Practice-anchored questions across three organisational layers, hardened by telemetry, resolved into one live AQ Coordinate. Under 40 minutes per respondent.
The organisation completes the Scoping Input (Template I-1): standard (AQDS/AQOS), units in scope, breadth units declared, telemetry sources available.
Respondents mapped to the three layers (Template I-2): Executive · Management · Practitioner. Minimums: 3 / 5 / 8 for a High-confidence read; any layer below minimum caps confidence at Medium.
Each respondent takes an adaptive questionnaire: domain locators first, then depth questions only around their located level. Median session: 25–40 minutes.
Attestation is cross-checked against telemetry bindings and artefact evidence (Templates I-3/I-4). Divergence between said and measured is flagged per domain.
The scoring engine computes domain levels, P and B, the organisational AQ Coordinate and AQ, with confidence and perception-gap analysis.
The output template is generated: scorecard, domain profiles, gap analysis, gate plan; advisor review optional before issue. See the sample report ↗
The same cell asks differently per layer — Executives are asked about mandate and accountability, Management about mechanism and cadence, Practitioners about lived reality. The divergence across layers is the perception-gap input.
One per domain. Each locator is a ladder question whose options are condensed level bands; the answer places the domain in a provisional band.
For each domain, the engine asks the depth questions of the provisional level and one level below (cumulative rule) and one above (headroom check). Confirmed = all lower-level conditions hold.
For the confirmed level: the level's gate checklist (which of the 10 gates are cleared → P) and the breadth counter (coverage across declared units → B).
A domain that locates at L0–L1 skips depth beyond L2 (no wasted questions); total questions per respondent stay within 60–90 depending on profile; hard cap 40 minutes with save-and-resume.
The full Closed-Tier bank extends each domain to 2–4 depth questions per active level using the same pattern.
"Who does the work in your delivery organisation today?"
"How are roles designed where humans and AI work together?"
"What may an AI system in your organisation decide without a human?"
"When an agent acts outside its boundary, what happens next?"
"When an AI system needs organisational knowledge, where does it come from?"
"How fresh and reliable is the context your AI tools draw on?"
"How are AI models chosen and run for your work?"
"How do you know a model change won't degrade a running service?"
"How is the contribution of AI to outcomes measured?"
"Could you defend the AI contribution number in a board meeting?"
"How do clients and partners experience your hybrid delivery?"
"When your delivery includes agents, how is that agreed with the client?"
"What do your contracts say about AI and autonomy?"
"If an agent's action caused a client loss tomorrow, is accountability already written down?"
"How are AI/agentic solutions built and approved for use?"
"Before an agent gets real autonomy, what does it have to pass?"
"Who runs your services day to day?"
"When something breaks at 3 a.m., what usually happens before a human is involved?"
"How do your AI capabilities improve once live?"
"Describe the path from 'the agent could do better' to 'the improved agent is live'."
"How prepared are you for AI-specific attacks and failures?"
"If autonomy had to be switched off in one service right now, what would happen?"
"If an engagement ended next quarter, what could the client take with them?"
"How early in an engagement is exit portability designed?"
DomainLevel(d) = highest L where the locator band and all depth confirmations at L and below hold (cumulative rule).
P(d) = gates cleared at DomainLevel ÷ 10 × 10 → 0–10.
B(d) = breadth counter coverage across declared units → 0–10.
L(org) = median of the twelve DomainLevels.
P(org) = mean P of domains at L(org) · B(org) = mean B of domains at L(org).
AQ = (L × 10) + (0.6·P + 0.4·B) × (10/11)
L(org) ≤ min(gov, ar) + 1 — autonomy may never run more than one level ahead of governance and assurance.
Attestation resolves per layer (E/M/P) unweighted; the published score uses the triangulated resolution (attestation × telemetry × artefacts), while the layer split and its gaps are reported, never averaged away.
High (all layer minimums met, telemetry connected, low variance) · Medium (one shortfall) · Low (two or more) — always printed beside the AQ.
Per respondent: name (held encrypted) · unit · layer (Executive / Management / Practitioner) · channel (email / link / messaging / voice).
Individual responses are encrypted; only aggregates are ever displayed.
Per connected source: signal name · cell bindings (e.g., boundary-breach rate → gov·L6) · period · collection method (API / export).
A "no telemetry" path is supported — it caps certification standing until connected.
Per artefact: type (autonomy charter / safety case / policy / contract clause / audit trail) · cells evidenced · date · reviewer.
AI-first review, human verification for certification-grade standing.
See what the instrument produces.
Read the sample reportThe specimens on this page are the published subset of the AutonowmyQ instrument (v1.0). The full question bank, gate definitions, telemetry bindings and calibration data are Closed-Tier and available only under licence. AutonowmyQ™, AQ™ (Autonowmy Quotient), AQDS™, AQOS™, AQCS™ and AQ Certified™ are marks of Autonowmy Technology Pvt Ltd; registrations pending.