Pricing

Agent Tiering / Inherent Risk Classification, full quality.

• A plain-English walkthrough plus a decision table you can actually run
• A worked example with real numbers
• Decides how much governance an agent needs, without ever granting it more autonomy or access
• A one-page summary you can hand to a non-technical stakeholder

92 governance patterns, with operating-control templates for Microsoft Agent 365.

• Five Agent 365 operating-control templates, a Conditional Access baseline and a step-by-step setup guide
• Every pack this library has, sold individually elsewhere on this page, in one download
• Includes the free ontology: the dependency graph and shared registry schema tying all 92 patterns together
• 30-day money-back guarantee: if any pack doesn't meet the standard shown in the free sample, tell us why and get a full refund
• No live platform to operate, no subscription, no seat count, unlike the enterprise AI governance tools this replaces

One workflow choice decides whether your CV-screening agent is R3 Critical or one step below, and whether it may act or only propose. Six governance patterns catch which one you're actually running.

• Whether it lands at R3 Critical or R2 High depends on who actually sends the rejection, this pack shows you which
• Registers the agent as governance-grade, with a real revocation path
• Catches a rejection logged under a shared service account instead of the agent's own identity
• A full worked example tying all six patterns to one CV-screening scenario
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For agents that can write someone up or flag them for termination without a human reviewing first, the kind of autonomous action AI laws worldwide are starting to put guardrails around. Six governance patterns catch it.

• Confirms whether autonomous write-ups put the agent at R3 Critical, and keeps it proposing until autonomy is explicitly authorised
• Catches a monitoring capability added after launch that was never re-authorised against what the agent can actually do
• Catches a write-up filed under a shared login instead of the agent's own identity, so the record holds up if challenged
• A full worked example tying all six patterns to one attendance-monitoring scenario
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

AI-text detectors misflag non-native English writers at a documented 61.3% false-positive rate. Six governance patterns catch an exam-scoring agent before it fails someone wrongly.

• Confirms whether autonomous integrity fails put the agent at R3 Critical, and keeps it proposing until autonomy is explicitly authorised
• Catches a scoring capability added after launch that was never re-authorised against what the agent can actually flag
• Catches an integrity fail filed under a shared login instead of the agent's own identity, so the record holds up if a candidate appeals
• A full worked example tying all six patterns to one exam-scoring scenario, grounded in documented detector bias
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

A written kill-switch procedure isn't evidence it works. A real test found a 47-minute stop time against a 30-minute target, a gap a paper review would have missed.

• Surfaces every agent running today, registered or not
• Separates a pilot that impressed a demo room from one actually fit for production
• Proves your kill-switch works, and how fast
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

A problem that makes your agent R3 Critical is expensive to catch after months of build work. Three governance patterns catch it before anyone writes code.

• Estimates the likely risk tier from a one-paragraph spec, before build time is spent
• Walks a proposed use case to documented harms and a mitigation for each, before code exists to argue over
• Locks in the approved scope at kickoff and catches a later unreviewed feature expansion
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

A large and growing share of the AI agents in your company weren't built by anyone internal, they arrived embedded in a SaaS tool you already pay for. The same vendor contract can hide features at very different risk tiers, four patterns catch it.

• Catches an AI capability a vendor rolled into an already-contracted tool, and whether you can actually turn it off
• Gives each discovered capability its own risk tier, from what it actually touches and can commit you to
• Moves an ad hoc legal question into a repeatable gate at signing or renewal, then keeps watching for a vendor expansion nobody asked about
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

Every agent needs to be watched, not just launched correctly. Nine governance patterns cover an agent for as long as it stays in production, not just at launch.

• A shared observability floor plus per-agent baselines for cost, latency, error rate, and escalation rate
• Catches an underlying model or prompt version change as soon as it happens
• Catches an agent that keeps responding while it has quietly stopped working correctly
• A full worked example tying all nine patterns to one claims-triage scenario
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

Rent or own, which model handles a request, and where any of it is allowed to run. Four governance patterns for the infrastructure decisions behind an agentic AI workload.

• Decides whether a sustained workload should still be rented or has earned owned infrastructure
• Checks whether an infrastructure choice was made deliberately or defaulted to whatever was easiest to procure
• Checks whether rented-model routing is defensible, resilient, and traceable
• A full worked example tying all four patterns to one bank's infrastructure decisions
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For agents that delegate work to other agents. Four governance patterns for the architecture, not any single agent: pairing approval, standing-composition governance, delegation depth limits, and cross-system audit log reconciliation.

