This document is our internal operating manual. It is how we become the global Business AI Operator: by living these values, running these rituals, and refusing these anti-patterns every single day.

1. Core Values

A. Customer economic outcomes are the product; time-to-value is sacred

Intent: Every decision is tied to a real customer problem and a measurable economic outcome (revenue, efficiency, reliability). We build for the many, a platform over bespoke.

Guardrails:

• No engineering without:
• (a) Customer Problem Statement
• (b) Reach & Revenue hypothesis (target reach ≥ 30% of active accounts or a named core segment)
• (c) Packaging & Pricing decided up front
• (d) Adoption Plan (90-day)

Do:

• Run 5 customer touchpoints per PM per month (calls, shadowing, or usability tests).
• Maintain a Top-10 Customer Problems backlog; each WBR pulls from it.

Don’t:

• Build features without an adoption plan or success metric.
• Close tickets without root-cause + prevent-recurrence note.

Measure:

• TTFV (days)
• Activation → Value rate
• Feature WAU/MAU 30 days post-launch
• Proof of revenue uplift, cost saving or reliability improvement where possible



B. Value-led growth is our scoreboard; evolve fast

Intent: Execute–learn–iterate as our default metabolism. We grow only when we create real, measurable value for customers.

Guardrails:

• Reversible decisions → move fast
• Irreversible decisions → quality gate

Do:

• Kill sunk-cost work.
• Run continue/kill/transform reviews.
• Prefer reversible bets.

Don’t:

• Preserve sacred cows.
• Launch “vanity” features without adoption plans or clear value logic.

Measure:

• NRR
• Activation → Value time
• Adoption velocity
• Time-to-sunset for laggards
• % growth tied to specific, measured value levers



C. Reality lives in data; we decide on conviction and then measure

Quality > Outcome > Result > Effort; measure everything

Intent: Data is the source of truth for scaling; conviction unlocks exploration.

Guardrails:

• Follow the loop: Believe → Explore (≤7 days / cheap test) → Scale only on signal.
• Define QORE: Quality > Outcome > Result > Effort in reviews.

Do:

• Pre-register success metrics / MDE.
• Publish a 24-hour post-experiment note.
• Track QORE in reviews and promotion discussions.

Don’t:

• Cherry-pick or shift goalposts.
• Reward “effort theatre.”

Measure:

• % experiments time-boxed
• Idea → signal cycle time
• % scaled with proof
• % reviews referencing QORE



D. Challenge sharply, then commit together

Intent: Hear dissent before commitment. Challenge ideas, not people.

Guardrails:

• Collaboration before decision; disagree-and-commit after.
• Refusal to be reviewed = performance risk.

Do:

• Get two strong counters for major decisions.
• Ask naïve questions.
• Submit pre-reads.
• Invite a cross-functional review.

Don’t:

• Run side-DM politics.
• Stonewall or stall reviews.
• Ridicule or use status play.

Measure:

• decision logs with counters
• Re-litigation rate
• Review participation rate
• Blocked-review escalations



E. Communication is our product and our operating system

Intent: Clear, kind, direct—in writing—inside the company and for customers.

Guardrails:

• 5-sentence memo standard: Context → Options → Risks → Ask → DRI + When.
• Feedback duty (monthly micro-360) with actions logged.

Do:

• Memo-before-meeting.
• Timestamp decisions.
• Ask–accept–act on feedback within 7 days.

Don’t:

• Send vague pings.
• Gossip or triangulate.
• Make undocumented decisions.

Measure:

• % decisions with memo
• Feedback turnaround time
• Feedback acceptance rate
• % commitments closed



F. Skin in the game decides; owners own the decision

Intent: The DRI decides and is accountable.

Guardrails:

• The owner hears two counters before deciding.
• Decision is logged; team disagrees and commits.

Do:

• Name DRI + deadline everywhere.
• Publish decision notes & rationale.

Don’t:

• Allow shadow ownership.
• Reopen decisions without a new signal.

Measure:

• % tasks with DRI + due date
• Reopen rate
• % decisions logged



G. Run a marathon, focus on the vital few, automate the rest

Intent: Frugal excellence over time; ruthless focus (80/20). Automation and AI should handle repeatable work so humans focus on the few decisions that truly matter.

Guardrails:

• Automate if payback < 2 quarters or removes recurring toil.
• Maintain Top-3 priorities per team.

Do:

• Run a Weekly Business Review.
• Run a Monthly Kill Session.
• Track cost per conversation/call.
• Keep a live kill list.
• Invest in AI/automation for repeat processes.

Don’t:

• Allow manual-repeat ops.
• Spend time on “nice-to-have” work that’s not in the Top-3.

Measure:

• Automation coverage %
• Cost per conversation/call
• Gross margin trend
• % capacity on Top-3
• Items killed per month



H. Close the loop with decisive action

Intent: Finish the loop with clear decisions, ownership and action.

Guardrails:

• Definition of Done (DoD) = decision, owner, due date, source-of-truth link, and customer impact note.

Do:

• Close threads with outcomes and next check-in.
• Archive/link artefacts to the source of truth.

Don’t:

• “Circle back” without closure.
• Leave orphaned tasks.

Measure:

• Closure SLA
• % threads with DoD complete
• Orphaned-task rate

2. Anti-Values (zero-tolerance behaviours & enforcement)

These describe what we refuse to tolerate because it breaks our operating system.

A. Who/What can’t grow or evolve can’t be continued

Intent: Adaptive or retired.

Guardrails:

• Quarterly continue/kill/transform review for products, projects and processes.

Do:

• Sunset laggards.
• Retire sacred cows that no longer serve.

