How many hours a week does a part-time Steward need?

There is no fixed number. The test is whether every intervention the Intervention Threshold allows can be handled inside the Attention Budget without delay. If not, the mandate is too wide.

What if an intervention arrives during working hours?

The Exception Architecture should hold it safely until the Steward is available, or the decision class that produced it should sit outside the mandate.

Does a small business carry less liability?

Its exposure may be lower when its decisions are few and reversible, but small scale does not remove legal responsibility for the decisions it makes.

The Mandate Fits the Hours

Attention Budget is the Steward time a person can reliably commit to governing an autonomous business in a given period. For a full-time Steward it is set by the business. For a person who keeps their job it is set first, by the rest of their working life, and the business has to fit inside it.

Attention Budget — The Steward time a person can reliably commit to governing an autonomous business in a given period, set by the rest of their working life rather than by the business — the constraint that caps the Mandate Scope a part-time Steward can hold, judged by the interventions that cannot wait rather than by average MTTI.

The target is necessary, not sufficient

The Log has used one number as the test of a working autonomous business since its early memos.

MTTI (Mean Time to Intervention) — The average time between required human interventions in an agentic system. Arco's target is greater than 72 hours.

A business that needs its Steward less than once every three days was described as a sign of Architectural Certainty. Read from the other side, it is a precondition for a business a person with a full-time job can govern, not proof of one. An autonomous business can be compatible with full-time employment when its mandate and safeguards are designed around the Steward's actual availability. Evenings and weekends can be enough, but only when every intervention the business generates can wait for them.

The limit is attention, not time

The condition is harder than it sounds. A part-time Steward does not lack hours in total; they lack attention at the moment the business needs it. An intervention that arrives during the working day waits. One that needs a long investigation is deferred to the weekend. The risk is not that the business fails loudly. It is that the Steward stops reading.

Nominal MTTI — The condition in which a system's measured Mean Time to Intervention is long not because it has achieved genuine Architectural Certainty but because the Steward has stopped engaging with the audit surface — producing a metric that appears to confirm autonomous operation while the system is in fact running unmonitored and unrefined.

A part-time Steward is more exposed to this than a full-time one, because the job always has a prior claim on their attention. The Steward Was Not Slow. The Steward Was Blind. (#127) described the same failure arriving through the wrong channel. Here it arrives through a fixed budget.

The answer is to size the business to the budget. Size Was Never Headcount (#151) made Mandate Scope the measure of a business's size. For a part-time Steward, Mandate Scope is set by the Attention Budget: the decision types, the Intervention Threshold and the Audit Surface all have to fit inside the time the person actually has. A mandate that would be right for a full-time Steward is too wide for this one.

MTTI is an average, and a part-time Steward is exposed to the tail. One intervention that cannot wait arrives during a working day, and the average says nothing about it.

Recovery Latency — The time from failure detection to restored autonomous operation — distinct from MTTI, which measures the average interval between required interventions rather than the duration of resolving any single one.

The sizing rule follows: any decision class whose wrong outcome cannot safely wait until the Steward's next available window sits outside the mandate. The Exception Architecture must be able to hold everything else until then.

The job is the source

The Drag Moved (#152) left open where a solo builder's judgment comes from. The Markets Too Small to Be Taken (#153) left open who sees sub-scale markets at all. The answer to both is the same, and it is the reason this route is open to professionals in particular.

The person currently doing a labour-heavy task by hand knows which parts of it are deterministic, where the exceptions come from, and which customers are badly served. That is the knowledge that sets a mandate and an Intervention Threshold correctly. It is also the knowledge that reveals a sub-scale market. A professional's job is where both are acquired, which makes keeping the job a source of the business's judgment, not only of the person's income.

The same knowledge also tells the professional which parts of the work the law reserves to them, and those parts stay with the person, not the mandate.

One caution follows directly. Knowledge acquired in employment is often bound by the employment contract. A business that serves the employer's customers or uses the employer's methods may conflict with it. That has to be resolved before the build, not after.

What the person risks

Income Is Not Yet a Business (#150) left open what a person risks by trying. Some risks fall sharply. The person keeps their salary, so income risk falls. Career risk changes form instead: a conflict with the employer, or a public failure in the person's own professional field, can cost the job or the reputation that keeps them employable. Purchased infrastructure keeps capital at risk small.

One risk does not fall at all. A business's decisions carry legal consequences whatever its size, and depending on the entity and jurisdiction, the people behind it can be answerable for them.

Accountability Trace — The specific subset of the Proof of Action record structured to establish legal and operational accountability for an autonomous decision after the fact — distinguishing routine operational logging, which exists to support debugging and quality review, from the evidentiary standard an actual liability claim requires.

A narrow mandate with cheap reversals can keep that exposure smaller. It does not remove it. A person who builds beside a job should hold the business in an entity, keep an Accountability Trace from the first transaction, and price insurance for the decisions the mandate allows. Silent AI, and Who Pays When the Silence Ends (#108) set out why the last of these matters.

The Operator's Verdict

Meeting the Log's own MTTI target is necessary for a business to fit beside a job, but not sufficient. A mandate sized for a full-time Steward does not fit at all. The part-time Steward's Attention Budget is fixed first, and the mandate, the threshold and the audit surface are sized to it. Their job is where the judgment and the market come from. What they risk is less their salary or capital than their standing with their employer and their liability, which a narrow mandate can keep smaller and never removes.

Technology changes what is possible. The hours left after work decide how much the business may be allowed to decide.

KEY TAKEAWAY

Can someone run an autonomous business without leaving their job?

Yes, if the business is sized to their Attention Budget. Arco's MTTI target of more than 72 hours between interventions is necessary but not sufficient: the business is compatible with a full-time job only if the Mandate Scope, Intervention Threshold and Audit Surface fit the time the person reliably has; decisions that cannot wait for the Steward's next window stay outside the mandate, or MTTI becomes Nominal MTTI. The job is an asset: it is where the domain judgment and the view of sub-scale markets come from, subject to the employment contract. Income and capital risk fall, career risk changes form, and legal liability does not fall at all, so the business needs an entity, an Accountability Trace and insurance from the start.