Why is a spectrum better than asking whether a business is autonomous or not?

Because the binary has no place to put real businesses, which are autonomous in some dimensions and human-dependent in others. A spectrum reading across independent axes can be inspected, disputed, and falsified — which is what makes it a measurement rather than a marketing claim.

Can a business score highly on one axis and low on another?

Yes, and the profile is more informative than the composite. A business with high Task Execution Autonomy (TEA) and low Decision Execution Autonomy (DEA) has automated its work while keeping its decisions human — the Judgment Layer / Execution Layer binary uncompleted. The Escalation Rate in that business will be high relative to its task execution volume, because every task that completes still requires a human decision about what happens next.

Who can apply the Autonomy Spectrum Framework?

Anyone. The axes, scoring semantics, and classification bands are published in the Arco Lexicon, and several axes can be scored from publicly available information. The framework is designed to be applied by operators, acquirers, and analysts without insider access.

What is the relationship between Intervention Dependency (ID) and MTTI (Mean Time to Intervention)?

Intervention Dependency is one of the five scored axes; MTTI is the operational metric that measures what the Intervention Dependency score predicts. A business with a low Intervention Dependency score requires frequent human intervention and will have a short MTTI. A business engineering toward the Architecturally Autonomous band targets an MTTI measured in days or weeks. The Intervention Threshold set at design time governs what triggers an escalation; the Escalation Rate confirms whether the threshold was calibrated correctly; MTTI is the duration measurement that results from both.

How does the Autonomy Spectrum Framework connect to the previous Arco vocabulary?

Each axis maps directly to existing terms and published memos. Task Execution Autonomy (TEA) maps to Labor-to-Compute Substitution and the Human-to-Logic Ratio. Decision Execution Autonomy (DEA) maps to the Judgment Layer / Execution Layer separation and the Exception Architecture that governs it. Process Continuity Score (PCS) measures what the Automation Paradox disrupts — workflows that halt at coordination points rather than flowing end-to-end. Intervention Dependency (ID) maps to MTTI (Mean Time to Intervention) and the Escalation Rate. Structural Headcount Independence (SHI) maps to Headcount Decoupling and Inverse Complexity Scaling. The framework does not introduce new concepts. It gives the existing vocabulary a shared scale.

Autonomy Is a Claim Until It Is Measured

Every company in the market now describes itself as AI-powered. The phrase costs nothing to claim and nothing to defend, because it is unfalsifiable — there is no measurement a sceptic can run against it, no threshold it must clear, no score it can fail. A claim that cannot fail is not a claim. It is decoration.

We have spent several dozen memos building the vocabulary of Autonomous Business design: what one is, which markets can support one, what infrastructure it runs on, how its failures are designed, how it acquires customers, and how it prices what it delivers. Under the Stewardship Model, a single operator governs an architecture that handles execution autonomously. If the distinction between an Automated Business and an autonomous one is real — and the entire Arco thesis rests on it being real — then it must be measurable. A distinction that cannot be measured is an opinion.

The wrong question

The industry's instinct is to ask a binary question: is this business autonomous or not? The binary fails immediately on contact with reality. A business can run its support operation on agents and its sales operation on people. It can execute tasks autonomously while a human approves every decision. It can run unattended for a week and then require three days of intervention. None of these businesses is autonomous, and none of them is not. The binary has no place to put them — so the market answers the binary with marketing, and every company that has deployed a chatbot calls itself autonomous.

The correct instrument is a spectrum. The Autonomy Spectrum Framework is that instrument: a five-axis scoring system, each axis scored 0 to 2, producing a composite score from 0 to 10 that places any business in one of four classification bands. Autonomy is a property a business has more or less of, along dimensions that can be inspected independently. A spectrum reading can be wrong, which is precisely what makes it useful: a claim that can be wrong is a claim that means something.

The five axes

The framework scores a business across five axes, each scored 0 to 2, producing a composite from 0 to 10.

The first axis is Task Execution Autonomy (TEA) — the proportion of a business's revenue-generating task volume executed by agents rather than people. It asks who moves the work.

