The Command Mapping Thesis

The apparatus gap and the discipline that fills it.

Dan Dollins · Mesa Group · May 2026

01 · The argument, in compressed form

The existing management apparatus does not specify decision authority for an automated decision. That gap is where most artificial intelligence initiatives fail.

Three independent measurements at three altitudes confirm the consequence. Roughly six percent of corporate artificial intelligence investment produces significant earnings impact. Roughly ninety-five percent of generative pilots fail to deliver measurable business value. RAND Corporation reports that more than eighty percent of artificial intelligence projects fail, approximately twice the rate of information technology projects that do not involve artificial intelligence.

The failures trace to a missing specification layer rather than to model performance.

Mesa Group introduces Command Mapping as the discipline that fills the gap. The deliverable is the Command Chart — an org chart extended with three additional layers: decision authority, human-in-loop placement, and agent placement.

02 · The four instruments

Modern organizations operate against four instruments of management. Org charts assign reporting relationships. Performance indicators measure outcomes. Objectives and key results set goals on a defined cadence. Statements of work scope and contract the work itself.

Each instrument does its job. None of them specifies decision authority for an automated decision.

The four instruments were built when every decision had a human owner. Authority was implicit in role. The question of who owns an automated decision did not exist when the apparatus was designed, and the apparatus does not contain an answer.

03 · What changes when automation enters the flow

Before automation entered the decision flow, every decision had a human owner. The apparatus appeared adequate because the missing layer was supplied by default.

When automation enters the flow, the default no longer holds. An automated process produces a decision, and the apparatus does not specify who owns it. The org chart does not include the automation. The performance indicator measures an outcome without identifying which agent produced it. The statement of work scoped the build but did not scope the decision authority.

A more capable model making an unauthorized decision is a more capable problem. Capability gains compound the consequence of unspecified authority rather than relieving it.

04 · The discipline

The work that specifies decision authority for an automated decision, before the automation is built, is a discipline. Mesa Group names it Command Mapping.

Command Mapping specifies four elements. Who owns the authority for each automated decision. What level of autonomy the automation is approved to exercise. Where the automation's authority starts and where it ends. How that authority sits inside the existing decision flow.

05 · The artifact

The Command Chart A diagram showing the four-layer Command Chart structure. Layer 1 at top is the existing org chart with an executive node at the center connected to three function nodes below labeled Function A, Function B, and Function C. Layer 2, decision authority, shows three blocks below each function naming the owner of decisions in that function's flow. Layer 3, human-in-loop, places review and escalation checkpoints between the authority blocks and the agents. Layer 4 at bottom shows three agent placement blocks with autonomy levels labeled autonomous, bounded, and assistive.
Figure 1 · The Command Chart. Layer 1 is the existing org chart. Layers 2 through 4 are added by Command Mapping.

The deliverable is the Command Chart. An org chart extended with three additional layers. The chart shows how decisions move through the organization when humans and automations share the work.

The Command Chart sits next to the org chart. It does not replace it. It extends the instrument the apparatus already uses to make organizational structure legible, into the layer the apparatus does not contain.

06 · Why this matters now

For organizations introducing artificial intelligence, the first specification is not the model, the vendor, or the use case. The first specification is decision authority. Without it, every downstream choice is made against an undefined frame.

For boards, the implication is direct. Board oversight of artificial intelligence initiatives requires legibility into where automated decisions are being made and who owns them. Without a specified Command Chart, board oversight of AI is an exercise in trust rather than governance. Approval without specification is governance theater.

07 · The test

Examine the next artificial intelligence initiative on your organization's roadmap. Identify, in writing, where decision authority for the automated decisions in that initiative is specified.

If the specification exists and is documented, the initiative is operating against a defined frame. If the specification does not exist, the initiative is operating inside the apparatus gap.

The apparatus gap is the diagnosis. Command Mapping is the discipline. The Command Chart is the artifact. Your next initiative is the test.

The full paper.

Eight sections. Three altitudes of empirical evidence. Five sources cited. The complete argument.

The next step.

Engage Mesa Group to map the decision authority of your next artificial intelligence initiative.

Begin