Glossary

Cryptographic proof that an authorized operator reviewed a finding or submission before the certification process advanced.

The intelligence loop that converts certified information into alerts, recommendations, workflows, and decisions.

The advisory AI capability that helps users interpret certified information but cannot alter or produce certification results.

A convincing AI output built on incomplete, inconsistent, or unverified data: statistically plausible, but factually wrong.

The distance between the confidence of an AI output and the reliability of the operational evidence beneath it.

The condition in which increasingly capable AI produces increasingly confident decisions from foundations the enterprise cannot prove.

A recorded statement that an authorized person or system reviewed, supplied, or affirmed defined information.

The append-only record of what was received, changed, tested, approved, certified, and delivered—and when.

Step 3: where data is normalized into canonical MGE structures.

The customer-approved definitions, contracts, policies, and operating rules against which its data is certified.

The sealed evidence package Sentry produces for a recovery claim — findings, calculations, source references, and the certification narrative, in a form built to survive challenge.

The CAAR and its complete supporting evidence, verification, attestation, and chain-of-custody materials delivered as one reviewable package.

A presentation-oriented name for the consolidated CAAR and supporting transaction-level evidence assembled for audit, dispute, arbitration, or recovery.

The complete supporting package for a CAAR, including the report, exhibits, manifests, hashes, rule versions, lineage, and chain-of-custody records.

The standard data structure into which different source formats are mapped so they can be evaluated consistently.

The process of establishing that an operational fact is provable: its source is authoritative, its formula is approved, its reconciliation succeeded, and the result is sealed. Certification happens before data reaches a system of record or an AI model.

One auditable execution of MGE against a defined data set, metric, period, entity, or transaction.

The point after which a certification must be refreshed because its data, rules, contracts, or operating conditions may no longer be current.

The missing control between data collection and downstream analysis: data may be unified, governed as metadata, and analyzed without ever being verified as operational truth.

The independent control layer between source systems and the systems, reports, or AI applications that consume their data.

The machine-readable record of a certification event, including inputs, rules, gate results, scores, exceptions, timestamps, and outputs.

MGE’s controlled sequence: Ingest, Seal, Normalize, Reconcile, Apply Rules, Score, Narrative Seal, and Output.

The status assigned to a certification result according to its evidence and Trust Score: Unverified, Provisional, Conditional, Validated, or Certified.

An alternate technical description of the KPI Vault emphasizing its role as the immutable store of certified calculation outputs and evidence.

The highest MGE certification state, assigned only when the required Trust Gates pass and the Trust Score reaches at least 85.

A governed metric or record whose definition, value, evidence, lineage, version, and certification status are preserved together.

The architectural barrier between certification and downstream intelligence. Only certified outputs may cross it.

The locked record of the inputs, formula version, rule results, Trust Score, and certified value produced by one calculation.

Data that has passed the applicable MGE rules and Trust Gates and carries a verifiable certification record.

The horizontal category for certified operational intelligence across industries. It describes the broader application of FohBoh.ai beyond restaurants.

Operational data and metrics that have passed deterministic governance, reconciliation, scoring, and sealing requirements and can be independently verified.

The governed flow of certified outputs from MGE to authorized systems, workflows, and AI applications.

The output of MGE. Not raw data and not an estimate — a specific figure (certified revenue, certified food cost) with the evidence chain attached.

The restaurant-specific application of Certified Intelligence delivered through FohBoh.ai products.

The governed record of differences between the Truth Source and Claim Source after deterministic reconciliation.

The documented history showing where evidence originated, who or what handled it, and whether it remained intact.

The evidence showing what was actually charged, paid, reported, or recorded by the counterparty or operating system.

A loss-prevention system that converts suspected leakage into rule-tested, quantified, and traceable findings.

A single governed record created by reconciling certified findings across multiple Sentry modules.

A structured representation of business entities, relationships, decisions, exceptions, and approvals that preserves operational context for systems of action.

The corruption of AI reasoning when inconsistent metrics, missing semantics, or unverified claims enter the context used for decisions.

Event-driven certification that validates data as it moves through the enterprise rather than only during periodic audits.

The certified AI operations intelligence platform. Every answer is grounded in a Vault-sealed KPI definition and gated by a Trust Score.

The retrieval approach used by Cortex to keep grounding honest — verifying that retrieved context actually supports an answer before it is produced.

A digital fingerprint—typically created using SHA-256—that reveals whether a file, record, or certification artifact has changed.

The condition in which organizations calculate, interpret, and report the same metric differently across systems and stakeholders.

A machine-readable agreement defining required fields, formats, meanings, quality standards, and responsibilities between data producers and consumers.

FohBoh.ai's rules-based system for producing consistent certification narratives from verified facts — without generative AI, speculation, or motive attribution. The current version is DCLS v2.0.

The current version of the Deterministic Certification Language System used to generate controlled, reproducible certification language.

The recorded facts, rules, approvals, exceptions, and reasoning context associated with an operational action.

