FluencyTracr

Agents and contributors: Read AGENTS.md before making any changes. It defines the V1 invariants, the canonical event and suppression sets, and the ordered prompt roadmap. Influences and credits: see ATTRIBUTION.md.

FluencyTracr is the behavioral evidence layer for Glean value realization. It exists because the current time-saved pipeline can show acceleration while leaving a critical question unanswered: roughly 64% of chat runs have no quality signal today.

VBD joint methodology, synthetic validation hold: VBD Coverage Trajectory describes a governed organization-first replacement for the illustrative weighted VBD posture: Eligible -> Active -> Embedded, with Embedded Adoption Coverage on X and Net Coverage Velocity on Y. Adoption Reach and Persistence are supporting ratios, and retained, newly embedded, and lapsed counts expose churn. VBD is the observed human-behavior pathway in a docs-only joint Bayesian design that keeps stated capability, actual work behavior, and the customer-owned outcome separately measured while propagating uncertainty. Status: SYNTHETIC_V4_REPLICATED_VALIDATION_PROTOCOL_FROZEN; decision: HOLD_FOR_MODEL_REPAIR. The isolated synthetic model has run bounded, permanently nonqualifying smoke matrices; full validation, real data, runtime integration, UI migration, and customer output remain unauthorized.

FluencyTracr sits on top of that pipeline as bounded evidence services and documented value-realization layers:

