97% of Your AI Agent Work Is Invisible to Your Contracts.

What the Research Covers

• The three-tier taxonomy: how CPAG sizes a market with no standard definition — Tier 1 (billed resolutions, ~3%), Tier 2 (verifiable but unpriced, ~22%), and Tier 3 (untracked dark matter, ~75%)
• Why Salesforce invented a proprietary measurement unit rather than adopt a standard — and what that reveals about the state of the market
• The three criteria that separate a commercially durable Atomic Resolution from a disclosure metric a CFO can dispute: verifiable, attributable, and finite
• Enterprise vertical breakdown: customer support (~38B completions), IT operations (~22B), sales and marketing (~18B), finance (~10B), HR (~7B), legal (~4B)
• SMB adoption data: the large-to-small AI adoption ratio narrowed from 1.8x to 1.2x between 2024 and 2025 — faster than any prior technology cycle
• Three-scenario cost model through 2030: base case ($0.12/resolution), accelerated ($0.04), plateau ($0.20) — and what each means for the 1-to-4 Rule
• The governance gap in numbers: 41% of failed deployments trace to unclear success criteria, 40%+ of agentic projects at cancellation risk by 2027
• Vendor taxonomy: who is billing per resolution today, who is transitioning, and who is dangerously exposed
• The forcing events: why the 2028 to 2029 renewal cycle is the structural moment that cannot be deferred