The bar every production automation clears.
An open, versioned definition of what it means for an AI automation to be production-ready: the kind your team acts on and ships to clients and investors. This is the line we hold every build to, published in full.
How grading and certification work
Graded against published checkpoints
Every automation is assessed against the same 30 checkpoints, grouped into 8 disciplines. Each has one definition of passing. Nothing is hidden, the bar is the same for everyone.
Automated assessment by default
A snapshot grade is produced by our grading engine reading the automation you provide: code, a conversation, a workflow, or a description. Snapshot grades are labeled as automated assessments.
Certification is earned, not granted
A certificate is issued only when all 30 checkpoints pass and hold for 60 days. It then re-verifies, so a live certificate always reflects current state.
A quality mark, not compliance
This is a production-readiness quality mark, like a code-quality grade. It is not a security, legal, or regulatory compliance certification, and it is not a warranty.
30 checkpoints, 8 disciplines
Each discipline maps to a way output goes wrong before it reaches a client. Open one to see the checkpoints inside.
- 01Row Count ReconciliationNo rows quietly go missing. A 240-unit rent roll comes out as 240 units, not 228.
- 02Schema Drift DetectionIf a source renames or moves a column, the run stops instead of putting rent in the expense field.
- 03Unit and Currency NormalizationMonthly versus annual and dollars versus thousands never get mixed up, so NOI is not off by 12x.
- 04Period and Date AlignmentA trailing-12 and a current rent roll are not blended across different dates into a wrong NOI.
- 05Null and Sentinel HandlingBlank or 'N/A' values are treated as missing, not silently turned into $0.
- 06Deduplication and IdentityThe same property or comp is not counted twice because the address was written two ways.
- 07Input ProvenanceEvery number can be traced back to the document and cell it came from when someone asks.
- 08Source-of-Truth Conflict ResolutionWhen the rent roll, T-12, and OM disagree, it flags the conflict instead of silently picking one.
- 09Document-Ingestion RobustnessIt handles messy real files (scans, phone photos, merged cells) or refuses them. It never guesses a number.
- 10Idempotency and Re-Run SafetyRunning the same job twice does not create duplicates or double-send anything.
- 11Scraper Layout-Change DetectionIf LoopNet or a county site changes its page, you get alerted instead of pulling garbage comps.
- 12Graceful Source DegradationOne source going down does not crash the whole run or store a blank as if it were real data.
- 13Partial-Failure IsolationOne bad file in a batch is set aside so the rest still finish.
- 14Timeout and Hang ProtectionA slow source cannot hang the whole job forever.
- 15Liveness AlertingIf the job stops running, you are told. Silence is not treated as everything-is-fine.
- 16Volume Anomaly DetectionIf a run processes 4 deals instead of the usual 80, it flags it instead of passing green.
- 17Data Freshness and Comp DecayStale data and six-month-old comps are flagged before you underwrite on them.
- 18Actionable Alert RoutingImportant alerts reach a person by text or Slack, not a log nobody reads.
- 19Numeric Source-TracingThe model cannot invent a cap rate or NOI that no document supports.
- 20Arithmetic in Code, Not ProseThe math runs in code, so the same deal always produces the same number.
- 21Regression Eval SetBefore any change ships, it is tested against known-good deals so accuracy does not quietly drop.
- 22Abstention on Unreadable InputOn a blurry scan it says 'I cannot read this' instead of confidently guessing.
- 23Output Range and Sanity BoundsImpossible numbers (140% occupancy, $4/SF office rent) get flagged, not passed downstream.
- 24Human Approval GatesNothing irreversible (an LOI, an offer, a tenant message) goes out without a person approving it.
- 25Model and Dependency PinningThe AI model is pinned, so an overnight provider change does not silently shift your outputs.
- 26Secret ManagementAPI keys and passwords live in a secure place, never hardcoded in the code or repo.
- 27Least-Privilege AccessThe tool only has the access it needs, read-only where it just reads, not admin keys to everything.
- 28PII and Untrusted-Input HandlingTenant SSNs and bank details are masked, and a malicious document cannot hijack the tool.
- 29Runbook, Ownership, and ConfigSettings live in one place a non-engineer can change, with a doc so it is not stuck in one person's head.
- 30Load Behavior and Cost CeilingsIt is tested at real volume and has spend caps, so a runaway loop cannot burn your budget overnight.
Versioning and provenance
This is Production Standard v1.0.0. The checkpoint set is versioned: when the bar changes, the version increments and prior certificates record the version they were issued under, so historical grades stay reproducible. Certificates are issued by LunoMotion and are publicly verifiable at their certificate link.
The Production Standard is a production-readiness quality mark issued by LunoMotion. A grade or certificate is an assessment of the automation as provided to us, not a security, legal, or compliance certification, and not a warranty of fitness or correctness. See terms.