Four Enterprise Control Surfaces

An enterprise agent control plane must connect technical execution with business accountability.

Four control surfaces organize the operating decisions.

They represent distinct management responsibilities supported by shared execution infrastructure.

Control Surface 01. Outcome Economics

The economic unit of agentic execution is the accepted business outcome.

Inference cost measures one component of the operating system.

A complete economic assessment also includes tool execution and infrastructure consumption. Human review and remediation contribute additional costs.

The enterprise must establish whether the workflow produces accepted results at an economically sustainable cost.

A useful measure is:

Cost per Accepted Outcome = Total Attributable Workflow Cost / Number of Accepted Outcomes

The Accounting Boundary

The measurement follows the canonical economic model established in Agentic Execution, Chapter 12.

Total attributable workflow cost includes human execution and AI consumption. It includes infrastructure and integration expenditure. Review and recovery costs form part of the same accounting boundary. Allocated transformation investment completes the measurement.

Costs and accepted outcomes must share a defined measurement period and outcome class.

The numerator includes failed attempts and rejected work. Retry expenditure and recovery effort remain attributable to the workflow.

Shared costs require a consistent allocation basis. Transformation investment is allocated over a disclosed useful-life horizon.

Each cost is assigned once. Human execution labor and human review labor are classified consistently to preserve accurate attribution.

Each distinct accepted outcome enters the denominator once after satisfying its acceptance criteria.

A period with zero accepted outcomes reports the incurred expenditure separately. The unit-cost measure remains undefined for that period.

This accounting discipline enables meaningful comparison across execution configurations while preserving the distinction between operating efficiency and realized business value.

Connected reference: Agentic Execution, Chapter 12, The Economics of Agentic Execution.

https://sentient-review.com/perspectives/agentic-execution/12-the-economics-of-agentic-execution

The numerator includes expenditure associated with failed attempts and recovery.

The denominator includes outcomes that satisfy defined business acceptance criteria.

The measure should be accompanied by quality and workload coverage.

A workflow that processes only simple cases may achieve attractive unit economics while transferring difficult work to human teams.

Economic accountability therefore requires a complete view of the operating process.

The control plane supports this responsibility through attribution and operating thresholds.

Control Surface 02. Authority and Action Safety

Authority determines which participants may perform consequential actions.

An agent may receive permission to read records or prepare recommendations.

A separate authorization may permit a financial commitment or a change to an enterprise system.

These permissions belong to explicit identities and scopes.

The control environment establishes delegated authority and approval requirements.

Enforcement services apply those requirements at the relevant action boundaries.

The enterprise must also preserve intervention capabilities.

A workflow may require suspension when permissions change. A compromised credential may require revocation. An uncertain transaction may require reconciliation.

The control surface connects delegated capability with accountable decision rights.

Control Surface 03. Evaluation and Assurance

Agentic execution requires continuous evidence about operating performance.

Predeployment evaluation establishes an initial qualification baseline.

Production monitoring identifies changes in behavior and operating conditions.

Incident analysis connects observed failures with corrective action.

Evaluation must examine the deployed configuration and the resulting business state.

A model comparison may establish differences in task quality.

An execution assessment must also examine tool effects and recovery behavior.

The enterprise should maintain evaluation suites appropriate to the workflow's consequence.

Higher-impact actions require stronger evidence and more explicit acceptance conditions.

Assurance becomes an ongoing operating process.

Control Surface 04. Capacity and Capital Discipline

AI execution consumes compute and supporting infrastructure.

Routing decisions influence inference expenditure.

Concurrency and serving architecture influence capacity utilization.

Infrastructure placement affects latency and operational exposure.

These decisions connect workload design with infrastructure economics.

The control plane can support operating decisions through resource accounting and capacity policy.

Longer-term capital decisions require an additional perspective.

An enterprise may purchase model services through APIs. It may reserve capacity or operate dedicated infrastructure.

Each arrangement creates a different economic exposure.

Infrastructure ownership introduces commitments concerning hardware and utilization. Power availability and equipment lifecycle can also affect investment economics.

These decisions belong to enterprise technology and capital planning.

The relationship between execution and capital becomes visible when workload demand informs capacity commitments.

The Shared Execution Foundation

The four control surfaces depend on common technical capabilities.

Durable state supports recovery.

Identity and policy services support authorization.

Execution traces support operational investigation.

Authoritative business records establish completed external effects.

Resource accounting connects consumption with operating limits.

These capabilities may be supplied through several platforms.

The enterprise architecture must establish how they operate together.

The control plane creates coherence across the four decision domains. The execution infrastructure provides the mechanisms through which those decisions become effective.

This relationship connects AI architecture with enterprise operating performance.