The second is Decision Execution Autonomy (DEA) — the proportion of operational decisions made by encoded logic rather than human judgment. It asks who decides how work is routed, prioritised, escalated, and resolved. A business with high Task Execution Autonomy and low Decision Execution Autonomy has automated its work while keeping its decisions human — an incomplete Judgment Layer / Execution Layer separation, where agents run the work but humans still govern every transition.

The third is the Process Continuity Score (PCS) — whether workflows run end-to-end without halting for human input. It asks whether the work flows or waits.

The fourth is Intervention Dependency (ID) — the frequency with which core systems require human intervention to continue operating. It is measured by the Escalation Rate and governed by the Intervention Threshold that was set at design time. It asks how long the system runs before it needs a person — which is what MTTI (Mean Time to Intervention) tracks as an operational metric.

The fifth is Structural Headcount Independence (SHI) — the degree to which revenue growth is decoupled from headcount growth. It measures Headcount Decoupling directly: whether the business can increase output without proportional hiring. It asks whether growth requires people or only compute.

Each axis isolates one dimension and is deliberately blind to the others. That separation is the point. The most common autonomy misreadings — the tool-saturated business that thinks it is autonomous, the agent-heavy business where humans approve everything — are each visible as a high score on one axis and a low score on another. The composite tells you how autonomous a business is. The profile tells you where it is not.

The four bands

The composite score places a business in one of four classification bands. A score of 0 to 3 is Automated: technology accelerates human execution, and humans remain the operating system of the business. A score of 4 to 6 is Transitional: agents own material work, but the architecture still depends on people at its joints. A score of 7 to 8 is Operationally Autonomous: the core operation runs without human execution or routine intervention. A score of 9 to 10 is Architecturally Autonomous: the business was engineered for autonomy from first principles, and every axis confirms it — the condition the Arco body of work has called Architectural Certainty.

The bands are deliberately strict. Most businesses describing themselves as AI-native today score in the Automated band, because tool adoption — however enthusiastic — transfers no execution, no decisions, and no continuity. The framework is not designed to flatter the market. It is designed to sort it.

What a published instrument changes

A framework that lives in our heads is no better than the marketing claims it criticises. Publishing the instrument changes its nature: the axes are defined in the Lexicon, the scoring semantics are fixed, and anyone — including a sceptic — can apply the framework to any business, including ours, and check the result. The next memos in this series take each axis in turn, because an instrument is only as credible as the precision of its dimensions. After that, we will address the way the framework can be gamed, and what a composite score is actually worth to the people who will use it.

A measurement system invites a measurement. That is the point of building one.

The Operator's Verdict

The autonomy claim is currently free, which is why everyone makes it. A published scale makes the claim expensive: it must now survive five independent inspections, each of which can return a zero. Businesses that are genuinely autonomous gain from this — measurement is the only thing that separates them from the decoration around them.

Technology changes what is possible. Measurement determines what is credible.

KEY TAKEAWAY

How do you measure whether a business is autonomous?

Business autonomy is measured on the Autonomy Spectrum Framework, which scores a company from 0 to 10 across five axes: Task Execution Autonomy (TEA) — who executes the revenue-generating work; Decision Execution Autonomy (DEA) — who makes operational decisions, expressed as the Judgment Layer / Execution Layer separation; Process Continuity Score (PCS) — whether workflows run end-to-end without human input; Intervention Dependency (ID) — how often the system requires human intervention, measured by the Escalation Rate against the designed Intervention Threshold and tracked over time as MTTI (Mean Time to Intervention); and Structural Headcount Independence (SHI) — whether revenue growth is decoupled from headcount growth. Each axis is scored 0 to 2, and the composite places the business in one of four bands: Automated (0–3), Transitional (4–6), Operationally Autonomous (7–8), or Architecturally Autonomous (9–10). The Architecturally Autonomous band corresponds to Architectural Certainty — the state in which every axis confirms autonomy from first principles. Source: Arco Venture Studio, arcoventure.studio.