Glossary

Every Rocket Routine OS term in one place — explained briefly.

Also see: Strategy & Leadership Wiki

A

Adoption Levels: The three levels Shadow, Copilot and Autopilot. You set the level for each AI: whether it suggests, works alongside you, or takes care of routine work on its own. How often its work is right the first time (FTT) is the evidence you use to decide whether it gets more freedom; a person decides.
AI Agent: An AI teammate with a defined role, for example developer, tester, content marketer or customer success. It does the hands-on work and can be talked to directly. You set its adoption level, and it acts with the rights of the person who asked. The AI heads of the areas are AI agents too. AI agents carry the execution load, not decision accountability.
Alignment Check: An automatic comparison of a new or changed goal, outcome or decision with the company's values, principles and mission. The result is shown as a marker.
Autopilot: The highest adoption level: the AI also does structural things itself and tells you in the chat. At every level it asks before deleting anything or changing the company's direction.

B

Build Log: The monthly bilingual email (never called "Newsletter" in copy); reports the month's article of the month, key takeaways, a Company 0 update, and what's next.

C

Cadence: The rhythm of the recurring meetings: Daily, Weekly, Monthly, Quarterly (or Trimester) and Annual. All are on by default, the Navigator leads them, and you switch off what you do not need. Any agenda can be defined freely, for example the EOS agenda.
CEO-Bench: Princeton benchmark (ceobench.com) that gave 13 AI models and one rule-based baseline $1M each to run a simulated company for 500 days, used as evidence that autonomy without governance fails unpredictably.
Company 0: Rocket Routine GmbH itself, run live on Rocket Routine OS as the first production environment and proof point (always written with a space, never "Company0").
Constitution: The company's rules for how AI and people decide. The alignment check and agents use it: a new or changed goal, outcome or decision is compared with your values and principles, and the result is shown as a marker.
Constitution Lite: The starting, minimal form of the Constitution used from day one: a handful of core rules plus a map of who decides what.
Control Tower: The work board: work items from Backlog to Released, ordered by status, with exceptions (blocked or escalated) in their own column. Work items hang directly on an outcome. Capitalized always.
Copilot: The middle adoption level: the AI does everyday things itself (setting a status, entering a metric value, writing a comment, adding someone to a chat) and tells you in the chat. For structural things, such as a new goal, it opens the filled-in form for you to confirm. The Navigator works at this level. Written as one word.

D

Decision Impact Classification: Every decision carries a rating of how far it reaches. What an AI agent may do on its own is decided by the adoption level and the kind of action.
Definition of Done: What "finished" means for a piece of work, as a checklist. The AI suggests it, and a person checks the result against it.
Domain: A part of the business with its own head and plan. There are ten fixed areas: Strategy, Marketing, Sales, Product, Engineering, Customer, Finance & Legal, People, Administration, and Quality & Excellence. Together they map any company completely, independent of headcount; distinct from a "department" (an org-chart/people concept). In text it is called an area; the settings page is named Domains. Every area has a VP and a responsible person.
Domain Advisory Board: Multiple Knowledge Personas combined for one area, giving the Navigator a synthesized (not smoothed-over) view.

F

FTT (First Time Through): How often work is right the first time: the percentage of work items passing the quality check on the first attempt without rework. FTT is the evidence you use to decide whether an AI gets more freedom, and it is part of the weekly figures. Always written in all caps.

I

Impact: The fourth of the four phases of Rocket Routine OS: verifying quality and writing Learning that updates the system when reality diverges from Intent.
Implementation: The third of the four phases of Rocket Routine OS: executing routines, routed through the Control Tower, linked back to the originating Intent.
Inbox: The one place where you approve things and receive requests and reports. Inbox is a Rocket Routine OS product word. "My Inbox" shows the items that are yours. Posts for LinkedIn, Instagram, YouTube and X wait there as drafts for your approval.
Insight: The second of the four phases of Rocket Routine OS: explicitly selecting the decision method (data, rule, procedure) before work begins.
Intent: The first of the four phases of Rocket Routine OS: defining outcomes and constraints as verifiable intent with explicit boundaries.

