Approach

Start with the decision. Build the capability behind it.

Every engagement begins with your situation: the outcome you need, the decisions involved and what an error would cost.

Argument Vectors set out the intent, context, constraints, evidence and authority behind an AI-assisted decision. AIF7™ examines the enterprise capabilities needed to develop, govern and run the system that supports it. Together, they connect the business problem to architecture, engineering and organisational action.

Discuss a RequirementRead the approach
The ARVEC approachThree levels
  1. 01

    Decision level

    Argument Vectors

    Define what the decision requires.

    IntentContextConstraintsEvidenceAuthority

  2. 02

    Enterprise level

    AIF7™

    Examine the capability behind it.

    7 capabilities × 7 lifecycle stages

  3. 03

    Intervention level

    Six areas of expertise

    Address the gaps that matter.

01Argument Vectors · Signature methodology

Make the foundations of the decision explicit.

An Argument Vector sets out what an AI-assisted decision depends on, before architecture or implementation begins. It is how Evidence, Meaning and Control are applied to a single decision.

  • Evidence

    Every finding traces back to its source documents and data.

  • Meaning

    Shared business definitions, so AI works with the right entities, relationships and rules.

  • Control

    Ownership, governance and operating discipline, so AI runs repeatably.

01IntentWhat are we trying to achieve?Purpose, intended outcome and the decision being supported.Control
02ContextWhat does the system need to understand?Business circumstances, relevant entities, relationships and terminology.Meaning
03ConstraintsWhich requirements and limits apply?Policies, rules, boundaries and conditions for use.Control
04EvidenceWhat supports or challenges the conclusion?Sources, provenance, contradictions, gaps and uncertainty.Evidence
05AuthorityWho or what may decide and act?Permissions, accountability, review, approval and escalation.Control

02Worked example

One decision, five dimensions.

A single document-verification decision in trade finance, followed through each dimension of the Argument Vector.

Documentary credit presentationIllustrative example. Not a client engagement.

What the system found

Commercial invoice

Quantity2,400 units

Bill of lading

Quantity2,200 units

Discrepancy flagged: 200 units

Both source passages linked. Finding drafted and routed to a documentary checker for decision.

01Intent
Decide whether the presented documents comply with the credit and can be honoured.
02Context
Applicant, beneficiary, goods and quantities; the credit's terms; the document set (commercial invoice, bill of lading, packing list).
03Constraints
The credit's terms, the applicable ICC rules (UCP 600), the bank's internal policy and the examination time limit.
04Evidence
The invoice states 2,400 units and the bill of lading states 2,200. Both passages are linked to their source, and no other document resolves the difference.
05Authority
The system flags the discrepancy and drafts the finding. A documentary checker makes the decision, and refusals above a set value go to senior trade operations.

The method shapes how the system is designed and evaluated. It does not expose a model's internal reasoning or guarantee a correct outcome.

03AIF7™ · Enterprise capability framework

Build the enterprise capability the decision requires.

A decision may be well specified while the organisation lacks the data foundations, engineering practices, governance or operational ownership needed to support it. AIF7™ provides a structured framework for examining those capabilities and prioritising improvements.

Seven enterprise capabilities

  1. 1AI Strategy & VisionDo we know where AI matters most — and why?
  2. 2AI Value RealisationCan we demonstrate the value AI is creating?
  3. 3AI Organisation & Operating ModelCan the organisation repeatedly turn AI ideas into operational capability?
  4. 4AI People & CultureCan people adopt, govern and improve AI as part of everyday work?
  5. 5AI Governance & TrustCan AI operate at scale without losing control, accountability or trust?
  6. 6AI Engineering & DeliveryCan we move AI reliably from prototype to production?
  7. 7AI Data FoundationCan each important AI input be traced to an approved, reliable data source?

