Canonical definitions

AI shaping glossary: core terms in plain language

Canonical website short definitions for the AI shaping website and paper landing pages. Use this page to avoid re-reading the full concept stack on every page.

The glossary explains the website architecture from two-sided work-burden shift through the direct-shaping approach or the mediated-shaping approach, shaped intelligence and productive AI work. It also keeps work pattern, reusable work basis and work state distinct, and maps practice-guided and subject-aware outcome language to domain-practice standards and subject-context reasoning.

Glossary term headings link to the most relevant explanatory page where one exists, rather than looping back to the same definition.

Core terms first

Read only these first if you are new

The full glossary is a reference list. Cold readers only need the six terms below before moving back to the plain pages.

Read only these first if you are new
Term	Plain role	Best next page
AI shaping	The operating-model discipline whose central operational aim is two-sided work-burden shift under human direction.	AI shaping
Direct-shaping approach	Under human direction, the AI-shaping discipline is applied directly to establish a bounded shaped capability without first materialising separately reusable AI-shaping intelligence.	AI shaping
Mediated-shaping approach	Under human direction, the AI-shaping discipline is applied through AI-shaping intelligence to develop, govern or renew domain-facing shaped intelligence.	AI-shaping intelligence
Shaped intelligence	The reusable domain-facing capability that carries more recurring work burden through a shaped work pattern and reusable work basis.	Shaped intelligence
Productive AI work	AI-enabled work that can be reviewed, continued and reused.	Productive AI work
Work-burden shift	More two-sided recurring burden moves from the responsible person to shaped intelligence while source basis and human authority remain explicit.	Work-burden shift

