Execra — Theory of Operations

How the execution OS turns raw events from your connected systems into auditable outcomes, with attribution, control, and no chat-first overhead.

1. The pipeline in one picture

Every unit of work Execra does passes these five stages. The UI is a direct rendering of the pipeline — the Feed is the event stream, Tasks is the action queue, Impact is the outcome rollup, Attribution is the credit ledger.

2. Events — what enters the system

3. The autonomy dial — you are in control

Every workflow has its own autonomy tier. Start conservative per workflow; graduate as trust grows. Default for every new workflow is Observe.

4. Actions — what Execra does

An action is a discrete, typed operation: send an SMS, reschedule a booking, refund a charge, open a task, post to Slack, call an outbound API. Every action carries a reason — the rule or signal that triggered it — and is reversible within its action type's defined undo window.

Guardrails

• Rate limits per workflow and per customer — configurable in Workflows.
• Blast-radius caps — an autonomous workflow cannot exceed N actions per hour without escalating to a human.
• Kill switch — flip any workflow back to Observe from the Workflows page; change is immediate.

5. Outcomes — what changed in the world

Execra links each action back to a measurable outcome: recovered revenue, avoided cost, time saved, a conversion, a schedule saved. You define outcome definitions per workflow; Execra computes them against your connected source-of-truth system (not a second ledger).

6. Attribution ledger — who gets credit

Every outcome is assigned credit across the human operator, Execra, and any connected systems that contributed. Default splits:

• 100% Execra when a Fully Autonomous action produced the outcome.
• 50/50 human / Execra when a human approved the action Execra proposed.
• 100% human when a human overrode Execra and took a different action that produced the outcome.
• Custom — configurable per workflow.

Attribution is informational — it is not a legal determination of commission or liability. Use it to reason about contribution and to compensate operators fairly.

7. Audit — PolicyLog

Every autonomy-policy change, every manual override, every export, every admin action is recorded in PolicyLog with timestamp, actor, and (where provided) reason. Retained seven years.

8. Replay — rehearse without touching production

Replay re-runs an event through the current workflow catalog and shows what Execra would do today, without executing. Useful for training, for regulatory review, and for validating a workflow change before you flip a tier up.

9. Patent-IP boundary

The autonomy dial, attribution ledger, and cross-product conductor are patent-pending. The public UI surfaces the behavior; internal ranking, classification, and cross-tenant signal logic are not exposed.