Which figure a percentage is calculated against — for instance, whether food cost uses gross or net sales. Most disputes about a number are really disputes about its denominator.

Certification produced by fixed rules, exact calculations, controlled definitions, and reproducible logic—not probabilistic AI.

The protected MGE certification path in which fixed rules, exact arithmetic, and versioned controls produce reproducible results without AI inference.

A concise architectural description of MGE: the engine establishes operational truth through fixed definitions, evidence, exact calculations, and reproducible rules.

A reusable package of canonical data definitions, KPIs, rules, and controls configured for a particular operating domain.

The MQ6 control that identifies changes in data patterns, configurations, definitions, or source behavior that may invalidate prior assumptions.

A third-party delivery platform. Commission terms, promotional fees, and adjustments are frequent sources of unreconciled loss.

MGE’s two-cadence processing model: Loop A evaluates current events in real or near-real time; Loop B evaluates patterns across certified history.

FohBoh.ai’s separation of truth certification from operational action: the Trust Loop establishes what can be relied upon; the Action Loop determines what to do with it.

Total card processing cost divided by card volume — the true rate paid, as opposed to the quoted rate.

Data strengthened by definitions, source records, lineage, controls, and proof sufficient for independent verification.

The portable collection of reports and supporting files required for independent review, dispute resolution, or audit.

The structured index connecting every file in an evidence package to its source, hash, rule references, and chain-of-custody record.

The controlled index of every source document, data set, exhibit, hash, and reference used in a certification run.

A condition that fails a rule, exceeds a threshold, lacks required evidence, or requires review before certification can continue.

Sentry’s role in an automated ecosystem: it continuously tests financial transactions and findings through deterministic rules before the books close or automated action proceeds.

Certified intelligence infrastructure that governs operational truth before systems of record, reporting platforms, and AI applications use it.

The versioning principle that a changed KPI, formula, or rule creates a new governed version while preserving the certified history of the prior version.

The risk that fluent AI turns unreliable input into an answer users accept as authoritative.

The principle that no autonomous system should act on operational data until its meaning, integrity, and permitted use have been governed.

A non-negotiable control that halts certification or downstream action when a required condition fails, regardless of other scores.

The inventory of cryptographic fingerprints for every file and artifact contained in an evidence package.

Running MGE as an API-only certification layer behind another vendor's interface, with no re-platforming required by the operator.

MGE delivered through APIs and embedded inside a partner platform without the FohBoh.ai user interface.

A record designed to be append-only so prior events cannot be silently altered or removed.

The durable dependency created as certified records, rule histories, evidence chains, integrations, and IUM accumulate over time.

The distinction that FohBoh.ai creates embedded trust records, definitions, evidence, and governance dependencies—not merely replaceable dashboards or workflow software.

The controlled receipt of source data and documents into the certification environment.

The verified operational facts supplied to analytics or AI before interpretation begins.

A transaction settling at a higher interchange tier than expected, usually from missing data or late settlement. A common and rarely audited source of leakage.

The body of governed operational intelligence an enterprise controls, measures, and can safely use. AUM manages financial capital; IUM manages operational intelligence.

The cryptographically verifiable history of certified intelligence as it progresses from governed data to actionable findings and evidentiary outputs.

The progression through which raw operational data becomes governed, certified, actionable, and economically valuable intelligence.

A defined business measure whose formula, inputs, scope, timing, and governance requirements are explicitly controlled.

Layer 1 of the definition model: the 50 canonical FohBoh-certified KPI definitions, each with formula, denominator, unit, and data dependencies.

Layer 2: an organization's own customizations. A changed definition enters as Pending and is not cited until formally committed.

Layer 3: the append-only, SHA-256 sealed store of committed definitions. Each commit carries the hash of the one before it, so any past answer can be reproduced against the definition in force at the time.

One controlled execution of approved KPI definitions and rules against a defined evidence set.

The cryptographic lock proving that a KPI definition, certification result, and supporting evidence have not changed since certification.

Step 2 of the pipeline: where data is SHA-256 sealed immediately on receipt, establishing chain of custody before any transformation.

Revenue lost to fee errors, rate drift, misapplied charges, and unreconciled settlements. Industry sources suggest 3–7% of annual revenue is recoverable.

Built to hold up under challenge: sealed chain of custody, reproducible calculation, and a documented narrative. Used in preference to “court-admissible”, which asserts an outcome no vendor can guarantee.

The traceable path showing how data moved and was transformed from source through certification to final output.

The fast execution loop that applies deterministic rules to current data and produces immediate certifications, exceptions, or alerts.

The slower scheduled loop that examines locked certified history for trends, recurring error signatures, and cross-period patterns without overwriting prior results.

The degree to which a metric is consistently defined, correctly calculated, reconciled, current, traceable, and reproducible.

The deterministic certification engine at the centre of the platform. MGE ingests operational data, reconciles it across systems, applies governance rules, scores the result, and seals it. It contains no machine learning — the same inputs always produce the same outputs.