• AI Work Evidence: org-agnostic core layer that separates reusable aggregate evidence primitives from source-specific adapters such as Glean dogfood and value-evidence mappings. It defines surfaces, workflows, approved cohorts, interventions, trust evidence, source coverage, outcome evidence, and value hypotheses without adding events, suppression reasons, individual scoring, rankings, or realized ROI calculation. When aggregate evidence is trusted, reliable, and customer-approved, it may feed customer-owned value-investigation metric routing and governed value-scenario review without customer-facing economic output.
• Quality Multiplier: discounts, preserves, or amplifies time-saved estimates when aggregate workflow behavior shows enough evidence quality.
• Causal Delta: compares pre/post workflow patterns around a known change moment so value teams can ask what changed after rollout without claiming statistical causality.
• Reliability Factor: qualifies whether surfaced workflow evidence looks operationally dependable based on verification, recovery, abandonment, and friction-loop behavior.
• Outcome Evidence: stores and replays customer-attested aggregate KPI outcomes next to the unchanged workflow verdict so external consumers can do their own correlation.
• Velocity Index: adds V2 aggregate usage-distribution context across frequency, engagement, and breadth, surfaced only as cohort percentiles after the same fail-closed gates clear. Dogfood velocity coverage now follows the full surface taxonomy in docs/concepts/SURFACES.md, spanning workflow and standalone AI surfaces without adding person-level output.
• Depth: frames cross-surface work integration through surface repertoire and repeated meaningful use, then qualifies that evidence with verification, delegation, reuse, recovery, and judgment behavior. Depth is aggregate evidence that adoption is becoming durable enough to support defensible value claims.
• AGENT sub-surfaces: V2.3 splits AGENT into agent:autonomous, agent:workflow_named, and agent:ephemeral so delegation, reusable Skill, and exploratory agent behavior can be evaluated independently.
• Skill Read Evidence: V4 research path for Skills usage in agent span logs. It treats raw span skill-reader attributes, legacy skill-name inputs where present, plus dbt skill_reader_skill_name, as availability signals only until unspecified-share, parent attribution, canonical identity, versioning, invocation mode, UGC coverage, and personal/shared/org Skill separation are validated.
• AI Fluency measurement-model calibration contract: defines the aggregate 24-item aggregate ordinal evidence boundary and accepted internal synthetic proof. Current state is SYNTHETIC_CALIBRATION_ACCEPTED_REAL_DATA_PENDING: 800 primary slots and 800 fresh recomputations passed exact-byte review, while dimension summaries alone remain insufficient and real-data admission, output, persistence, routes, UI, connectors, ROI, causality, productivity, finance, HR, ranking, and economic use remain blocked.
• VBD Coverage Trajectory replacement concept: defines the proposed organization-first work-pattern-family adoption model: Adoption Reach is Active / Eligible, Persistence is Embedded / Active, Coverage is Embedded / Eligible, and Net Coverage Velocity is adjacent Coverage change normalized to 30 days. Stable observable source coverage is required for comparison; retained, newly embedded, and lapsed counts expose churn; and an exact alignment receipt is required before capability, adoption, and outcome movement may be interpreted together. A docs-only joint methodology now treats the aggregate family-state trajectory as the observed human-behavior pathway, carries its uncertainty into a predeclared outcome model, and keeps stated capability as a separate measured term. It prohibits weighted Integration and overall VBD scores, holds function drilldowns pending an approved aggregate join, and defaults to noncausal interpretation. Current state is SYNTHETIC_V4_REPLICATED_VALIDATION_PROTOCOL_FROZEN with decision HOLD_FOR_MODEL_REPAIR; see VBD Coverage Trajectory.
• Legacy VBD trajectory-model calibration contract: defines the earlier non-overlapping aggregate frequency, engagement, and Breadth trajectories, with Depth kept as source-bound context outside the likelihood. Current state remains SYNTHETIC_IMPLEMENTATION_HELD_FOR_NUMERICAL_PRECISION_REPAIR. It has a different estimand from the proposed Coverage Trajectory and is not its runtime implementation. PR #434's collapsed-target algebra remains a held, non-evidentiary diagnostic oracle under REPAIR_DIAGNOSTIC_LINEAGE_ONLY; it does not authorize execution. Its prior bounded roadmap remains as legacy planning context in the VBD Live Product Roadmap. Its three pending queue items are dormant behind their recorded prerequisites and require separate human activation; they are not completed work or an alternative implementation of the current Coverage Trajectory.
• Hypothesis and metric longitudinal admission: allows companies to define different aggregate metric catalogs without an arbitrary count cap while binding each longitudinal analysis unit to one approved hypothesis and one immutable primary metric version. Current eligibility remains limited to the proved synthetic continuous-normal aggregate specification; unsupported metric families HOLD and no AI-attribution confidence or customer output is authorized.
• V3 production ingest: moves beyond manual CSV dogfood by running a customer-side transformer in the customer's cloud, sending only aggregate cohort distributions to FluencyTracr, and storing immutable verdicts against governed calibration baselines. This is bounded aggregate ingest, not a production Sigma/BigQuery AI Value pipeline: FluencyTracr does not run live Sigma/BigQuery queries for the AI Value spine, ingest raw rows, or authorize customer-facing value output from this path.
• AI Value controlled aggregate pipeline dry run: proves that a BigQuery/Sigma-shaped scrubbed aggregate export manifest can enter the existing source-review path and produce a compact internal Measurement Cell candidate proof. It does not run BigQuery/Sigma, persist pipeline runs, create Measurement Cell snapshots, or emit customer-facing value/finance/confidence output.
• AI Value Measurement Cell snapshot projection: proves that a promoted backend-internal Measurement Cell snapshot row can be reduced to a compact internal operator-review product shape. It does not create read routes, frontend UI, exports, rendered readouts, live connectors, model output, finance output, or customer-facing output.
• AI Value customer data model promotion gate: proves that the compact Measurement Cell snapshot projection is safe enough only to enter a separate exact-scope customer data model persistence promotion decision. It holds by default, validates ready gates against the source projection, and blocks persistence writes, schemas, routes, UI, exports, rendered readouts, live connectors, model output, confidence/probability/score output, finance output, and customer-facing output.
• AI Value customer data model persistence promotion decision: proves that the customer data model gate can move only through the exact-scope implementation-decision lane. It holds by default and keeps any broader persistence writes, Prisma schemas, migrations, repository write paths, routes, UI, exports, rendered readouts, live connectors, model output, finance output, and customer-facing output blocked.
• AI Value customer data model persistence implementation decision: names and bounds the only promoted physical slice for customer-ready product data: one append-only compact ai_value_customer_data_model_snapshots table plus internal repository write/read paths derived from valid Measurement Cell Snapshot Projection inputs. It does not by itself expose stored rows, exports, rendered readouts, live connectors, confidence/probability/score output, finance output, or customer-facing economic output.
• AI Value customer data model route projection: opens the exact read-only customer evidence status projection over ai_value_customer_data_model_snapshots: one source-bound backend route and one compact workspace panel. It exposes approved customer-safe labels, windows, evidence status, caveats, blocked outputs, and next evidence action only; it only projects clear validated rows; it does not derive customer labels from compact IDs, apply stale local filters, or expose stored rows, org/client IDs, snapshot IDs, source refs, hashes, raw rows, exports, rendered readouts, live BigQuery/Sigma/Glean execution, confidence/probability/score output, ROI, finance output, causality, productivity, or customer-facing financial output.
• AI Value Measurement Cell Series persistence decision: repeated Day 0 / 30 / 60 / 90 / 180 / 365 validation now passes through the contract-only Measurement Cell Series layer, but durable measurement_cell_series_snapshots remain held because no product read-path need has been proven. Evidence Continuity stays inside the Series contract output for now; do not extend evidence_snapshots until a later exact-scope decision authorizes it.
