Glossary

Create a credible initiative narrative that explains the problem, timing, outcome, uncertainty, trade-offs and non-goals without hype.

alignment

One of the most powerful alignment tools in AI Transformation is a clear and compelling “why” narrative. People are rarely resistant to AI itself; what they resist is uncertainty. When questions like “Why this AI?”, “Why now?”, or “Why are we investing here?” remain unanswered, the gaps are quickly filled with assumptions, anxiety, and unrealistic expectations.

alignment

Convert Integration findings and current-to-target capability gaps into a sequenced, owned 90-day roadmap of usable operating improvements and evidence-producing decisions.

integration

Creates bounded experiments, decision gates and measurable progress before large-scale process expansion.

ethics

Accessibility means that authorised consumers, whether people, applications or agents, can discover and use information with reasonable effort. Discovery works differently for each: a person can ask a colleague, whereas an agent can only use what is described in a form it can read. It is different from availability, which describes whether information can be obtained at all, and from access control, which determines who may use it. Interoperability means that information can move across systems without losing meaning, identity or essential constraints.

data

Determine whether authorised human and machine consumers can obtain and interpret required information with acceptable effort.

data

Accessibility means that authorised consumers, whether people, applications or agents, can discover and use information with reasonable effort. Discovery works differently for each: a person can ask a colleague, whereas an agent can only use what is described in a form it can read. It is different from availability, which describes whether information can be obtained at all, and from access control, which determines who may use it. Interoperability means that information can move across systems without losing meaning, identity or essential constraints.

data

Accountability connects business authority to measurable obligations. Owners decide meaning, acceptable use, quality tolerance and investment priority. Stewards coordinate definitions and remediation. Custodians operate storage, access and protection. Product or service owners ensure that information delivered to consumers remains useful and supported. Producers remain answerable for what they create at source, and consumers, including AI services and agents, are accountable for using assets within their stated limits. Where an agent operates outside those limits, accountability rests with the team that deployed it, not with the source that supplied the data.

data

Defines decision rights, lifecycle obligations, contextual access and exception handling.

data

Clarify who decides, who advises, who executes and who provides independent challenge for critical data-use decisions.

data

Quality rules should be derived from business decisions and failure modes. Accuracy asks whether information reflects reality; completeness asks whether required elements are present; consistency tests whether representations agree across contexts; and validity checks conformance to defined rules. The relevant thresholds vary by use case, risk and reversibility. These four dimensions assume structured records. Content in document and conversational form requires a different set of tests, which are addressed in D.2.2.

data

Actual model usage cost is calculated from provider- and model-specific usage records and applicable billing rules. In token-priced services, the calculation commonly separates billable input, cached input where applicable, and output because these may be priced differently. The calculation also includes usage created by system prompts, tool schemas, orchestration, retries and discarded responses. Forecasts may use a model-specific token counter, but actual reporting reconciles API usage, billing exports and workflow telemetry.

impact

Triggers translate evidence into action. They define conditions for continuing, investigating, mitigating, pivoting, pausing, scaling or retiring an initiative. Trigger design combines target outcomes, guardrail breaches, cost variance, adoption, evidence confidence and changes in strategic relevance.

impact

Adoption friction is often the employee-side expression of poorly redesigned roles, increased cognitive burden, weak trust foundations, and unclear performance logic. A major risk is to interpret all resistance as an employee attitude problem and never question the system design. Another is to evaluate early adoption only through usage rates while ignoring depth, trust, quality, or sustainability. In such cases, the organisation may believe AI use is increasing while a superficial, fragile, and obligation-driven adoption model is actually taking shape. A mature operating pattern is evident when the organisation learns from friction instead of suppressing it. Progress is evident when resistance signals are identified earlier, diagnosed more accurately, and used to improve the adoption approach itself.

values

Auto-register owner, purpose, current risk classification, AI authority mode, human-control mode, delegated-authority envelope, version, model, endpoints, tools, data scope, delegation rights, expiry, cost limits and lifecycle state.

agentic operating model

Within the approved Autonomy Appetite and decision authority, define an agent’s permitted actions as an explicit, time- and scope-bounded authority envelope that remains attributable to an accountable human or service owner.