The governance layer that controls whether operational data may advance to reporting, automation, or AI action.

A 0–100 measure of evidence coverage across six dimensions, expressed in five bands: Unverified (0–39), Provisional (40–59), Conditional (60–79), Validated (80–84), and Certified (85–100).

The comparison of independent operational sources to establish whether they describe the same event, value, or obligation consistently.

The unique cryptographic fingerprint of a DCLS narrative used to verify that its language has not changed.

The cryptographic lock connecting a certification narrative to the exact facts, rules, scores, and evidence from which it was produced.

The controlled collection of pre-approved DCLS document structures for each output type and certification state.

The conversion of differently formatted source data into a consistent canonical structure without changing its meaning.

A metric or fact established from governed definitions, reconciled sources, deterministic rules, and traceable evidence.

The human-review layer where authorized users apply business judgment and acknowledge defined findings or submissions.

The enforcement function of a Trust Gate: management policy is converted into a hard machine control that permits, blocks, or routes action for review.

A governed history of decision traces that allows prior certified decisions and outcomes to inform future action without changing the certification core.

Combined cost of goods sold and total labour — the headline operating metric, and the one most often calculated inconsistently across systems.

The documented origin, ownership, and authority of a data element or piece of evidence.

Cross-system validation between authoritative sources — POS sales against DSP orders, POS sales against deposits, inventory against purchasing, payroll against schedules.

Cortex's policy threshold at 70. Below it, no model call is made at all — the system returns a refusal naming the missing evidence rather than an answer.

FohBoh.ai’s initial market-entry strategy: use certified findings to recover financial leakage, generate measurable ROI, and prove MGE’s broader value.

A versioned collection of deterministic tests and calculations applied to a particular domain or use case.

The exact edition of a rule used in a certification run, preserved so the result can be reproduced later.

The immutable, mathematically sealed record for one certified transaction—the atomic unit of transaction-level governed intelligence.

The governed meaning and context that connect operational entities, metrics, and relationships so every system interprets them consistently.

The loss-prevention and recovery product. Modules M01 (merchant fee recovery) and M02 (delivery fee recovery) are live.

A domain-specific loss-prevention application—such as merchant fees, delivery fees, royalties, supply chain, labor, or inventory—powered by the shared MGE certification core.

The architectural rule that AI may interpret certified information but cannot modify the deterministic certification path.

The hashing standard used to establish chain of custody. Data is sealed on receipt in the Landing Zone, and again at narrative generation.

A calculated reference ledger representing what should have occurred under certified contracts, policies, and rules.

Step 7: where the certification narrative is generated and sealed.

The initial cryptographic fingerprint applied to source evidence at ingestion to establish integrity and chain of custody.

Verification of a transaction or metric against three or more independent sources when the certification policy requires it.

MGE’s governed reference model for the meaning and relationships of operational data. It is versioned and controlled to preserve reproducibility.

The market analogy for MGE as infrastructure governing money leaving a business. Stripe optimized getting paid; FohBoh.ai optimizes keeping what was earned.

A platform or agent that initiates operational decisions or workflows, subject to MGE certification and Trust Gate controls.

The accounting or ERP platform that stores transactions. MGE certifies data before it enters the system of record, so neither the source nor the destination can influence the outcome.

The deterministic logic that chooses the approved DCLS template for a specific output and certification state.

The reconciliation of an expected obligation, proof of receipt or performance, and the final invoice or settlement before approval, payment, or recovery action.

The DCLS component that retrieves certified values from the KPI Vault and inserts them into an approved narrative template.

The governed catalog connecting every DCLS placeholder to an approved certified value and source in the KPI Vault.

An authoritative evidence source used to resolve competing claims or complete reconciliation, such as a verified bank deposit record.

The principle that incoming data is treated as a claim, not a fact, until it passes certification controls.

A hard control that prevents data or metrics from advancing unless defined evidence, rule, and score requirements are satisfied. MGE operates 11 Trust Gates, TG01–TG11; earlier references to eight gates are superseded.

The coordinated set of 11 MGE controls—TG01 through TG11—that determines the Trust Score, certification state, and CAAR eligibility.

Infrastructure that determines whether data is sufficiently reliable to enter a system of record, report, workflow, or AI process.

The deterministic loop that governs, tests, scores, and certifies operational reality.

A deterministic 0–100 measure of the integrity of a certification process and its evidence—not an AI confidence score.

The evidence showing what should have occurred under the governing contract, policy, rate, or approved business rule.

Data that has not yet passed certification or did not satisfy the required controls. It remains a claim, not established truth.

A measurable difference between an expected, contracted, recorded, or calculated value and the value actually observed.

The cryptographic signature binding a KPI definition to a specific version, allowing any citing answer to be verified later.

The machine-readable information needed to validate a certification artifact, including timestamps, versions, hashes, identifiers, and status.