• AI Value Measurement Cell Series persistence promotion gate: adds the executable hold-by-default gate after repeated milestone validation. It can become ready only for a later exact-scope Series snapshot implementation decision after a separate durable read-path decision proves compact customer data model projections cannot satisfy continuity needs. Caller-supplied proof strings alone still hold. It does not create measurement_cell_series_snapshots, extend evidence_snapshots, add routes, UI, exports, rendered readouts, live BigQuery/Sigma/Glean wiring, model output, confidence/probability/score output, finance output, ROI, causality, productivity, or customer-facing output.
• AI Value customer evidence history read-path proof: proves that current Day 0 / 30 / 60 / 90 / 180 / 365 customer evidence history can be served from compact ai_value_customer_data_model_snapshots plus Measurement Cell Series contract output. It binds only compact milestone counts and hashes, ignores stale superseded rows, holds on missing or held latest rows, and does not create Series persistence, extend evidence_snapshots, add routes, UI, exports, rendered readouts, live connector execution, model output, confidence/probability/score output, finance output, ROI, causality, productivity, or customer-facing economic output.
• AI Value durable Series read-path decision: consumes the customer evidence history read-path proof and records the current decision: compact customer history satisfies the read path, so measurement_cell_series_snapshots remain blocked. The allowed next step is to continue history reads from ai_value_customer_data_model_snapshots; the decision does not emit a ready Series implementation state, persistence write, schema, migration, repository path, route, UI, export, rendered readout, live wiring, model output, finance output, ROI, causality, productivity, or customer-facing economic output.
• AI Value confidence-engine Series read-path decision: names the held internal contribution-alignment Bayesian execution runtime as a distinct durable-series read-path consumer and, per the approved OpenSpec change add-ai-value-series-confidence-read-path, authorizes Measurement Cell Series persistence solely as append-only internal confidence-engine observation input at Day 0/30/60/90/180/365 milestones. It narrows exactly one feed — research_model_feed becomes the scoped token internal_confidence_engine_only — while the durable customer-history decision stays unchanged and the promotion gate gains an internal_confidence_observation lane that can reach READY only from a fully source-bound authorized decision. It does not create measurement_cell_series_snapshots, schemas, migrations, repositories, routes, UI, exports, rendered readouts, live BigQuery/Sigma/Glean execution, customer connectors, customer-facing model/confidence/probability/score output, finance output, ROI, causality, productivity, or customer-facing economic output; physical persistence still requires the separate exact-scope implementation decision behind the promotion gate.
• AI Value data model spine readiness lock: records that the compact customer data model spine is ready for hardening toward real source wiring only as a Boolean readiness contract. The implemented equation is measurement_cell_snapshots_promoted AND ai_value_customer_data_model_snapshots_promoted AND customer_data_model_route_projection_ready AND customer_evidence_history_read_path_proven AND durable_series_read_path_holds_series_persistence AND all_blocked_outputs_false. It explicitly reports no statistical model equation, confidence math, or numeric weights; keeps Series persistence held; and blocks model output, confidence/probability/score output, finance output, live BigQuery/Sigma/Glean execution, ROI, causality, productivity, and customer-facing economic output.
• AI Value compact source wiring hardening: turns the data model spine lock's allowed next step into non-live compact source-descriptor posture for prepared bigquery_export and sigma_export only. glean_query remains held. It does not authorize live BigQuery/Sigma/Glean execution, credentials, SQL/query storage, warehouse/dashboard handles, raw rows, routes, UI, exports, rendered readouts, model/confidence/probability/score output, finance/ROI/causality/ productivity output, customer-facing output, or Measurement Cell Series persistence.
• AI Value connector promotion readiness sequence: records the next four actions as a governed requirements path: non-live connector promotion decision requirements, held Glean source adapter boundary planning, source descriptor promotion checklist, and exact-scope BigQuery/Sigma live connector gate design. It names the future target of a full data model with weights and Bayesian readiness, but keeps numeric weights, Bayesian model execution, model output, confidence/probability/score output, live connectors, finance output, and customer-facing output blocked.
• AI Value controlled aggregate connector adapter: proves that the reviewed aggregate dry-run proof can become a compact internal BigQuery/Sigma connector review packet without credentials, live execution, persistence, snapshots, routes, UI, or customer-facing output.
• AI Value aggregate connector boundary plan: proves that a source-owner-attested BigQuery/Sigma aggregate export plan can pass a validator-only boundary review before any live connector exists. It does not run BigQuery/Sigma, store connector or pipeline runs, persist manifests, create Source Packages or Measurement Cells, or emit model, finance, or customer-facing output.
• AI Value controlled aggregate manifest validation: proves that the saved BigQuery/Sigma-shaped connector review packet can be represented as Source Inventory, Aggregate Extraction, and Pipeline Run Review manifests for operator promotion review. It remains non-persistent and does not authorize live connector execution, Source Package clearance, Measurement Cell or Series persistence, confidence math, finance output, or customer-facing output.
• AI Value research promotion readiness packet: proves that repeated governed Day 0 / 30 / 60 / 90 / 180 / 365 aggregate evidence can become a compact internal research-design gate. It emits only refs, hashes, caveats, blocked uses, and false boundary feeds; it does not authorize model math, numeric weights, persistence, exports, finance output, or customer-facing output.
• AI Value contribution alignment internal prototype runner: proves that the research-design packet can become a local compact internal review envelope without becoming a model. It emits only refs, hashes, source-bound posture, caveats, blocked uses, and false boundary feeds; it does not authorize confidence math, numeric weights, persistence, exports, finance output, or customer-facing output.
• AI Value contribution alignment runner review packet: proves that the compact internal prototype-runner envelope can become a source-bound internal model-prototype design review packet without becoming a model. It keeps AI Fluency construct context, psychological context, observed VBD context, and selected customer metric movement separate; it does not authorize model implementation, numeric weights, persistence, exports, finance output, or customer-facing output.
• AI Value contribution alignment model prototype design review: records the internal candidate model frame and alignment-review components without implementing a model. It remains method-design-only and blocks model math, numeric weights, score/probability output, persistence, exports, finance output, and customer-facing output.
• AI Value contribution alignment internal model prototype: replays the governed design-review components as compact internal component traces without emitting a model result. It remains non-persistent and blocks model math, numeric weights, confidence output, probability/score output, persistence, exports, finance output, and customer-facing output.
• AI Value contribution alignment internal model prototype review packet: turns the compact internal prototype into a source-bound review packet for a separate internal research-design gate. It remains non-persistent, compact-ref-only, and blocks model feeds, model implementation, numeric weights, score/probability output, finance output, persistence, exports, live connectors, and customer-facing output.
• AI Value contribution alignment internal research-design gate review: closes the current internal design chain by reviewing whether the compact prototype review packet is safe enough for a later exact-scope method-prototype decision. It remains non-persistent, compact-ref-only, and blocks model feeds, model implementation, numeric weights, score/probability output, finance output, persistence, exports, live connectors, and customer-facing output.
• AI Value contribution alignment method prototype decision: records the exact-scope decision to promote only a small internal qualitative method prototype after the internal research-design gate review passes. It remains non-persistent, compact-ref-only, and blocks model feeds, model implementation, numeric weights, score/probability output, finance output, persistence, exports, live connectors, UI, routes, schemas, and customer-facing output.