Core terms

AI shaping: The operating-model discipline for establishing shaped intelligence that carries a reusable and iteratively improvable domain work pattern and reusable work basis. Two-sided work-burden shift is the operational effect while source basis and the human-authority boundary remain explicit.
Direct-shaping approach: Under human direction, the AI-shaping discipline is applied directly to establish a bounded shaped capability without first materialising separately reusable AI-shaping intelligence. Direct approach is controlled shorthand after the formal term is established; direct shaping names the corresponding activity.
Mediated-shaping approach: Under human direction, the AI-shaping discipline is applied through AI-shaping intelligence to develop, govern or renew domain-facing shaped intelligence. Mediated approach is controlled shorthand after the formal term is established; mediated shaping names the corresponding activity.
AI-shaping intelligence: The reusable shaping capability used in the mediated-shaping approach to develop, govern or renew shaped intelligence. It combines self-shaping capability, which develops and renews the shaping pattern, with domain-shaping capability, which uses that pattern to develop shaped intelligence for a target work domain under human direction.
Self-shaping capability: The capability surface of AI-shaping intelligence that develops, audits, improves and renews its own shaping pattern under human direction. It is one side of AI-shaping intelligence, not the whole capability.
Domain-shaping capability: The capability surface of AI-shaping intelligence that uses the shaping pattern to develop domain-facing shaped intelligence for a target work domain under human direction. It is paired with self-shaping capability.
Domain extension: Widening shaped intelligence within a related work family, where an earlier shaped capability becomes a subset of a broader shaped capability. In the current public evidence, vendor-evaluating shaped intelligence extended into project-managing shaped intelligence because vendor evaluation sits inside broader project-management work.
Domain extensibility: The capability-evaluation property that shaped intelligence can be widened within a related work family, where an earlier shaped capability may become a subset of a broader shaped capability. Domain extensibility is evidenced only when that extension is actually shown or reviewed; it should not be treated as unrestricted transfer.
Domain transfer: Materialisation or evaluation of shaped intelligence in a different work domain through target-domain rematerialisation. Domain transfer is not proven by domain extension; it requires its own evaluation and does not authorise implementation, commercial reliance or protected-method transfer.
Domain transferability: The evaluation property that a shaping pattern may be assessed for target-domain rematerialisation in a different work domain. Domain transferability does not prove domain transfer, implementation readiness, commercial reliance or protected-method transfer.
Target-domain rematerialisation: Developing shaped intelligence against a different target domain's work pattern, source expectations, domain-practice conditions, subject context, review boundaries, work state and resumption needs rather than assuming a shaped capability can simply be copied across domains.
Transfer-evaluation map: A strictly illustrative public-safe route for showing how a later protected-stage evaluation could consider domain transfer without proving the transferred capability, enabling implementation or disclosing protected method.
Shaped intelligence: The reusable domain-facing capability that carries a reusable and iteratively improvable domain work pattern and reusable work basis into recurring work so more two-sided burden can shift from the responsible person while source basis and human authority remain explicit.
Product-managing shaped intelligence: The product-level capability operationally established within the bounded AI Shaping knowledge and publication product. It carries recurring product-state, owner-selection, sequencing, reconciliation and resumption burden through governed automation under human review. It is supported across repeated accepted cycles; the three-cycle maturity gate remains satisfied while broader evidence limits remain.
Project-managing shaped intelligence: The formally published mediated-shaping capability-evidence instance in project and business-as-usual (BAU) IT support work, carrying recurring project-support work burden while human authority remains explicit.
Productive AI work: AI-enabled work that is grounded, practice-guided, subject-aware, reviewable, repeatable and resumable under human direction; the work outcome of AI shaping that can be reviewed, approved, continued and reused.
Work burden: The recurring human load required to turn useful but fragile AI output into productive AI work, including context rebuilding, source-basis repair, domain-practice application, subject-context reasoning, work-state preservation and resumption work. Source-basis rebuilding can contribute to that load, but source basis remains the grounding condition around the two-sided model rather than a third burden side.
Work-burden shift: The visible before/after operational effect in which more recurring work burden moves from the responsible person to shaped intelligence. In AI shaping, two-sided work-burden shift is the primary capability-evidence signal while source basis is carried forward and the human-authority boundary is preserved.
Two-sided work-burden shift: The AI-shaping operational effect in which shaped intelligence carries more domain-practice and subject-context burden, and more work-state and resumption burden, while judgement, approval and real-world action remain human-owned.
Stock-evaluating shaped intelligence: A planned capability direction for repeated evaluation of one listed stock. It is deliberately not portfolio-aware and remains planning context, not current capability evidence, financial advice, autonomous trading, guaranteed performance or authority to act.
Portfolio-managing shaped intelligence: A possible later superset extension adding holdings and transaction state, sizing, concentration, combined exposures, allocation, cash constraints, aggregate risk, rebalancing and portfolio history. It remains planning context, not current capability evidence, portfolio-management assurance or financial advice.
Code-developing shaped intelligence: A planned capability direction for recurring code-development work across research, specification, implementation, testing, defect handling, work state and resumption. It is planning context, not current capability evidence or implementation guidance.
Paper-authoring shaped intelligence: A bounded form of shaped intelligence established through the direct-shaping approach for developing, maintaining, assessing and releasing a governed paper set across successive revisions. The current publication system distinguishes private and published paper-authoring capabilities by dominant work object and disclosure status.
Private paper-authoring shaped intelligence: A bounded private form of shaped intelligence established through the direct-shaping approach for developing and maintaining the complete private AI Shaping paper set across successive authorised revisions. Its existence and role may be described publicly; its papers, controls, rubrics and operating mechanics remain private.
Published paper-authoring shaped intelligence: A bounded form of shaped intelligence established through the direct-shaping approach for developing and maintaining the formal public Category Definition and Capability Evidence paper set across authorised revisions under human publication authority. Its published outputs are inspectable; complete revision controls remain private.
Public web-developing shaped intelligence: A bounded website-specific form of shaped intelligence established through the direct-shaping approach for recurring development and revision of the public website across content, information architecture, visual explanation, interface implementation, publishing and validation under human direction. It is not claimed as a generic or cross-site capability.
Private knowledge-publishing shaped intelligence: A bounded private form of shaped intelligence established through the direct-shaping approach that synthesises, structures and presents authorised source material as a digestible, reusable knowledge surface under human direction. Its name, private status and purpose may be disclosed without publishing its content or operating machinery.
Publication co-evolution through reciprocal refinement: A principal operating mechanism used by Product-managing shaped intelligence across Private knowledge-publishing, Private paper-authoring, Published paper-authoring and Public web-developing shaped intelligence. It is not the umbrella capability, a fifth publication capability, automatic propagation or simultaneous autonomous operation.