Seven lifecycle stages

  1. S1Initiate · Assess & AlignWhat problem are we trying to solve?
  2. S2Design · Plan & PrioritiseWhat must be designed before we build?
  3. S3Build · Develop & PrepareAre we ready to move beyond the concept?
  4. S4Launch · Deploy & AdoptCan this work reliably in production?
  5. S5Optimise · Scale & ImproveCan we make it better and replicate what works?
  6. S6Transform · Industrialise & LeadHas AI become part of how the organisation operates?
  7. S7Sustain · Evolve & Future-ProofCan the organisation keep improving as AI evolves?

Explore in depth

The 7 × 7 capability map

The seven capabilities and seven lifecycle stages form 49 intersections. Each defines an expected activity that can be examined through evidence, ownership and risk review.

Select any square to view its expected activity, example evidence and decision gate. Rows identify capabilities by colour; columns become darker from Initiate to Sustain. Colour is not an assessment score.

Select a pillar

AI Strategy & Vision

Do we know where AI matters most — and why?

  1. 1.1AmbitionInitiate — Assess & AlignSigned AI ambition, named executive sponsor and defined scope
  2. 1.2Market & optionsDesign — Plan & PrioritiseEvaluated market/technology options and decision rationale
  3. 1.3PrioritiseBuild — Develop & PreparePrioritised portfolio using value, risk and feasibility criteria
  4. 1.4CommunicateLaunch — Deploy & AdoptStakeholder communication and accountable decision log
  5. 1.5Measure strategyOptimise — Scale & ImproveStrategy KPIs with observed measurements
  6. 1.6Refine strategyTransform — Industrialise & LeadStrategic review and reprioritisation decisions
  7. 1.7Embed strategySustain — Evolve & Future-ProofAI strategy embedded in enterprise planning and funding

Stages 1–7 describe the lifecycle

Columns show expected work and approval gates. They do not represent maturity ratings.

Maturity 1–7 is evidence-based

Every score requires evidence, rationale and confidence, assessed separately at each applicable intersection.

“Not evidenced” is not level 1

Unknown evidence and justified N/A are separate statuses. Critical legal, safety, security and model-risk controls override averages.

The seven independent maturity levels

  1. Maturity 1

    Not established

    Verified absence after sufficient examination; missing evidence is not a level 1 rating.

  2. Maturity 2

    Ad hoc

    Informal, person-dependent activity without demonstrated repeatability.

  3. Maturity 3

    Emerging

    Pilots or early practice with uneven coverage and limited repeatability.

  4. Maturity 4

    Defined

    Approved practice implemented across the stated scope with clear ownership.

  5. Maturity 5

    Managed

    Performance and controls are measured; corrective actions are recorded.

  6. Maturity 6

    Optimised

    Repeated measured improvements to reliability, effectiveness or cost.

  7. Maturity 7

    Industrialised

    Sustained, embedded and resilient operations with periodic revalidation.

Lifecycle stage and maturity are separate: a Stage 4 (Launch) activity can have Maturity 2 (Ad hoc), while a Stage 2 (Design) activity can have Maturity 6 (Optimised).

AIF7™ is ARVEC's proposed enterprise operating and assessment methodology, developed for field validation. It is not a certification, independent benchmark, assurance opinion or regulatory approval.

04From findings to intervention

From gap to the right expertise.

Findings from the decision and capability review point directly to the work required, across ARVEC's six areas of expertise.

  1. 01Argument Vectorsdefine the decision requirements
  2. 02AIF7™examines the supporting capability
  3. 03Six areas of expertiseaddress the gaps

Focused engagements use only the parts of the approach the problem requires. Not every assignment needs a full AIF7™ assessment.

05What you receive

What an engagement produces.

Sample outputs, illustrative

  1. 01Decision specificationOne Argument Vector per priority decision.
  2. 02Capability findingsAn AIF7™ view of the gaps and their root causes.
  3. 03Architecture recommendationThe target design for evidence, knowledge and controls.
  4. 04Prioritised actionsA sequenced plan, mapped to the relevant expertise areas.

Next step

Bring us the decision, workflow or capability you need to improve.

Whether it is a single decision or an enterprise programme, we start with what your AI must get right.

Discuss a Requirement