K

Key figures and KPIs: A handful of figures computed before each Weekly and reviewed in the Monthly: how often work is right the first time, open work items, time from start to release, decision speed and overdue metric values.
Knowledge: Your company's knowledge: documents and company knowledge are stored, indexed and searchable. Every agent finds the right passage and cites its source. Documents can be marked as a source of truth.
Knowledge Persona: An AI expert modelled on a thinker, for advice only. You create it by describing it or from any material. It knows your company's data, owns nothing, decides nothing and runs no routines. The Navigator brings it into a meeting when needed. "A Knowledge Persona speaks. It does not decide."

L

Learning: A lesson from a routine run, a meeting review or an unclear answer. A person accepts or discards it, and an accepted one becomes a concrete next step.

M

Maturity Ladder: Four stages of operational maturity — Ad hoc, Governed, Verified, Compounding — distinct from company growth in revenue or headcount.
Metrics: A metric is the number that measures progress or current state (for example the FTT). Each outcome has one metric with a target, warning lines and a traffic light. A metric has an owner (a person or an AI head) and a source: entered by hand, read from a connected tool, or calculated by an agent.

N

Navigator: Your AI guide. It leads every meeting and is your everyday chat partner for the whole system: tell it what you need, and it creates, changes, closes or looks up what the menus can, and hands specialist questions to the right VP.

O

Outcomes: An outcome is the measurable result that proves a goal. An outcome has one metric, and work items hang directly on an outcome. A project is an outcome with its work items.

P

Paperclip: Named open-source competitor: a self-hosted, agent-agnostic orchestration platform pursuing "zero-human companies," positioned as strong on execution/governance-of-config but lacking a methodology or verification layer.
PDCA: The loop the whole system runs. Plan: goals, outcomes and work items. Do: the work and the routines. Check: metrics, quality checks and the weekly figures. Act: what the team learned becomes concrete improvements.
Person: A person in the company: holds decision rights, accountability, and judgment; holds every decision right not explicitly delegated; the only sovereign actor under the Constitution. In the product every person is an Admin or a Member.
Poka Yoke: Toyota Production System principle, applied to how AI agents are configured: design the process so an error structurally cannot occur, instead of relying on human oversight to catch it after the fact.
Principles: Your company's principles: guidelines that actively steer decisions (for example "data beats opinions") rather than hanging on a wall. The alignment check compares new and changed goals, outcomes and decisions with them.

Q

Question Card and Result Card: A question card is a small form in the chat with which an agent gets structured answers (text, number with unit, date, yes or no, a choice, a person or area picker; never secrets). "Later" moves the question to your Inbox. A result card shows a chart, figure or table in the chat, read-only.

R

Responsible Person: The person who owns an area and for whom its agents work. By default the VP is an AI and the responsible person is human. They receive the area's escalations and check its quality.
Routines: A routine is a schedule that runs a workflow, for example a daily AI news digest or an intake that turns channels such as email into draft tasks. A workflow is a described procedure. Failed runs reach the owner at once.

S

Shadow: The lowest adoption level: the AI only suggests. A form opens, and nothing is saved until you confirm. You set the level for each AI.
Source of Truth: A document you mark as the authority. Every AI agent reads it and quotes it exactly, for example for pricing and business model.

T

Team: A saved group of AI agents and Knowledge Personas; one click starts a chat with the whole group. People are added per chat.

V

Verification-first: The principle of checking work against defined criteria before it counts as finished. In Rocket Routine OS you mark a work item as needing a quality check ("Quality check required"). A person, never the one who did the work, then checks the result against its definition of done and either accepts it or sends it back with a reason.
Virtual CEO: The AI agent at the top that looks across all areas. It supports you and does not replace you. It asks the VPs questions and hands requests to them.
VP (AI Head): The AI head is the AI agent that leads an area. In the app it is called the VP, for example VP Sales. It knows its area's goals, metrics, work and meetings, answers "How are we doing?" with exact numbers and drafts its area's plan; you decide. A company can also make a person the VP of an area. The AI head then stays on as that person's assistant.

W

Weekly Operating Review: The Weekly: the weekly meeting the Navigator leads. It shows which metrics crossed a warning line this week (with the latest value, trend and traffic light, and names any value that is still missing), whether the weekly figures show a problem (for example how often work is right the first time, open work items, time from start to release, decision speed), whether the quality matches what was agreed, and what the team learned that should change a routine. At the end of each topic the Navigator offers a work item and asks who owns it and when it is due.
Work Item: One piece of work, always linked to an area and an owner, usually to an outcome. Work items run on the Control Tower from Backlog to Released. A project is an outcome with its work items.