Work-pattern and work-state terms

Practice-guided: Short-form wording for AI-enabled work shaped by relevant domain-practice guidance, standards, expectations and review boundaries. It connects the shaped work pattern to how work should be done in that domain.
Subject-aware: Short-form wording for AI-enabled work shaped around the specific matter being worked on. It connects subject-specific context to subject-context reasoning, so the work is not generic AI output detached from the actual project, issue, facts or constraints.
Domain-practice guidance: The domain-practice expectations supplied into the shaped work pattern, including relevant work-practice standards, review expectations and operating patterns.
Domain-practice standards: The standards, methods, expectations and operating patterns that shaped intelligence carries so AI-enabled work follows the relevant domain practice rather than a generic response pattern.
Domain-practice application: The work of applying relevant work-practice standards, review expectations and operating patterns inside shaped intelligence, so AI-enabled work follows the domain’s work pattern rather than a generic response pattern.
Subject-specific context: The matter-specific information the AI-enabled work depends on, such as the project, issue, stakeholder, support context, constraints or supplied facts.
Subject-context reasoning: The work of reasoning with subject-specific project, issue, stakeholder or support-context information so AI-enabled work is reviewable and fitted to the actual situation.
Domain-practice and subject-context burden: The side of work-burden shift concerned with carrying domain-practice standards, subject-specific context, subject-context reasoning and relevant interpretation into the shaped work pattern.
Work pattern: The reusable and iteratively improvable organising structure for how recurring work should run: its role relationship, source relationship, domain-practice expectations, subject-context reasoning, review boundaries and output pattern. It is not one side of the two-sided burden model.
Reusable work basis: The basis carried into recurring work: source basis, relevant domain-practice standards, subject context, current work state and resumption logic. AI shaping arranges this basis; shaped intelligence carries it into recurring work.
Source basis: The relevant source material, facts, assumptions, provenance and supplied context carried forward as the grounding basis for AI-enabled work. Source basis is a condition around the two-sided burden shift, not a third burden side.
Work state: Where the AI-enabled work is up to: what happened, what changed, what remains unresolved, what needs review and what happens next, so the work can resume without rebuilding the whole work basis.
Work-state preservation: The component of work-state and resumption burden concerned with preserving current state, unresolved matters, assumptions, decisions and review status so the work can continue coherently.
Reconstruction: The repeated work of rebuilding the AI work basis before AI-enabled work can continue: context, source material, prior decisions, constraints, risks, review status and next action.
Resumption: Continuing AI-enabled work from a preserved reviewable state after a pause, handover, context loss or new information, instead of asking AI to restart from a rebuilt prompt.
Work-resumption logic: The component of work-state and resumption burden concerned with preserving next action, version context and continuation logic so work can resume without rebuilding the whole work basis.
Work-state and resumption burden: The side of work-burden shift concerned with work-state preservation and work-resumption logic: carrying current state, unresolved matters, review status, next action and continuation logic forward. Source basis is a separate grounding condition around the shift.
Context rebuilding: A plain-language version of reconstruction. In ordinary AI use, this often means writing a long or complex prompt to re-explain the work before the AI can help again.
Grounded: AI-enabled work is grounded when it remains tied to source basis, context, review state and relevant constraints rather than becoming detached AI output.
Reviewable: AI-enabled work is reviewable when a person can check, correct and approve it from a visible work basis and review state.
Repeatable: AI-enabled work is repeatable when it can recur through a reusable work pattern without the person rebuilding the whole work basis each time.
Resumable: AI-enabled work is resumable when it can continue after pause, handover or new information while preserving work state and next-action logic.
Specification-led AI work: AI work in which the person supplies a comparatively complete problem statement, source set, requirements, constraints and desired result before substantial AI production begins.
Iterative AI work: AI work in which the problem definition, requirements, work product and next action develop through repeated human review and AI revision cycles.

Operating-model and authority terms

AI-enabled work systems: A practical umbrella for work systems that use AI to support complex decision, delivery, review and coordination workflows while keeping human authority explicit. On this site, AI shaping is the specific operating-model discipline used to explain how useful AI output can become productive AI work.
Operating-model design: The central design activity and work-system lens for AI shaping: intentionally arranging burden allocation, source relationship, capability and human-authority boundaries, review controls, domain-practice conditions, subject-context reasoning, work state and resumption so recurring AI-enabled work can operate coherently.
Decision and delivery workflows: Organisational plain-language wording for workflows where decisions, delivery status, evidence, unresolved matters, review state and next action need to remain coherent across pauses, changes and handover.
Human-authority boundary: The governing boundary under which purpose, supplied context, judgement, review, approval, disclosure, escalation, communication, decisions and real-world action remain human-owned while shaped intelligence carries more recurring work burden.
Human product ownership, stewardship and decision authority: The human-owned product relationship in which product ownership retains purpose, value, audiences, priorities, scope, architecture decisions and acceptance; stewardship retains continuity and integrity; and decision authority retains judgement, rejection, disclosure, publication, deployment and consequential action.
Human-owned authority: The observable evidence-test expression of the human-authority boundary: supplied context, judgement, review, approval, disclosure, escalation, communication, decisions and real-world action remain with the responsible person even when shaped intelligence carries more recurring work burden.
Domain-bounded shaped intelligence: Shaped intelligence developed for a defined work domain or evidence environment rather than an unrestricted all-purpose autonomous assistant.
Governed automation: AI-carried recurring coordination performed through accepted controls and subject to human review and acceptance. It is not autonomous product management; purpose, judgement, disclosure, publication, deployment and consequential action remain human-owned.

Public and protected-stage terms

Public-stage: The public reading or evaluation stage for AI-shaping material, limited to category understanding or bounded capability evaluation without implementation reliance.
Public-stage material: Material that supports public category understanding or bounded public-stage capability evaluation of AI shaping, without providing method transfer or deployment assurance.
Protected-stage: The non-public stage for selected practice, capability, evidence or implementation questions that cannot be evaluated adequately from public material alone.
Protected-stage discussion: The minimum scoped discussion pathway used where a defined question may justify non-public review, subject to recipient, purpose, confidentiality, ownership, scope and permitted use.

Use the glossary for

Quick concept lookup.
Consistent wording across the paper landing pages.
Reader-friendly entry before the full papers.

Reading sequence

Page 29 of 31