Consultation edition 0.51 · 6 November 2026
AI Native Business Benchmark
The AI Native Business Benchmark measures how much useful value a business produces for every hour of human involvement, and whether that holds or improves over time. It is open and vendor neutral: any business can run it with any AI tools. aBOS is one implementation, not a requirement.
This is a consultation and pilot edition. The metrics, thresholds and comparisons still require independent validation. It is not a certification, a ranking, or a validated standard.
A licence has not been granted. Until then, "open" describes the intent, not a licence grant. The draft has not been publicly released or assigned final licenses.
Founding sponsor: Alien Cortex (aliencortex.com), credited without preferential scoring or control of the rules.
Value per Human Hour
The headline is realized value divided by all human hours, for the same scope and period. For most businesses, value is gross profit. A trajectory is the change against that business's own first period.
- Value per Human Hour Realized value divided by all attributable human hours. Written as V / H.
- Human minutes per $1,000 The same result inverted: 60,000 x H / V. Publish it only when value is positive.
- Who counts in the hours Human hours include everyone: owners, staff, contractors and outsourced services, plus time spent prompting, reviewing, approving and fixing AI work.
No headline without its guardrails
- Quality status One of: qualified, breached, insufficient evidence, or provisional. A rising headline with a breached quality condition is not an improvement claim.
- Full AI and program cost Show the full AI and program cost beside the headline, so a gain bought with runaway spend stays visible.
- AI operating coverage The share of the reference workload completed by qualified AI execution.
The headline is a communication aid. It is not a composite maturity score, and it is not used here to rank unlike businesses. Observed change is not, by itself, evidence that AI caused the change.
Lite track
The Lite track is for a small business without a data team. It uses the same definitions, with proportionate evidence. Results are self-reported, for that business's own trend. Do not compare Lite results between companies or publish rankings from them.
- Freeze the boundary before you start: the entity, one full quarter, what counts as value, and whose time counts.
- Fill in one quarter: value, all human hours and how you counted them, AI cost, and a quality status.
- Optionally add the top 3 workflows.
- Add a second quarter after 90 days, under the same rules. Report the change against the first period. If the rules changed, start a new series.
Downloads
These are the consultation edition 0.51 files. The starter kit runs locally, uses only the Python standard library, and uploads nothing. It is an unvalidated reference calculator prototype.
- Lite calculator Executable 0.5.1-prototype, input schema lite-0.5-1, for edition 0.51. It computes Value per Human Hour, Human minutes per $1,000, coverage for the workflows you list, and a descriptive change against the first period.
- Full-track calculator Method 0.3 core, executable 0.1.0-prototype, input schema 0.3-prototype-1. Unchanged and compatible with edition 0.51. Full-track reports still say working_draft 0.3. That label is intentional.
Neither calculator assesses statistical trajectory, uncertainty or AI attribution. Passing the included tests checks a limited set of implementation properties. It does not validate the method or certify a business.
Open, and not tied to one vendor
Any company must be able to run a released version without an aBOS account, proprietary endpoint or permission from a vendor. No normative calculation may depend on an exclusive commercial service.
Disclosed interest
aBOS (Alien Business Operating System) is the commercial product of Alien Cortex, the founding sponsor. It installs a private team of AI agents for a business, with people approving consequential actions and finance work kept on local hardware. Because the sponsor sells aBOS, this section is a disclosed interest, not an endorsement.
Nothing in the protocol requires aBOS, its agent structure, hardware, local-first policy or commercial model. Other systems must be able to conform under the same evidence rules.
Pilot participation and comments
Pilot participation, the starter kit and updates to this consultation are coordinated through aliencortex.com. Comments are welcome from any business, provider or reviewer, and every comment is considered under the same open process (section 16).
Anyone may propose a defect, improvement or profile without being a customer. The sponsor's pilot proves instrument feasibility only. Industry validation must include competing implementations and organizations outside the sponsor's network.
Received
Received. This confirms the form was accepted. It is not a pilot acceptance, a score, or a certification.
Your reference is . Quote it if you write again.