Enforcement & Signals:

• The owner must table a “continue/kill/transform” memo.
• Non-submission auto-flags for review.

Metrics:

• % of portfolio reviewed per quarter
• Median time-to-sunset for laggards



B. Who/What can’t collaborate or be questioned can’t be continued

Intent: We are a challenge-welcoming culture.

Guardrails:

• Refusal to be reviewed is a performance risk.

Do:

• Submit pre-reads.
• Surface hard questions.

Enforcement & Signals:

• Missing pre-reads → meeting cancelled.
• Repeated offences → manager escalation.

Metrics:

• Review participation rate
• Blocked-review escalations



C. You can’t give feedback; it’s your fault. People can’t give you feedback; it’s your fault

Intent: Coachability is a core skill.

Guardrails:

• Feedback duty (monthly 360 micro-loop); action items logged and revisited.

Do:

• Ask, accept, act.
• Reflect before defending.

Don’t:

• Defend ignorance.
• Ignore repeated feedback.

Enforcement & Signals:

• No action on feedback in 30 days → performance risk noted in review.

Metrics:

• Feedback acceptance rate
• % commitments closed

3. Conflict Rules (so values don’t collide)

To avoid confusion when values pull in different directions, we follow these conflict rules:

• Conviction vs Data: Try fast and cheap (≤7 days), then scale only on signal.
• Collaboration vs Ownership Debate openly → DRI decides → disagree-and-commit.
• Speed vs Quality Reversible → ship fast. Irreversible → gate behind QORE.
• Frugality vs Investment Automate when payback < 2 quarters or removes high-toil work.

4. Rituals (with owners & SLAs)

Our rituals make the values real. They are non-optional parts of how we operate.

A. Weekly Team Review (≤45 min)

• Owner: Team heads
• Cadence: Weekly

Inputs:

• NRR
• Activation → value
• Adoption
• Defects
• Cost per conversation/call
• Top-3 progress

Flow:

• 10’ metrics
• 15’ experiments (results → scale/kill)
• 15’ blockers & automation
• 5’ decisions recap

Outputs:

• Decision links (memos)
• Owner + due dates
• Updated kill list

SLA:

• Decisions logged in 24h
• Blockers assigned owners in-meeting



B. Decision Log (continuous, async-first)

• Owner: DRI
• Cadence: Continuous

Template:

• 5-sentence memo + two counters + DRI + due date + source-of-truth link.

SLA:

• Log within 24h of the decision.
• Close with DoD meta on completion.

System:

• Single index (Notion/Jira/Linear) with filters by team/date.



C. Monthly Kill Session (60 min)

• Owner: COO / Chief of Staff
• Cadence: Monthly

Prep:

• Each team proposes 3 things to kill, de-scope or automate.

Rules:

• Not in Top-3 and no signal → kill or transform.
• Create automation tickets with ROI & owner.

Output:

• Updated portfolio
• Published the stopped-doing list

SLA:

• Tickets created within 48h
• Follow-up in next WBR



D. Feedback Micro-360 (30 min)

• Owner: People Ops
• Cadence: Monthly

Flow:

• Ask–accept–act: 2 peers + 1 manager.
• “Start / Stop / Continue” (3 bullets).
• 1-page action plan.

SLA:

• Action items logged within 72h.
• Reviewed next month.

Metric Roll-up:

• Acceptance rate
• Commitments closed



E. Experiment Week Cadence

• Owner: PM / DRI
• Cadence: As needed

Design:

• Pre-registered metric / MDE.
• ≤7 days.
• Reversible.

Outcome:

• 24-hour write-up.
• Scale/kill / iterate decision recorded.

SLA:

• Experiment start → decision ≤10 days end-to-end.

5. Anti-People (do not hire or keep)

These behaviours violate our operating system and are grounds for no-hire, 30-day coaching with clear outcomes, or exit.

A. Mindset & Integrity

Victim-mindset players – externalise blame, low agency.

• Signals: excuses > ownership; “they” > “I”; patterns of helplessness.

Narcissists – self-serving, credit hoarding, empathy deficit.

• Signals: steals credit, blames others, hoards information.

Integrity risks – self-benefit at others’ cost.

• Signals: gossip, side deals, data access misuse, opaque decisions.



B. Competence & Learning

Dunning–Kruger operators – high confidence, low competence; resist correction.

• Signals: argues from certainty, fails basic work-sample, repeats errors.

Uncoachable – can’t give/receive feedback; no action after feedback.

• Signals: defensive replies, missed follow-ups, zero change logs.

Data-deniers – cherry-pick, shift goalposts, avoid measurement.

• Signals: no pre-registered metrics, post-hoc success stories.



C. Collaboration & Ownership

Non-collaborators – refuse review, run side-DM politics, block dissent.

• Signals: missing pre-reads, backchanneling, meeting theatrics.

Non-owners – dodge DRI & deadlines; reopen decisions without a new signal.

• Signals: “someone should…”, unclear ownership, churned threads.

Open-loopers – don’t meet the Definition of Done (DoD).

• Signals: threads without closure; artefacts missing links/owners/dates.



D. Execution & Product Fit

Pet-project crusaders – belief without data; refuse ≤7-day tests.

• Signals: “trust me” launches; no experiment notes; no kill criteria.

Single-logo customizers – push bespoke builds that won’t scale or lift ARPA.

• Signals: roadmap items for one account; no packaging/pricing; endless pilots.

Effort-theatre hustlers – high activity, low QORE; vanity launches.

• Signals: status spam; weak adoption; bugs/defects escaping.

Sacred-cow keepers – block continue/kill/transform; won’t evolve.

• Signals: aged features with near-zero usage; defensive “history” arguments.