How we score Forbes Global 2000 incumbents on AI exposure and AI defensibility.

The Stork Quadrant for Forbes Global 2000 incumbents is an editorial framework that places one named public company per day on a 2-axis grid: AI Defensibility (Y) by AI Exposure (X). It's a coordinate, not a prophecy. We don't predict bankruptcies. We describe where the company sits today on two axes, with primary sources cited for every claim.

This is not investment research. It does not recommend trading actions. It does not estimate cash flows, target prices, or probabilities of survival. Read it as a journalist would: a structured editorial verdict based on what the company has itself disclosed, with our reading of the implication for AI and agent disruption.

AI Defensibility (Y-axis, 0–100) — how protected is the company's core business from being replaced by LLM/agent automation? Scored by which of seven moats materially apply:

• Physical — hardware, logistics, real estate, payment terminals, physical labor coordination.
• Regulatory — banking charters, FDA approval, defense clearance, HIPAA-gating, licensed-profession barriers.
• Network — two-sided or N-sided marketplaces where users bring more users; liquidity is the product.
• Data — proprietary refreshing data nobody else has (terminal data, MLS, exchange data).
• Trust — catastrophic-mistake workflows where someone must bear liability (legal, dosing, fund transfers).
• Coordination — multi-stakeholder orchestration where the value is making N parties act in concert.
• Brand — taste, community, cultural authority, identity that customers refuse to switch from.

AI Exposure (X-axis, 0–100) — how much of the company's revenue / workflow is in the path of LLM/agent disruption? Scored by which of seven signals materially apply:

• Revenue at risk — the company's own SEC filings name AI as a material business risk.
• Workforce automatable — large share of the company's (or its customers') labor force is in roles LLMs already perform competently.
• Customer AI adoption — customers are actively substituting AI for what this company sells.
• Competitor AI-native — an AI-native competitor is materially eating share or named in filings.
• AI spend disclosed — material AI capex / opex / acquisition disclosed; the company itself sees AI as material.
• Layoffs technical — recent layoffs in categories plausibly attributable to automation.
• Pivot in progress — explicit AI pivot or AI-first restructuring announced.

Each signal has a weight (sums to 100 per axis). Scores are computed deterministically from which moats/signals apply — we don't hand-pick numbers.

Placement is computed from the two scores at a median threshold of 50:

• Insulated — low exposure, high defensibility. Core business is mostly out of the AI disruption path and well-moated.
• Reinventing — high exposure, high defensibility. AI is hitting the business, but the moats give time + leverage to pivot.
• Slow Drift — low exposure, low defensibility. Not immediately threatened by AI, but secular pressures dominate.
• Acute Risk — high exposure, low defensibility. Core product overlaps with LLM/agent capabilities and the moat is thin.

A label is a coordinate, not a prediction. A company in "Acute Risk" is not "going bankrupt next quarter" — it's in a quadrant where the disruption vector intersects with limited natural defense. Whether management responds successfully is a separate question outside this score.

Every published assessment cites primary sources. Our default stack (free, mostly US-listed; extending to non-US):

• SEC EDGAR — 10-K (Item 1A Risk Factors, Item 7 MD&A), 10-Q, 8-K filings.
• GDELT — global news with org-tagged sentiment, 15-minute refresh.
• WARN Firehose — US state WARN layoff notices plus DOL claims.
• USPTO PatentsView — AI-tagged patent filings as an R&D / build-vs-buy proxy.
• SEDAR+ (Canada), HKEXnews (Hong Kong), TDnet (Japan — English required for Prime Market since April 2025) for non-US filers.
• Public press releases and the company's own investor relations site.

We do not use Glassdoor / Blind data (ToS), private analyst reports, paid earnings-transcript services like AlphaSense, or any non-public source.

For each assessment we:

1. Pull the latest SEC filings + recent news for the company.
2. Score both axes with Claude Sonnet 4.6, structured-output prompted, three independent runs.
3. Aggregate moats and signals by majority vote across runs.
4. Compute final scores and quadrant placement deterministically from those moats/signals.
5. Record the spread between runs; flag low-confidence assessments where spread > 20.
6. Editorial review before publication. No automated posting.

Each assessment is dated. Re-scores happen on capability shifts (new 10-K filed, major news, capability change in the AI engines that altered the disruption math) and are recorded in an append-only history table so drift is auditable.

• We do not predict which companies will "die" or "go bankrupt."
• We do not estimate fair value, target prices, or probability of stock returns.
• We do not score private companies (only Forbes Global 2000 constituents).
• We do not score AI tools through this framework (that's the separate Stork Quadrant for AI tools).
• We do not name individual executives critically. Our subject is the company's disclosures, not management character.
• We do not solicit, accept, or honor paid placement, promoted positioning, or "right of preview" for covered companies.

Every covered company can submit a response to its assessment by emailing hello@stork.ai. If the response is on-record and factual (corrections to evidence we cited, additional disclosures we missed, links to filings we should have referenced), we will:

• Append the response inline to the published LinkedIn post (or our website mirror) within 7 calendar days.
• Re-score the company against the new evidence and publish the diff.
• Add an entry to /errata for any factual claim we got wrong.

We do not edit or remove the original assessment. Errata are additive; the public record is preserved.

• Stork holds no position, long or short, in any covered company's securities at time of publication.
• Stork has no commercial relationship — paid, sponsored, or trade — with any covered company at time of publication.
• If either changes for any specific company, that company will be removed from coverage permanently.
• Stork employees and contractors are barred from trading covered companies in the 7 days before and 30 days after a published assessment.

This document is methodology version f2000-v1.0, published 2026-05-28. Material changes to scoring (new signals, weight changes, threshold adjustments) trigger a new version. Historical assessments retain the version they were scored under; the history page surfaces both.