The AI Likelihood Test asks whether an AI exposure appears reasonably likely. The next test asks whether that exposure crosses into materiality.

When AI affects core workflows (e.g. revenue, customers, compliance, reporting, …), management should look beyond governance and test material exposure. It should ask whether the exposure could matter to the enterprise.

Materiality changes the conversation.

A technical team may describe an AI failure as an error rate. A business unit may describe it as a process issue. Legal may frame it as a contractual, indemnity, or liability question. Risk may see a control gap. CFOs and Audit Committees must ask a different question: 𝐂𝐨𝐮𝐥𝐝 𝐭𝐡𝐢𝐬 𝐀𝐈 𝐞𝐱𝐩𝐨𝐬𝐮𝐫𝐞 𝐫𝐞𝐚𝐬𝐨𝐧𝐚𝐛𝐥𝐲 𝐚𝐟𝐟𝐞𝐜𝐭 𝐨𝐩𝐞𝐫𝐚𝐭𝐢𝐨𝐧𝐬, 𝐟𝐢𝐧𝐚𝐧𝐜𝐢𝐚𝐥 𝐩𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞, 𝐢𝐧𝐯𝐞𝐬𝐭𝐨𝐫 𝐜𝐨𝐧𝐟𝐢𝐝𝐞𝐧𝐜𝐞, 𝐨𝐫 𝐞𝐧𝐭𝐞𝐫𝐩𝐫𝐢𝐬𝐞 𝐯𝐚𝐥𝐮𝐞?

That requires financial judgment, not just technical assessment, and the relevant questions remain straightforward:
• Does the AI use case touch a material business process?
• Could failure create a material cost, liability, delay, claim, restatement, disruption, or customer impact?
• Can management reasonably estimate the exposure in dollars?
• Has the company mitigated, insured, transferred, reserved for, disclosed, or otherwise priced the residual risk?
• Does the exposure reflect a known trend or uncertainty reasonably likely to materially affect operations, liquidity, capital resources, or results?

Under Regulation S-K Item 303’s MD&A framework, management cannot treat known trends and uncertainties as background noise when they appear reasonably likely to have a material effect. AI exposure deserves the same discipline.

Aggregation can be the difference. One AI use case may not cross a materiality threshold alone. However, ten, twenty, or fifty deployments across business units may collectively create a dollar-quantified exposure that does. 𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐢𝐛𝐥𝐞 𝐀𝐈 𝐬𝐜𝐨𝐫𝐞𝐬, maturity ratings, and governance dashboards can assess process quality, but they 𝐝𝐨 𝐧𝐨𝐭 answer the financial materiality question.

Not every AI issue requires disclosure. Not every AI exposure calls for reserve treatment. But scaled AI systems should not become enterprise dependencies before management tests both likelihood and materiality.

AI ROI cannot measure only efficiency gained. It must also include reasonably likely material exposure introduced. That moves AI beyond governance and into financial oversight.