• AI Value contribution alignment small internal method prototype: turns the promoted method scope into compact qualitative component posture for internal review only. It keeps AI Fluency construct context, psychological context, observed VBD context, and selected customer metric movement distinct, and it blocks model results, numeric weights, score/probability output, finance output, persistence, exports, live connectors, UI, routes, schemas, and customer-facing output.
• AI Value contribution alignment internal method prototype review record: reviews the compact qualitative method prototype and promotes only a separate exact-scope research math finalization review. It remains non-persistent, compact-ref-only, and blocks research math output, model feeds, model implementation, numeric weights, score/probability output, finance output, persistence, exports, live connectors, UI, routes, schemas, and customer-facing output.
• AI Value contribution alignment research math finalization review: closes the promised exact-scope review step before any data-model promotion. It consumes compact source-bound review record refs only and authorizes only a later research math data model promotion decision. It remains non-persistent and blocks research math output, model feeds, model implementation, numeric weights, score/probability output, finance output, persistence, exports, live connectors, UI, routes, schemas, and customer-facing output.
• AI Value contribution alignment research math data model promotion decision: consumes the finalization review and promotes only a compact internal research-math data model layer. It explicitly holds if only the older review record is supplied, and it blocks physical tables, persistence, research math output, model feeds, numeric weights, score/probability output, finance output, exports, live connectors, UI, routes, schemas, and customer-facing output.
• AI Value contribution alignment internal research math data model: defines the compact internal context grain for later research design: one approved expectation path ref, one source-bound Measurement Cell context ref, repeated Day 0 / 30 / 60 / 90 / 180 / 365 milestone refs, context-only component registry, and separate AI Fluency construct, AI Fluency psychological, observed VBD, and selected metric movement partitions. It remains non-persistent and emits no model result, numeric weights, score/probability output, finance output, exports, live connectors, UI, routes, schemas, or customer-facing output.
• AI Value contribution alignment feature stability review: reviews the internal research math data model feature inputs against a world-class pre-weight standard: source-bound id/hash, stable component registry, distinct context partitions, repeated milestone requirement, and false forbidden-output feeds. A pass authorizes only a separate internal numeric weight decision; it does not authorize weights, weighted model output, confidence/probability output, Bayesian execution, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, or customer-facing output.
• AI Value contribution alignment internal numeric weight decision: consumes the feature stability review and authorizes only a later versioned internal weight object. It contains no weight values, emits no weighted model frame, and keeps confidence/probability output, Bayesian execution, score-like output, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, and customer-facing output blocked.
• AI Value contribution alignment versioned weight object: consumes the internal numeric weight decision and creates the first internal-only structural weight set: internal_structural_equal_weights_2026_06. It assigns neutral equal weights across the ten governed source-bound feature inputs, records initial_internal_structural_weights_not_empirical_confidence, and feeds only a later weighted internal model frame. It does not emit weighted model output, research model feed, confidence/probability output, Bayesian execution, score-like output, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, or customer-facing output.
• AI Value contribution alignment weighted internal model frame: consumes the versioned weight object and attaches its source-bound weights to the ten governed feature inputs as internal weighted feature composition only. It is the full internal weighted data model frame, not a model result: it keeps weighted internal model output, aggregate score output, research model feed, confidence/probability output, Bayesian execution, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, and customer-facing output blocked while allowing only a later internal Bayesian readiness review.
• AI Value contribution alignment internal Bayesian readiness review: consumes the weighted internal model frame and authorizes only a later Bayesian model specification contract. It names bayesian_hierarchical_difference_in_differences_candidate as the specialized DiD candidate model_family field value, not the canonical Bayesian architecture family, and does not define priors, likelihood, estimands, posterior output, confidence/probability output, Bayesian execution, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, or customer-facing output.
• AI Value contribution alignment Bayesian model specification: consumes the internal Bayesian readiness review and defines the internal-only candidate model contract: bayesian_hierarchical_did_spec_2026_06. It records the aggregate Measurement Cell window unit, candidate difference-in-differences estimand, and uncalibrated prior/likelihood placeholders, but it does not run Bayesian execution, emit posterior/confidence/probability output, create score-like output, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, or customer-facing output.
• AI Value contribution alignment internal Bayesian execution gate: consumes the Bayesian model specification and authorizes only a later internal execution runtime implementation. It binds to the specification id/hash, readiness-review ref, weighted-frame ref, and governed feature weights; records aggregate-only runtime prerequisites; and requires a later posterior/output review gate before any confidence or probability language can appear. It does not run Bayesian execution, emit posterior/confidence/probability output, create score-like output, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, or customer-facing output.
• AI Value contribution alignment internal Bayesian execution runtime: consumes the execution gate and aggregate Measurement Cell windows to create a contained fixture/prototype Bayesian difference-in-differences artifact. It keeps the closed-form normal-normal update for delta_ai_post inside the internal fixture artifact, records missing diagnostics, and authorizes only internal_diagnostics_and_model_adequacy_review_only. It does not emit posterior/confidence/probability language, create score-like output, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, or customer-facing output.
• AI Value contribution alignment posterior output review gate: consumes the internal Bayesian execution runtime and reviews the internal fixture artifact by ref/hash without echoing posterior numeric values. It is an artifact-containment review only, authorizes only internal_diagnostics_and_model_adequacy_review_only, and keeps internal posterior interpretation specification, posterior output, confidence output, probability output, finance output, ROI, causality, productivity, persistence, exports, live connectors, UI, routes, schemas, and customer-facing output blocked.
• AI Value contribution alignment diagnostics evidence packet: consumes the contained internal Bayesian fixture runtime and represents the evidence needed for diagnostics/model-adequacy review: data adequacy, suppressed/missing/held windows, comparison-design adequacy, convergence diagnostics, posterior predictive checks, prior sensitivity, residual/fit checks, calibration/backtest evidence, and feature-weight provenance. The current packet is ready for promotion-decision review but not promotion-sufficient: data/window/weight evidence is satisfied while comparison-design and model diagnostics remain unsatisfied. It cannot authorize promotion, posterior interpretation, confidence/probability language, customer-facing output, ROI, finance, causality, productivity, persistence, exports, live connectors, UI, routes, or schemas.
• AI Value contribution alignment governed diagnostics sufficiency evidence source: provides the internal-only, aggregate-only source contract that may supply diagnostics and comparison-design sufficiency evidence to the Diagnostics Evidence Packet. It holds by default unless each satisfied dimension has an explicit reviewed source evidence ref, an independent reviewed source evidence hash, and a derived source evidence hash bound to the runtime and fixture artifact. It cannot authorize promotion, posterior interpretation, confidence/probability language, customer-facing output, ROI, finance, causality, productivity, persistence, exports, live connectors, UI, routes, or schemas.
