Like a credit score for autonomy. AQ runs 0–100 — measured from how your teams actually work, not what the strategy deck says. Behind it, three plain facts tell you what to do next.
The Q is the point: as IQ measured intelligence for a person, AQ measures autonomy capability for an enterprise.
Humans, forward-deployed engineers, agentic development and IT agents now deliver together. Capability means orchestrating all four.
Services learn while running — improving, drifting, re-basing. A capability model must measure the loop, not a snapshot.
Autonomy without governance is risk. What an agent may decide, and who answers for it, is now a core capability.
AutonowmyQ is the standard that measures all three — continuously.
Cars have autonomy levels — everyone knows the ladder from cruise control to full self-driving. Companies have one too. AutonowmyQ measures where yours sits, from L0 (everything by hand) to L10 (it runs, you steer).
People do the work. AI is a personal helper at best.
Work is redesigned so people and AI deliver together.
Agents own whole tasks inside guardrails you set.
Strategy stays human. Execution largely runs itself.
Your level and score — one page your CEO reads in a minute.
Where you're strong and where you're exposed — twelve areas of the business, ranked, weakest flagged.
The next three moves — concrete actions with owners and a 90/180-day horizon.
Everything below is the full architecture behind the number — the twin models, the three-dimensional capability map, the delivery loop, the twelve domains and the scoring math. Skip it if you just want your score; read it if you're the one who has to defend it.
For GCCs, managed-service firms, agentic delivery organisations.
MEASURESthe capability to deliver with a hybrid human+agent workforce — embedded engineering, agentic build, autonomous operations, outcome accountability.
For the organisations that commission and govern AI-enabled delivery.
MEASURESthe capability to be an intelligent customer of autonomy — defining outcomes, governing agents you don't operate, absorbing transferred capability.
Assess both sides and you get what no single model can: a relationship maturity picture.
Pick a phase and a domain on the floor plane, then read the tower — the address is the intersection. Drag the lattice to look at it from anywhere.
An organisation's capability profile is the set of addresses it genuinely occupies — verified, not claimed.
Five phases turn clockwise around a core of six Always-On domains. The ring advances on its own — select a phase to hold it at the front.
{{ lpDesc }}
FLOW DOMAIN ANCHORED HERE {{ lpAnchor }}Five phases run clockwise; six Always-On domains run through all of them. Capability is read against both the phase and the ring.
A capability performed only in Operate does not satisfy an Always-On intent.
Roles across humans and agents; FDE capability; fluency development; agent registry, versioning, retirement.
Autonomy calibration; guardrails, sandboxing, kill-switches; auditability; escalation design; responsible-AI conformance.
Data and context pipelines; knowledge graphs; context quality and freshness; institutional memory agents can use.
Model portfolio and tiered routing; LLMOps/AgentOps; evaluation harnesses; token economics; vendor independence.
Live outcome measurement; human-vs-agent attribution; cost-to-serve analytics; the substrate for the P/B axes.
Client intimacy in embedded delivery; multi-party ecosystems; agent-to-agent interfaces across boundaries.
Outcome-based commercials; autonomy scope as a contractual object; AI service-level objectives; accountability for agent actions.
Agentic engineering as first-class SDLC; FDE co-build method; simulation and safety cases before autonomy is granted.
IT agents running services; self-healing, ticketless ops; human exception management; continuity across model changes.
Structured agent improvement under governance; model refresh without regression; drift management.
AI-class threat defence (prompt injection, data poisoning, agent compromise); privacy; autonomy-fallback-to-human recovery.
Portability by design; client capability uplift as a deliverable; exit without erasing institutional memory.
Every level has content — there is no empty starting rung.
L10 is a direction, not a destination. The cumulative rule applies: a level is held only while every level below it stays effective.
Each level defines ten concrete Gates — small, achievable steps. Clearing gates moves P; P = 10 → eligible for promotion assessment. Continuous forward motion; no multi-year plateaus.
Across functions, engagements and teams. B 2 = one pioneering pocket; B 10 = the organisation's normal state.
Racing ahead in one pocket, fragile elsewhere. → Consolidate before climbing.
The level is culture; momentum has stalled. → Open the next gates.
Progression and breadth move together; the level holds as it climbs.
AQ = (L × 10) + (0.6·P + 0.4·B) × (10/11)
Level dominates; progression outweighs breadth within a level; 100 only at a perfect L10. Worked: L4, P6, B5 → 40 + (3.6 + 2.0) × 0.909 ≈ AQ 45
The AQ is a headline. The AQ Coordinate is the diagnosis. They always travel together.
Practice-anchored questionnaires, behavioural five-point anchors, captured across three organisational layers — executive, management, practitioner — with perception-gap analysis: where leadership belief and practitioner reality diverge, per domain.
Live signals from the delivery estate: agent action logs, oversight-intervention rates, autonomy-boundary breaches, routing economics, learning-loop cadence, drift metrics. The gap between what is said and what is measured is itself a maturity signal.
Autonomy charters, guardrail policies, safety cases, audit trails, contract clauses — reviewed AI-first, human-verified for certification-grade standing.
The coordinate updates as evidence flows; certification is a point-in-time attestation of the live score by an accredited assessor. Regression is visible — capability is a state, not a trophy.
Every Domain × Level cell carries a capability statement (behavioural, metric-free) → 2–4 anchored questions tagged by layer → the P-gates it feeds → telemetry bindings → a breadth counter. Adaptive questioning keeps a respondent under 40 minutes.
Kestrel builds well but governs late: composition and platform capability run half a ladder ahead of governance and evolution. The pattern is a classic spearhead — strong pockets of L5 practice on a governance base still at L2. Two moves unlock the next band: stand up autonomy charters (gov gates G1–G3) and instrument one production learning loop (ev G1).
Run the AQDS or AQOS assessment: multi-layer attestation, connect available telemetry. Receive your first AQ Coordinate and AQ score.
Review the scorecard: domain profile, perception gaps, spearhead-vs-plateau shape. Agree where you actually are — the honesty step.
Select the next 3–5 gates from your current level. Each gate is small, owned, and dated. This is the roadmap — not a transformation programme, a gate list.
Clear gates; telemetry and re-attestation move P and B live. Celebrate standings (Achieved → Institutionalised) — breadth is culture-building.
Take a point-in-time certification for market signalling; join the anonymised benchmark to see your Coordinate against peers by size, sector and model.
Typical first-year arc: one level of ascent, or one standing upgrade — both are wins. The ladder rewards motion, not leaps.
The level is real: demonstrated in at least a meaningful pocket.
The level is the organisation's norm.
Endurance, made continuous.
AQ credentials are alive: every badge carries a live-verification link to the current coordinate. A score that regresses, shows.
AutonowmyQ™ is authored and maintained by Autonowmy Technology Pvt Ltd. The model is versioned; capability content evolves with the field through a formal change process, while the architecture — the ladder, the coordinate, the domains — stays stable so scores remain comparable over time. Assessment runs on the Autonowmy platform; certification is delivered under AQCS by Autonowmy and accredited partners.