The Unmanaged Resolution Economy: Sizing AI Agent Task Completions Across Enterprise and SMB Markets, 2026–2030
Approximately 117 billion AI agent task completions will occur across US enterprise and SMB markets in 2026. Fewer than 3% carry outcome-based pricing. CPAG research quantifying the governance gap and what closes it.
“Until every AI agent task completion can answer three questions — Was it verifiable? Was it attributable? Was it finite? — the resolution economy will remain unmanaged.” — The Unmanaged Resolution Economy, CPAG Research Brief, May 2026
Download the Full Research Brief →
Free. No credit card. PDF delivered by email.
Executive Summary
Resolution as a Service (RaaS) is the pricing and architectural model in which enterprise software is priced on problems solved rather than the number of users who log in. The foundational unit of that model is the Atomic Resolution: a discrete outcome that is verifiable, attributable, and finite. In 2026, the market is generating those outcomes at a scale that has outrun every governance framework available to price, audit, or dispute them.
Approximately 117 billion AI agent task completions are estimated to occur across US enterprise and SMB markets in 2026. Fewer than 3% carry any outcome-based pricing. The remaining 97% are absorbed into flat-rate seat contracts or unmonitored infrastructure costs.
Key Findings
- Estimated 86 billion AI agent task completions occurred across US enterprise and SMB markets in 2025. The 2026 projection reaches approximately 117 billion at the mid-case estimate. By 2030, total annual volume reaches between 1.1 trillion (plateau scenario) and 4.2 trillion (accelerated scenario).
- Only approximately 3% of enterprise agentic task completions in 2026 carry Tier 1 outcome-based pricing. The remaining 97% are absorbed into legacy contract structures designed for human-paced workflows.
- Salesforce, the largest enterprise CRM company in the world, invented a proprietary measurement unit — the Agentic Work Unit (AWU) — rather than adopt a standard, because no standard exists. As of Q4 FY26, Salesforce reported 2.4 billion AWUs delivered to date, growing 57% quarter-over-quarter. The AWU cannot be priced, audited, or disputed at the contract level by a CFO.
- Gartner projects a 90% reduction in LLM inference costs by 2030. Yet enterprise AI bills are rising because agentic workflows consume 10 to 50 times more tokens per task than single-query interactions.
- Only 21% of organizations have a mature governance model for autonomous AI agents. Gartner projects more than 40% of agentic AI projects will be cancelled by 2027. The primary cause is not model quality. It is the absence of audit trails and attribution frameworks that can survive enterprise procurement review.
- SMB AI agent adoption is closing the enterprise gap faster than any prior technology cycle. The SBA reports the large-to-small business AI adoption ratio narrowed from 1.8x in February 2024 to 1.2x by August 2025.
The Definitional Problem
The unmanaged resolution economy exists because the market has no standard definition of what constitutes a completed unit of agentic work. Every major vendor has invented its own. Salesforce measures Agentic Work Units. Intercom counts resolutions. Zendesk tracks automated resolutions and deflection rates as separate metrics, explicitly noting that deflection and resolution are not the same event. Microsoft Copilot reports assisted actions. ServiceNow measures flow executions. None are compatible.
A resolution that can function as a pricing unit must satisfy three criteria:
Verifiable: The completion must be objectively confirmable against a defined state. A ticket marked closed with no reopen within 48 hours is verifiable. A task the AI “attempted” is not.
Attributable: The completion must be traceable to the AI agent’s execution, not to a human override or an ambiguous handoff. Attribution requires a complete action log.
Finite: The resolution must have a clear endpoint. Open-ended agentic loops cannot be priced per-resolution without a defined scope boundary.
Zendesk’s automated resolution metric comes closest among Tier 1 vendors to satisfying all three criteria, which is why it carries the highest per-unit price in the market: $1.50 to $2.00 per automated resolution, versus Intercom’s $0.99. Definitional rigor commands a price premium. The market is already demonstrating this.
Market Sizing: The Three-Tier Taxonomy
Tier 1 — Vendor-disclosed, explicitly billed resolutions (~3% of volume): Completions for which a specific per-resolution fee is charged and disclosed. Examples: Intercom Fin at $0.99, Zendesk AI at $1.50 to $2.00. Confidence: HIGH.
Tier 2 — Verifiable completions, not billed as resolutions (~22% of volume): AI agent task completions measurable via vendor metrics but absorbed into seat contracts rather than priced per event. Examples: Salesforce AWUs, Microsoft Copilot assisted actions, ServiceNow flow executions. Confidence: MODERATE.
Tier 3 — Estimated untracked agentic completions (~75% of volume): AI-driven task completions with no attribution, no measurement, and no pricing. Background agent loops, unlogged automation runs, AI completions inside bundled copilot features. Confidence: LOW by definition.
The enterprise total is estimated at approximately 99 billion task completions in 2026. Of those, fewer than 3 billion carry Tier 1 outcome-based pricing. The remaining 96 billion represent unpriced agentic work absorbed into legacy contract structures.
The Pricing Gap
If the average Tier 1 resolution price is approximately $1.25, then 96 billion unpriced enterprise completions represent a theoretical revenue gap of approximately $120 billion annually. Even a 10% capture rate at a conservative $0.50 per resolution represents a $4.8 billion annual pricing opportunity that does not currently appear on any enterprise software vendor’s revenue line.
The Forcing Events: 2027 Through 2030
Gartner projects 70% of vendors will refactor pricing away from pure per-seat models by 2028. IDC forecasts seat-based pricing will be effectively dead for most software categories by 2028 to 2029. The 2028 to 2029 renewal cycle is the structural forcing event: contracts signed before AI agent deployment are coming up for renegotiation by customers who can now document the headcount reduction AI delivered.
Download the Full Research Brief, Including All Data Tables and Three-Scenario Cost Model →
What the Full Research Brief Contains
- The complete three-tier taxonomy with methodology notes and confidence levels for each estimate
- Full enterprise vertical breakdown: customer support (~38B), IT operations (~22B), sales and marketing (~18B), finance and accounting (~10B), HR and recruiting (~7B), legal and compliance (~4B)
- SMB adoption analysis with SBA, US Chamber of Commerce, and McKinsey sourcing
- Three-scenario cost model (base, accelerated, plateau) with 2030 volume and inference cost projections
- Vendor taxonomy covering Tier 01 through Tier 03 with named vendor analysis
- The governance gap: four commercial costs already materializing in enterprise deployments
- The resolution diagnostic framework: how to move from untracked completions to a contract-ready resolution stack
- 8 footnotes tracing every material claim to primary disclosures
Download the Full Research Brief →
Free. PDF delivered by email. No credit card required.
This research is produced by Crown Point Advisory Group for informational and thought leadership purposes only. It does not constitute investment advice, legal advice, or a solicitation to buy or sell any security. All figures sourced from primary vendor disclosures, Tier 1 analyst research, or CPAG Research estimates; confidence levels stated explicitly throughout.
Crown Point Advisory Group · crownpointadvisorygroup.com · @CrownPointAG