• AI Value contribution alignment internal diagnostics and model adequacy review: consumes the contained internal Bayesian fixture runtime and reviews data adequacy, comparison-design adequacy, and diagnostic placeholder status. It completes only as INTERNAL_DIAGNOSTICS_AND_MODEL_ADEQUACY_REVIEW_COMPLETED_PROMOTION_BLOCKED, may feed only bayesian_promotion_decision_gate_only, keeps feature weights structural/internal rather than confidence scores, and keeps Bayesian interpretation, confidence/probability language, customer-facing output, ROI, finance, causality, productivity, persistence, exports, live connectors, UI, routes, and schemas blocked. This slice does not implement the Bayesian Promotion Decision Gate.
• AI Value contribution alignment Bayesian promotion decision gate: consumes the diagnostics/model adequacy review and decides whether a contained fixture may move only to a later Internal Bayesian Execution Artifact v1 slice. It may authorize only internal_bayesian_execution_artifact_v1_only and keeps posterior interpretation, confidence/probability language, customer-facing output, ROI, finance, causality, productivity, persistence, exports, live connectors, UI, routes, and schemas blocked. This slice does not create the execution artifact.
• AI Value contribution alignment promotion gate passed artifact handoff: records a passive, hash-bound handoff for an already-passed Bayesian Promotion Decision Gate and its exact runtime, diagnostics review, evidence packet, and governed diagnostics sufficiency evidence source hashes. The default executable path remains held; the explicit governed-evidence path may produce a passed handoff only as PROMOTION_GATE_PASSED_ARTIFACT_HANDOFF_READY_FOR_INTERNAL_EXECUTION_ARTIFACT_V1_CONTRACT_HANDOFF_ONLY. The handoff itself keeps promotion_authorized=false, creates no Internal Bayesian Execution Artifact v1, and keeps posterior interpretation, confidence/probability language, customer-facing output, ROI, finance, causality, productivity, persistence, exports, live connectors, UI, routes, and schemas blocked.
• AI Value contribution alignment Internal Bayesian Execution Artifact v1: creates the internal-only, aggregate-only execution artifact record authorized by the passed Promotion Gate Passed Artifact Handoff and passed Bayesian Promotion Decision Gate. It is source/hash-bound to the handoff, promotion gate, runtime, diagnostics review, evidence packet, and governed diagnostics source. It does not rerun Bayesian execution, reinterpret posterior-like prototype values, or authorize posterior interpretation, confidence/probability language, customer-facing output, ROI, finance, causality, productivity, persistence, exports, live connectors, UI, routes, or schemas. Its only allowed next step is posterior_interpretation_specification_gate_only.
• AI Value contribution alignment Bayesian Hardening Orchestrator: emits a read-only, internal-only handoff report over the existing Bayesian hardening chain. Default execution stops at the first held governed diagnostics evidence gate; an explicit governed-evidence path may be reported only when the existing source-bound gates validate. The orchestrator does not create new model artifacts and does not create governed evidence. Orchestrator promotion authority remains false; existing Bayesian Promotion Decision Gate authority may be reported only as source evidence. It keeps posterior interpretation, confidence/probability language, customer-facing output, ROI, finance, causality, productivity, persistence, exports, live connectors, UI, routes, and schemas blocked. Its allowed next step is derived from the validated existing gate chain only.
• AI Value confidence-engine workspace: promotes the 16-module contribution-alignment Bayesian execution spine into the typed packages/confidence-engine workspace with byte-identical outputs: every schema version, state token, hash, and property insertion order is preserved and enforced by golden-fixture parity gates plus the migrated validation suites (npm run test:confidence-engine-workspace); the spine scripts/ runners remain as thin wrappers preserving CLI usage and named exports, and the ConfidenceModel contract module is types + Zod schemas only. The port changes no model math or gate semantics: all held states stay held, promotion stays blocked, and posterior interpretation, confidence/probability/score output, finance output, ROI, causality, productivity, customer-facing output, routes, UI, schemas, persistence, exports, and live connectors remain blocked.
• Confidence inference methodology contract: the normative specification for the current specialized comparison-supported Bayesian DiD proof module (docs/contracts/confidence-inference-methodology/README.md): a hierarchical Bayesian difference-in-differences estimand over aggregate Measurement Cell windows, a Python-owns-statistics / TypeScript-owns-governance boundary, an internal-only InferenceProofArtifactSchema, seven diagnostics as computed values with hard numeric gates (any failure holds the artifact), fixed posterior predictive checks, per-scenario calibration proof, fail-closed negative controls, the comparison-cohort rule ("no credible comparison cohort, no comparison-supported contribution estimate — evidence-tier label only") backed by a runnable adequacy rubric, fixed-horizon one-look peeking by default, empirically justified weakly-informative priors with mandatory sensitivity reporting, k>=5/k>=10 aggregate floors, and Value-Playbook-aligned claim language where customer-facing statuses are future ceilings only. Threshold-probability and expected-loss representations exist as internal-only types with customer output pinned false; no probability language is exposed anywhere pending a separate recorded human promotion decision.
• Bayesian AI Value Realization And Human Transformation model family: the single canonical Bayesian architecture family (docs/contracts/bayesian-ai-value-realization-and-human-transformation-model-family/README.md). It keeps the current DiD path as comparison_supported_bayesian_did_module and defines components for aggregate Fluency, VBD, hypothesis outcome, economic-value context, pathway coherence, claim caps, and portfolio review. The first non-DiD component now has an independently accepted synthetic longitudinal state-space model proof: 30 full-setting NUTS concordance fits, 1,200 deterministic replicated-validation slots, null/floor/lag/shock controls, and strict Python/TypeScript evidence bindings. AI Fluency measurement calibration is accepted only at the internal synthetic boundary; VBD tasks 2.2 through 2.5 are implemented, but numerical-precision repair, canaries, concordance, evidence, and acceptance remain held. This is not production promotion. "Human transformation" means aggregate work-pattern and capability-change context only; it does not authorize HR analytics, individual scoring, employee productivity measurement, manager/team ranking, customer-facing confidence/probability output, ROI proof, finance output, causality claims, persistence, routes, UI, exports, live connector reads, or economic output.
• V4 Value Confidence Layer: combines Velocity and Depth with governed V3 verdicts to qualify the defensibility of AI value claims.
• AI Scale Readiness Portfolio: V4 internal readout contract that turns aggregate evidence into action postures for scale, enablement, workflow redesign, trust calibration, adoption expansion, value investigation, or hold. It is not customer-facing and does not calculate economic value.
• Organizational Segmentation: future V4 concept for aggregate-only intervention contexts such as tenure, function, role family, or behavior bands; never person, manager, or comparative team evaluation.
• Economic Impact Bridge: future V4 concept that maps trusted readiness patterns to customer-owned value investigations, value metric candidates, and governed scenario review without proving ROI or causality.
• AI Manager Outcomes Recommendations: docs-first V4 layer that recommends which customer-owned outcome signals and aggregate formulas to use next when testing cost, revenue, quality, capacity, risk, or experience value routes. Accepted aggregate evidence may be routed into value metric selection and scenario review, but not automatic economic proof or customer-facing economic output. "AI Manager" means AI program owner or value-realization leader, not people manager scoring.
• Data Boundary and ROI Evidence: defines which organizational data may be useful upstream for value analysis, how it must be transformed before crossing into FluencyTracr, and which aggregate evidence can feed value-evidence cases. Sensitive HRIS, finance, revenue, workflow, support, and quality data can inform value modeling only after aggregation, attestation, and identifier removal; FluencyTracr still does not store raw rows, calculate realized ROI, prove causality, run HR analytics, or emit customer-facing economic output.
• Work Mode Taxonomy: maps governed surfaces into durable AI work patterns such as retrieval, conversation, transformation, embedded assist, delegation, reuse, exploration, verification, and corroborative telemetry.

