Glossary
Every Bespoke Mentis term explained in plain English. If the product pages feel technical, start here. Each term includes both a plain-language version and the technical definition.
Platform Terms
MIOS
Mentis Intelligence Operating System
Plain language
An AI operating system that replaces your entire stack of business tools — CRM, analytics, newsletter, SEO, compliance, governance — with one governed platform. Custom-built for your organization.
Technical definition
A full-stack business intelligence OS with 20+ modules, 7 AI agents, 6 authority systems, and a cross-agent intelligence bus. Deployed on your infrastructure and configured to your workflows.
Mentis Console
Mentis Console (AI Engineering Platform)
Plain language
The platform that Bespoke Mentis's AI products are built on. Think of it as the foundation and governance layer that makes all AI systems auditable and safe.
Technical definition
A 15-core constitutional AI engineering platform. Each core is a specialist domain. All cores operate under a 5-layer governance architecture with SHA-256 evidence chains and mandatory human oversight gates.
Foresight
Foresight (Life Sciences Intelligence Platform)
Plain language
An AI intelligence platform for pharmaceutical and biotech organizations that produces evidence-backed competitive, clinical, and regulatory intelligence with a full audit trail.
Technical definition
A 5-core governed intelligence operating system for life sciences. Delivers clinical data synthesis, competitive landscape intelligence, payer analysis, and regulatory simulation with deterministic scoring and SHA-256 verified outputs.
Strategic Memory
Strategic Memory Agent (pgvector)
Plain language
The system's long-term memory. Every meeting outcome, deal decision, and strategic signal is stored and becomes searchable. The more you use it, the smarter it gets.
Technical definition
A pgvector-powered long-term intelligence layer. Every decision, meeting artifact, deal pattern, and business signal is embedded, indexed, and made queryable through natural language. Compounds over time.
Governance Terms
Constitutional Governance
Constitutional AI Governance (AC11)
Plain language
Rules baked into the system's code, not written in a policy document. The AI cannot do things it is not allowed to do — it is structurally blocked, not just warned.
Technical definition
A runtime enforcement architecture where every AI operation is evaluated against a compiled governance law set before execution. Violations trigger an immediate HALT — not a warning.
SHA-256 Audit Trail
SHA-256 Hash-Chained Evidence Ledger
Plain language
Every decision the system makes gets a unique, permanent fingerprint. You can prove, at any point in the future, that a specific decision was made, when it was made, and what led to it.
Technical definition
An append-only cryptographic ledger where each entry is SHA-256 hashed and linked to its predecessor, creating a tamper-evident chain. Any modification to a past entry is structurally detectable.
Fail-Closed Safety
Fail-Closed Execution Architecture
Plain language
When the system is uncertain, it stops and waits. It never invents an answer to fill a gap. If you ask it something outside its knowledge or authority, it says so.
Technical definition
On any governance boundary violation, undefined context, or confidence threshold failure, the system returns a null result and HALT signal — never a fabricated plausible response.
Bounded Retrieval
Bounded Retrieval Architecture
Plain language
The AI can only use information it has been explicitly given access to. It cannot go looking for answers outside its defined knowledge boundary.
Technical definition
RAG context is scoped to authorized, verified sources at query time. The model cannot access, infer, or hallucinate beyond the bounded context window — enforced by the governance layer.
Deterministic Execution
Deterministic Output Execution
Plain language
The same question always produces the same answer. You can run it again tomorrow and get exactly the same result — it doesn't change based on mood or randomness.
Technical definition
Temperature-controlled, reproducible model invocations with fixed context windows, deterministic scoring functions, and bit-perfect replay capability for audit verification.
MIOS Intelligence Agents
PIA
Prompt Intelligence Authority (MIOS Module 14)
Plain language
The system that checks every AI instruction before it runs. Think of it as a legal review for every question you ask the AI.
Technical definition
A constitutional prompt governance system that version-controls, scores against 87 governance laws, adversarially tests, and SHA-256 hash-chains every prompt before production deployment.
CIA
Compliance Intelligence Authority (MIOS Module 15)
Plain language
Automatic monitoring of every major regulation that could affect your business. It alerts you to compliance gaps before they become problems.
Technical definition
An AI compliance engine monitoring 24 live regulatory data sources (EU AI Act, GDPR, HIPAA, SOC 2, CCPA, PCI DSS +18) with automated run types and SHA-256 evidence-backed reports.
GIA
Governance Intelligence Authority (MIOS Module 16)
Plain language
A live score for how well your organization is actually following its own rules — not based on what your policies say, but on what the data shows.
Technical definition
A live Operational Trust Score (OTS) engine that scores 8 governance dimensions using machine-generated evidence. Includes a Ghost Drift Detector that flags policies existing only on paper.
CIT
Central Intelligence Terminal (MIOS Agent)
Plain language
A single command interface that lets you ask questions about any part of your business in natural language — and get a governed, cited answer.
Technical definition
A cross-system AI command interface with 12 intelligence layers. Synthesizes analytics, CRM, meetings, market data, and all 7 agents into a single governed response.
Architecture Terms
Intelligence Bus
Cross-Agent Intelligence Bus
Plain language
The system that automatically passes information between AI agents without you having to copy anything. When one agent learns something relevant, it immediately tells the others.
Technical definition
A real-time event routing system where agents fire typed intelligence events (signal_score_threshold_hit, financial_alert, etc.) to downstream agents based on predefined triggers and constitutional rules.
Modular AI Architecture
Specialist Modular AI Architecture (Brain-Inspired)
Plain language
Instead of one AI that tries to do everything, multiple specialist AIs each handle what they are best at. Validated by MIT, Berkeley, and EPFL as the most accurate architecture for complex tasks.
Technical definition
A cortical column-inspired architecture where specialist AI cores, each with domain-specific inductive biases, are coordinated by a constitutional meta-orchestrator. Validated by MAP, BIMT, MiCRo, Alpha Framework, Numenta CCA, and Salatiello 2026.
Still unclear? Talk to Mentis — our AI assistant can answer any question about how our systems work.