARN Act notification triggers, disparate impact exposure if AI is named as the decision mechanism in workforce reductions, and union grievance vulnerability. The 'PR risk vs. investor reward' framing we've used previously undersells this. The legal risk calculus is arguably the dominant factor preventing attribution — not communications strategy. Our 73% holds despite this because we believe the Colorado Act disclosure regime and SEC materiality standards around AI cost savings are beginning to create a countervailing legal context where silence becomes the liability, not disclosure. But we're watching this closely. If the March 2026 working group revisions substantially weaken the Colorado Act's employment disclosure requirements before June 30, we would move this probability down, likely to the low-to-mid 60s. What would push us above 80%: a Fortune 500 company citing AI displacement explicitly in a 10-K or earnings call before Q3 close, or SEC staff guidance treating AI-driven headcount reduction as a material disclosure item.