V4 is the Value Confidence Layer. It combines Velocity and Depth with governed V3 verdicts to qualify the defensibility of AI value claims.

The aggregate verdict layer uses AIVM vocabulary consistently: value_type communicates the kind of value claim, and evidence_grade communicates whether the support is OBJECTIVE, CALIBRATED, or QUALITATIVE. Value-realization services should preserve those fields when they consume verdict metadata, and suppression remains fail-closed.

The audience is AIOMs, value-realization PMs, and CIOs deciding which Glean value claims are defensible. This is not an HR, learning, or individual measurement product.

AI assistants — start every session here

Long-form or multi-session coding must begin with docs/agent/SESSION_START.md so work stays bounded, verified, and grounded in repo memory (queue + harness + git), not chat context alone. That doc aligns with Anthropic’s guidance on long-running agent harnesses as implemented under harness/.

Scope Guardrails

This project intentionally rejects surveillance and scope creep. Please read and follow the guardrails in SCOPE_GUARDRAILS.md before proposing changes.

Evidence Layer Invariants

The governance invariants are the proof of seriousness behind the value realization story. They are not the headline, but they make the headline credible:

• Outputs are signals, not facts, and are strictly binary: SURFACE or SUPPRESS.
• Default state is SUPPRESS; ambiguity is first-class and always suppressive.
• No content storage or individual attribution is permitted.
• Latency is corroborative only and never triggers surfacing on its own.
• No tunable thresholds or admin overrides are allowed; constants are compiled into code.

Suppression reason codes (one-hot, immutable):

• INSUFFICIENT_TIME
• INSUFFICIENT_VOLUME
• NO_CONVERGENCE
• BASELINE_UNSTABLE
• HIGH_AMBIGUITY

Complementary Stated-Evidence Layer

The AI Fluency Instrument remains useful as a stated-evidence layer: it captures what people report about adoption, confidence, and practice. FluencyTracr should be paired with that instrument when value teams want to compare stated evidence with observed aggregate workflow evidence. The instrument is complementary; it is not the lead positioning for this repository.