ethics

Expose enterprise data and actions to agents through bounded, interpretable and auditable business-capability interfaces.

data

Agent evaluation must test policy compliance, action quality, recovery and refusal alongside task success. Observability, safe failure and termination show whether control still holds when the system behaves unexpectedly. Evaluate task success together with policy compliance, tool-selection quality, action correctness, recovery, refusal, escalation and harm. Include messy inputs, missing context, malicious instructions, dependency failure and long-horizon tasks.

ethics

Agent identity makes delegated action attributable. Least privilege then limits that identity to the tools, data and operations required for the approved purpose and period.

ethics

Agent identity makes delegated action attributable. Least privilege then limits that identity to the tools, data and operations required for the approved purpose and period. Give each material agent or agent class an attributable non-human identity and avoid invisible sharing of human credentials. Bind actions to both the agent identity and the authorizing user, workflow or service where appropriate.

ethics

Give agents attributable identity and least-privilege, action-aware access to tools, data and destinations.

ethics

Effective delegation, monitoring, quality control and intervention within formal boundaries defined by Alignment and the Agentic Operating Model, demonstrated in realistic and live work.

values

Agent-to-agent delegation introduces another form of operational handoff. One agent may decompose a request and delegate parts of it, another may retrieve or interpret information, and another may execute a business action. The same handoff principles established in I.1 still apply, but the chain adds further requirements for identity, authority, task continuity and evidence across each transition. Architecture & Rhythm defines how multi-agent coordination should be designed; Integration examines whether the resulting chain remains bounded, attributable and reconstructable in operation.

integration

An interoperability standard for agent-to-agent communication. AIDIVE does not prescribe a protocol; identity, task continuity, authorization, context, lifecycle and evidence commitments remain explicit regardless of transport.

agentic operating model

The Agentic AI Operating Model is the execution architecture through which approved enterprise intent and all six AIDIVE dimensions are enacted across five operating layers. Alignment defines strategic intent, decision rights, portfolio policy, Autonomy Appetite and feedback contracts; this document defines how those choices are operated through layers, roles, cadences, value streams, workflows and controls.

agentic operating model · maturity journey

Purpose: To identify dependencies and bottlenecks between multiple Agentic Networks to resolve these using workflow design principles rather than ticket management. Cadence / trigger: Biweekly Core participants: •Cross-Agentic Networks Coordination Lead•Value Stream Owner / Cluster Owner•Value Stream Architecture / Flow Design Lead Additional participants: •Experience / Journey Lead•Data / Knowledge Representative•Selected Agentic Networks representatives•Platform representative Evidence inputs: • Dependency maps • Integration and context-transfer issues • Release synchronization needs • Shared capability / platform dependencies • Flow anomalies • Interface or contract failures Expected output / decision: • Dependency resolution decisions • Interface, context-transfer or contract changes with named owners • Synchronization decisions • Handoff improvement actions • Unresolved dependencies routed to the appropriate authority

agentic operating model

The Agentic Networks Layer synchronises multiple teams, agents and shared capabilities so they operate as one delivery engine around a common value stream. It is a coordination shell rather than a managerial hierarchy, and it keeps integration, assurance, experience, policy and runtime evidence coherent across teams.

agentic operating model

Coordinates multiple team types, shared AI capabilities and human-AI workflows so they operate as one delivery engine within the value stream. Maintains synchronization and escalation paths within approved boundaries and does not acquire business decision authority merely by coordinating the network.

agentic operating model

A critical threshold in agentic value-stream design is the ability to connect individual AI capabilities to the end-to-end flow of work. When multiple AI capabilities participate in the same end-to-end work, their system-level value and reliability depend on coherent coordination toward a shared operating objective, with the required context, authority boundaries and outcome evidence preserved across the flow.

agentic operating model

Value streams have always been central to understanding how organizations create value. They reveal how customer needs move through the enterprise, how work travels across teams and systems, where decisions are made, and where delays accumulate. In traditional operating models, value streams were primarily understood as human-centered flows: people interpreted signals, made decisions, transferred work, resolved exceptions, and corrected errors through meetings, escalations, dashboards, and experience-based judgment.

agentic operating model