Canonical Docs

• Every session (mandatory for agents): docs/agent/SESSION_START.md
• AI / agent navigation (harness, evaluation, task contracts): docs/agent/README.md
• Agentic execution harness concept: docs/concepts/AGENTIC_EXECUTION_HARNESS.md
• Cursor agent harness rules: docs/agent/cursor-agent-harness.md
• Optional OpenAI Agents SDK development harness: docs/agent/openai-agents-harness.md
• Partner ingestion facade: docs/api/ingest.md
• Unified telemetry ingest (UT_2026_04, feature-flagged): docs/api/ingest-unified-telemetry.md
• Trace reconstruction (PRD Phase 1, admin): docs/api/traces-reconstructed.md
• Workflow observability rollup (PRD Phase 4): docs/api/observability-org.md
• Architecture map: docs/ARCHITECTURE_MAP.md
• Connector mapping spec: docs/CONNECTOR_MAPPING_SPEC.md
• EvidenceBundle v1 contract: docs/contracts/evidence-bundle/v1/README.md
• Reportability contract: docs/contracts/reportability/README.md
• Agent-run development harness contract: docs/contracts/agent-run/README.md
• Agent-run ledger contract framing: docs/contracts/agent-run/ledger.md
• Unified cross-surface telemetry (UT_2026_04): docs/contracts/unified-telemetry/README.md
• AI Value Intelligence evidence pack contract: docs/contracts/ai-value-intelligence/README.md
• Customer Support AI Value pilot design: docs/contracts/ai-value-intelligence/customer-support-pilot-design.md
• Customer Support AI Value pilot dry run: docs/contracts/ai-value-intelligence/customer-support-pilot-dry-run.md
• Customer Support AI Value validation workshop kit: docs/contracts/ai-value-intelligence/customer-support-validation-workshop-kit.md
• Customer Support AI Value workshop response fixture: docs/contracts/ai-value-intelligence/examples/customer-support-workshop-response.json
• AI Value Blueprint schema: schemas/ai-value-intelligence/blueprint.schema.json
• Customer Support AI Value blueprint fixture: docs/contracts/ai-value-intelligence/examples/customer-support-blueprint.json
• AI Value Metric Definition schema: schemas/ai-value-intelligence/metric-definition.schema.json
• Customer Support AI Value metrics fixture: docs/contracts/ai-value-intelligence/examples/customer-support-metrics-library.json
• AI Value Scenario input schema: schemas/ai-value-intelligence/value-scenario-input.schema.json
• AI Value Scenario output schema: schemas/ai-value-intelligence/value-scenario-output.schema.json
• Customer Support AI Value scenario fixture: docs/contracts/ai-value-intelligence/examples/customer-support-value-scenario.json
• AI Value governed value-scenario schema: schemas/ai-value-intelligence/roi-scenario.schema.json
• Customer Support governed value-scenario fixture: docs/contracts/ai-value-intelligence/examples/customer-support-roi-scenario.json
• AI Value data boundary and ROI evidence schema: schemas/ai-value-intelligence/data-boundary-roi-evidence.schema.json
• Customer Support data boundary and ROI evidence fixture: docs/contracts/ai-value-intelligence/examples/customer-support-data-boundary-roi-evidence.json
• AI Value Evidence Case schema: schemas/ai-value-intelligence/value-evidence-case.schema.json
• Customer Support AI Value evidence case fixture: docs/contracts/ai-value-intelligence/examples/customer-support-value-evidence-case.json
• AI Value Agentic Platform Harness concept: docs/concepts/AI_VALUE_AGENTIC_PLATFORM_HARNESS.md
• AI Value Agent Handoff schema: schemas/ai-value-intelligence/agent-handoff.schema.json
• Customer Support AI Value agent handoff fixture: docs/contracts/ai-value-intelligence/examples/customer-support-agent-handoff.json
• AI Value Evidence Readiness schema: schemas/ai-value-intelligence/evidence-readiness.schema.json
• Customer Support AI Value readiness fixture: docs/contracts/ai-value-intelligence/examples/customer-support-evidence-readiness.json
• AI Value Hypothesis Readiness contract: docs/contracts/ai-value-value-hypothesis-readiness/README.md
• AI Value Hypothesis Readiness Packet Runner: docs/contracts/ai-value-hypothesis-readiness-packet-runner/README.md
• AI Value Real Data Intake Packet Runner: docs/contracts/ai-value-real-data-intake-packet-runner/README.md
• AI Value Operator Intake Adapter: docs/contracts/ai-value-operator-intake-adapter/README.md
• AI Value VBD + Token Operator Source Handoff: docs/contracts/ai-value-vbd-token-operator-source-handoff/README.md
• AI Value Customer Metric Operator Source Handoff: docs/contracts/ai-value-customer-metric-operator-source-handoff/README.md
• AI Value Assumption / Governance Operator Source Handoff: docs/contracts/ai-value-assumption-governance-operator-source-handoff/README.md
• AI Value Operator Source Handoff Bundle: docs/contracts/ai-value-operator-source-handoff-bundle/README.md
• AI Value Measurement Cell Series: docs/contracts/ai-value-measurement-cell-series/README.md
• AI Value Operator Time-Series Run: docs/contracts/ai-value-operator-time-series-run/README.md
• AI Value Operator Workflow: docs/contracts/ai-value-operator-workflow/README.md
• AI Value Operator Evidence Package Runner: docs/contracts/ai-value-operator-evidence-package-runner/README.md
• AI Value Controlled Aggregate Fixture Review: docs/contracts/ai-value-controlled-aggregate-fixture-review/README.md
• AI Value Controlled Measurement Cell Assembly: docs/contracts/ai-value-controlled-measurement-cell-assembly/README.md
• AI Value Controlled Aggregate Pipeline Dry Run: docs/contracts/ai-value-controlled-aggregate-pipeline-dry-run/README.md
• AI Value Measurement Cell Preflight Runner: docs/contracts/ai-value-measurement-cell-preflight-runner/README.md
• AI Value Measurement Cell Snapshot Projection: docs/contracts/ai-value-measurement-cell-snapshot-projection/README.md
• AI Value Customer Data Model Promotion Gate: docs/contracts/ai-value-customer-data-model-promotion-gate/README.md
• AI Value Customer Data Model Persistence Promotion Decision: docs/contracts/ai-value-customer-data-model-persistence-promotion-decision/README.md
• AI Value Customer Data Model Persistence Implementation Decision: docs/contracts/ai-value-customer-data-model-persistence-implementation-decision/README.md
• AI Value Customer Data Model Route Projection: docs/contracts/ai-value-customer-data-model-route-projection/README.md
• AI Value Data Model Spine Readiness Lock: docs/contracts/ai-value-data-model-spine-readiness-lock/README.md
• AI Value Compact Source Wiring Hardening: docs/contracts/ai-value-compact-source-wiring-hardening/README.md
• AI Value Connector Promotion Readiness Sequence: docs/contracts/ai-value-connector-promotion-readiness-sequence/README.md
• AI Value Controlled Aggregate Connector Adapter: docs/contracts/ai-value-controlled-aggregate-connector-adapter/README.md
• AI Value Aggregate Connector Boundary Plan: docs/contracts/ai-value-aggregate-connector-boundary-plan/README.md
• AI Value BigQuery Aggregate Export Review: docs/contracts/ai-value-bigquery-aggregate-export-review/README.md
• AI Value Controlled Aggregate Manifest Validation: docs/contracts/ai-value-controlled-aggregate-manifest-validation/README.md
• AI Value Research Promotion Readiness Packet: docs/contracts/ai-value-research-promotion-readiness-packet/README.md
• AI Value Contribution Alignment Feature Stability Review: docs/contracts/ai-value-contribution-alignment-feature-stability-review/README.md
• AI Value Contribution Alignment Internal Numeric Weight Decision: docs/contracts/ai-value-contribution-alignment-internal-numeric-weight-decision/README.md
• AI Value Contribution Alignment Versioned Weight Object: docs/contracts/ai-value-contribution-alignment-versioned-weight-object/README.md
• AI Value Contribution Alignment Weighted Internal Model Frame: docs/contracts/ai-value-contribution-alignment-weighted-internal-model-frame/README.md
• AI Value Contribution Alignment Internal Bayesian Readiness Review: docs/contracts/ai-value-contribution-alignment-internal-bayesian-readiness-review/README.md
• AI Value Contribution Alignment Bayesian Model Specification: docs/contracts/ai-value-contribution-alignment-bayesian-model-specification/README.md
• AI Value Contribution Alignment Internal Bayesian Execution Gate: docs/contracts/ai-value-contribution-alignment-internal-bayesian-execution-gate/README.md
• AI Value Contribution Alignment Internal Bayesian Execution Runtime: docs/contracts/ai-value-contribution-alignment-internal-bayesian-execution-runtime/README.md
• AI Value Contribution Alignment Posterior Output Review Gate: docs/contracts/ai-value-contribution-alignment-posterior-output-review-gate/README.md
• AI Value Contribution Alignment Diagnostics Evidence Packet: docs/contracts/ai-value-contribution-alignment-diagnostics-evidence-packet/README.md
• AI Value Contribution Alignment Governed Diagnostics Sufficiency Evidence Source: docs/contracts/ai-value-contribution-alignment-governed-diagnostics-sufficiency-evidence-source/README.md
• AI Value Contribution Alignment Internal Diagnostics and Model Adequacy Review: docs/contracts/ai-value-contribution-alignment-internal-diagnostics-model-adequacy-review/README.md
• AI Value Contribution Alignment Bayesian Promotion Decision Gate: docs/contracts/ai-value-contribution-alignment-bayesian-promotion-decision-gate/README.md
• AI Value Contribution Alignment Promotion Gate Passed Artifact Handoff: docs/contracts/ai-value-contribution-alignment-promotion-gate-passed-artifact-handoff/README.md
• AI Value Contribution Alignment Internal Bayesian Execution Artifact v1: docs/contracts/ai-value-contribution-alignment-internal-bayesian-execution-artifact-v1/README.md
• AI Value Contribution Alignment Bayesian Hardening Orchestrator: docs/contracts/ai-value-contribution-alignment-bayesian-hardening-orchestrator/README.md
• AI Value Source Inventory Manifest: docs/contracts/ai-value-source-inventory-manifest/README.md
• AI Value Aggregate Extraction Manifest: docs/contracts/ai-value-aggregate-extraction-manifest/README.md
• AI Value Pipeline Run Review Manifest: docs/contracts/ai-value-pipeline-run-review-manifest/README.md
• AI Value Claim Boundary schema: schemas/ai-value-intelligence/claim-boundary.schema.json
• Customer Support AI Value claim boundary fixture: docs/contracts/ai-value-intelligence/examples/customer-support-claim-boundary.json
• AI Value Executive Packet schema: schemas/ai-value-intelligence/executive-packet.schema.json
• Customer Support AI Value executive packet fixture: docs/contracts/ai-value-intelligence/examples/customer-support-executive-packet.json
• Local AI Value Workspace UI: frontend/src/pages/AIValueWorkspace.tsx
• Glean integration pack: docs/integrations/glean/01-overview.md
• Glean readiness demo guide: docs/integrations/glean/06-readiness-demo-guide.md
• Glean executive readiness prototype: docs/integrations/glean/prototypes/executive-readiness-demo.html
• Glean live-data decision gate: docs/integrations/glean/07-live-data-access-decision-gate.md
• Value-realization contract index: docs/integrations/value-realization/INDEX.md
• V3 production ingest walkthrough: docs/integrations/value-realization/V3_INGEST.md
• V4 value confidence integration: docs/integrations/value-realization/V4_VALUE_CONFIDENCE.md
• Internal Scale Readiness Readout contract: docs/contracts/value-confidence/internal-scale-readiness-readout.md
• AI Work Evidence concept: docs/concepts/AI_WORK_EVIDENCE.md
• AI Value Measurement Model and phase data support map: docs/concepts/AI_VALUE_MEASUREMENT_MODEL.md
• AI Value Platform architecture: docs/concepts/AI_VALUE_PLATFORM_ARCHITECTURE.md
• AI Value Intelligence MVP concept: docs/concepts/AI_VALUE_INTELLIGENCE_MVP.md
• AI Work Evidence pilot package: docs/integrations/value-realization/AI_WORK_EVIDENCE_PILOT_PACKAGE.md
• Velocity concept and V2 reference: docs/concepts/VELOCITY.md
• Depth concept and V4 reference: docs/concepts/DEPTH.md
• Delegation Depth concept and V4 Depth subdimension: docs/concepts/DELEGATION_DEPTH.md
• Work Mode Taxonomy concept: docs/concepts/WORK_MODES.md
• V4 Value Confidence Layer concept: docs/concepts/V4_VALUE_CONFIDENCE_LAYER.md
• AI Scale Readiness Portfolio concept: docs/concepts/AI_SCALE_READINESS_PORTFOLIO.md
• Organizational Segmentation concept: docs/concepts/ORG_SEGMENTATION.md
• Economic Impact Bridge concept: docs/concepts/ECONOMIC_IMPACT_BRIDGE.md
• AI Manager Outcomes Recommendations concept: docs/concepts/AI_MANAGER_OUTCOMES_RECOMMENDATIONS.md
• AI Manager Outcomes Recommendations contract: docs/contracts/value-confidence/ai-manager-outcomes-recommendations.md
• AI Manager Outcomes Recommendations pilot readout: docs/research/AI_MANAGER_OUTCOMES_RECOMMENDATIONS_PILOT_READOUT.md
• Signal promotion criteria: docs/research/SIGNAL_PROMOTION_CRITERIA.md
• V4 signal discovery probes: docs/research/V4_SIGNAL_DISCOVERY_PROBES.md
• V4 signal discovery readout: docs/research/V4_SIGNAL_DISCOVERY_READOUT.md
• V4 signal validation gate: docs/research/V4_SIGNAL_VALIDATION_GATE.md
• V4 signal validation runbook: docs/research/V4_SIGNAL_VALIDATION_RUNBOOK.md
• V4 signal validation readout template: docs/research/V4_SIGNAL_VALIDATION_READOUT_TEMPLATE.md
• Trust Episode Boundary research: docs/research/TRUST_EPISODE_BOUNDARY.md
• Trust Episode Boundary validation readout: docs/research/V4_TRUST_EPISODE_BOUNDARY_VALIDATION_READOUT.md
• Trust Episode Boundary input contract proposal: docs/contracts/value-confidence/trust-episode-boundary-input.md
• Trust Episode Boundary pilot runbook: docs/dogfood/TRUST_EPISODE_PILOT_RUNBOOK.md
• Trust Calibration external pilot brief: output/doc/FluencyTracr_Trust_Calibration_Pilot_Brief.docx
• V4 Depth readout runbook: docs/research/V4_DEPTH_READOUT_RUNBOOK.md
• V4 Depth stability decision: docs/research/V4_DEPTH_STABILITY_DECISION.md
• V4 Depth Repertoire stability readout: docs/research/V4_DEPTH_REPERTOIRE_STABILITY_READOUT.md
• V4 Depth Repertoire value-confidence review: docs/research/V4_DEPTH_REPERTOIRE_VALUE_CONFIDENCE_REVIEW.md
• V4 Depth Repertoire calibration plan: docs/research/V4_DEPTH_REPERTOIRE_VALUE_CONFIDENCE_CALIBRATION_PLAN.md
• V4 Depth Repertoire calibration decision: docs/research/V4_DEPTH_REPERTOIRE_VALUE_CONFIDENCE_CALIBRATION_DECISION.md
• V4 Scale Readiness caveat propagation decision: docs/research/V4_SCALE_READINESS_CAVEAT_PROPAGATION_DECISION.md
• V4 Trust Calibration caveat propagation decision: docs/research/V4_TRUST_CALIBRATION_CAVEAT_PROPAGATION_DECISION.md
• V4 Glean dogfood scale readiness readout: docs/research/V4_GLEAN_DOGFOOD_SCALE_READINESS_READOUT.md
• V4 full dogfood rehearsal readout: docs/research/V4_FULL_DOGFOOD_REHEARSAL_READOUT.md
• V4 behavior cohort classification readout: docs/research/V4_BEHAVIOR_COHORT_CLASSIFICATION_READOUT.md
• V4 data analysis closeout: docs/research/V4_DATA_ANALYSIS_CLOSEOUT.md
• V4 next sprint plan: docs/research/V4_NEXT_SPRINT_PLAN.md
• V4 measurement build plan: docs/research/V4_MEASUREMENT_BUILD_PLAN.md
• V4 segment overlay test plan: docs/research/V4_SEGMENT_OVERLAY_TEST_PLAN.md
• V4 intervention tracking research design: docs/research/V4_INTERVENTION_TRACKING_RESEARCH_DESIGN.md
• V4 outcome join test plan: docs/research/V4_OUTCOME_JOIN_TEST_PLAN.md
• V4 outcome source inventory readout: docs/research/V4_OUTCOME_SOURCE_INVENTORY_READOUT.md
• V4 support outcome join test readout: docs/research/V4_SUPPORT_OUTCOME_JOIN_TEST_READOUT.md
• V4 support join-key test readout: docs/research/V4_SUPPORT_JOIN_KEY_TEST_READOUT.md
• V4 Time-Saved Defensibility test plan: docs/research/V4_TIME_SAVED_DEFENSIBILITY_TEST_PLAN.md
• V4 readout zone model: docs/research/V4_READOUT_ZONE_MODEL.md
• V4 behavior feature backlog: docs/research/V4_BEHAVIOR_FEATURE_BACKLOG.md
• V4 value hypothesis map: docs/research/V4_VALUE_HYPOTHESIS_MAP.md
• V4 value realization strategy layer: docs/research/V4_VALUE_REALIZATION_STRATEGY_LAYER.md
• V4 readout zone data test: docs/research/V4_READOUT_ZONE_DATA_TEST.md
• V4 Velocity x Depth zone test: docs/research/V4_VELOCITY_DEPTH_ZONE_TEST.md
• V4 Glean dogfood decision: docs/research/V4_GLEAN_DOGFOOD_DECISION.md
• V4 closeout decision: docs/research/V4_CLOSEOUT_DECISION.md
• V4 canonical contract decision: docs/research/V4_CANONICAL_CONTRACT_DECISION.md
• V4 internal readout runbook: docs/research/V4_INTERNAL_READOUT_RUNBOOK.md
• V4 internal readout rehearsal: docs/research/V4_INTERNAL_READOUT_REHEARSAL.md
• V4 Trust Attribution method: docs/research/V4_TRUST_ATTRIBUTION_METHOD.md
• V4 Trust Attribution test readout: docs/research/V4_TRUST_ATTRIBUTION_TEST_READOUT.md
• V4 Trust and Cohort classification plan: docs/research/V4_TRUST_AND_COHORT_CLASSIFICATION_PLAN.md
• V4 value-confidence caveat propagation runbook: docs/research/V4_VALUE_CONFIDENCE_CAVEAT_PROPAGATION_RUNBOOK.md
• V4 TSDR caveat propagation decision: docs/research/V4_TSDR_CAVEAT_PROPAGATION_DECISION.md
• V4 AI Value Leakage Map caveat propagation decision: docs/research/V4_VALUE_LEAKAGE_CAVEAT_PROPAGATION_DECISION.md
• V4 validation plan: docs/research/V4_VALIDATION_PLAN.md
• Surface taxonomy concept: docs/concepts/SURFACES.md
• Agent types concept: docs/concepts/AGENT_TYPES.md
• Production ingest concept: docs/concepts/INGEST.md
• Calibration governance concept: docs/concepts/CALIBRATION.md
• Depth readout contract: docs/contracts/depth/README.md
• Depth Repertoire contract: docs/contracts/depth/depth-repertoire.md
• Value Confidence contract: docs/contracts/value-confidence/README.md
• Time-Saved Defensibility Range: docs/contracts/value-confidence/time-saved-defensibility-range.md
• Trust Calibration Index: docs/contracts/value-confidence/trust-calibration-index.md
• AI Value Leakage Map: docs/contracts/value-confidence/value-leakage-map.md
• AI Scale Readiness Portfolio: docs/contracts/value-confidence/scale-readiness-portfolio.md
• AI Manager Outcomes Recommendations: docs/contracts/value-confidence/ai-manager-outcomes-recommendations.md
• AI Value Finance Investigation Readiness: docs/contracts/value-confidence/ai-value-finance-investigation-readiness.md
• Velocity Index contract: docs/contracts/velocity-index.md
• Behavioral Signals spec: docs/BEHAVIORAL_SIGNALS_SPEC.md
• MCP adapter server: docs/mcp/fluencytracr-mcp-server.md
• API reference: docs/api/API_REFERENCE.md
• System architecture overview: docs/architecture/SYSTEM_ARCHITECTURE_OVERVIEW.md