• Reviews a specific agent-to-agent pairing rather than assuming two individually-approved agents are safe together
• Governs an ongoing composition of agents with shared definitions, change-triggered re-review, and a named owner
• Checks that a delegation chain has a tracked, deliberately-set depth limit
• Confirms cross-system audit logs actually agree before a decision trail is trusted
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For the identity layer underneath every agent, any industry, any lifecycle stage, any architecture. Six governance patterns: sponsor assignment, orphaned credential detection, standing access review, static credential elimination, runtime permission escalation, and single-source-of-truth ownership.

• Confirms a named, current, meaningfully accountable sponsor exists for an agent identity
• A recurring audit that catches a credential still holding access after its agent, sponsor, or use has lapsed
• Confirms granted access is actually re-justified on a defined schedule, not just granted once
• Confirms one authoritative record of an identity's ownership exists and is actually used
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For the data layer underneath every agent's output, any industry, any lifecycle stage. Five governance patterns: shared definitions, lineage, vocabulary, quality, and access classification.

• Confirms an authoritative, owned, shared, and versioned definition exists before an agent consumes a data entity
• Confirms a data feed's origin and full transformation path is tracked and reviewable, not assumed
• Confirms a defined minimum quality standard exists, is currently met, and is checked on an ongoing basis
• Confirms an agent's access to a data table is approved for its stated purpose, not granted by its risk tier
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For the moment an agent changes, any industry, any lifecycle stage. Five governance patterns: model drift re-approval, impact classification, approval workflow, emergency handling, and post-change validation.

• Triggers re-approval when a vendor changes the foundation model behind an already-approved agent
• Classifies a proposed change's impact against the agent's operating mode and risk tier before approval
• Routes a classified change through the workflow its risk tier requires, with the right approver authority
• Confirms a deployed change actually worked, checked against criteria defined in advance
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For how an agent's operating mode is set, changed, and defended over time, any industry, any lifecycle stage. Six governance patterns: escalation trigger definition, override/intervention mechanism, autonomy regression trigger, approval authority matrix, time-bound autonomy trials, and autonomy level audit trail.

• Turns "escalate when uncertain" into a checkable criterion with a confirmed live destination
• Confirms a real, tested override capability exists, not an assumed safety net
• Defines exactly when an automated downgrade trigger fires, not just that one exists
• Confirms every oversight-level transition is captured in an immutable, attributed, retained log
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For the operational mechanics of running an agent under ongoing incident-response and regulatory obligations, any industry, any lifecycle stage. Six governance patterns: audit trail requirements, incident response, post-incident review, compliance attestation cadence, regulatory change monitoring, and a whistleblower/escalation channel for AI concerns.

• Confirms decision inputs are captured, logs are tamper-evident, and retention meets a defined minimum
• Confirms a documented plan, an active named responder, and a real containment action actually exist
• Confirms a post-incident review found a real root cause and assigned a corrective action to a named owner
• Confirms a protected, independently-reviewed channel exists for employees to raise a concern about an agent
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For the specific moment a deployment moves from built to launched, any industry, any lifecycle stage. Six governance patterns: security review gate, data access scope definition, rollback plan requirement, environment segregation, user acceptance criteria definition, and a capstone go-live readiness checklist.

• Confirms a formal, independent security review was conducted with high-severity findings resolved
• Confirms this deployment's data access is documented, minimal, and reviewed by the data owner
• Confirms a rollback plan exists with defined trigger criteria and a named owner
• Capstone checklist confirms all prior gates hold, an approver is named, and rollback is armed at launch
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

For the governed transitions an agent moves through across its whole operating life, any industry, any lifecycle stage. Seven governance patterns: ownership assignment, naming and identity convention, sponsor/business-owner sign-off gate, promotion gate, demotion/downgrade trigger, agent deprecation process, and agent retirement and decommissioning.

• Confirms a real owner was nominated, accepted the role, and has a documented scope of responsibility
• Confirms sign-off was requested, given by the accountable sponsor, and obtained before the transition
• Confirms operating-mode promotions are objectively verified and independently approved, not self-assessed
• Confirms retirement means access genuinely revoked, retention obligations met, and owner confirmation obtained
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

73% of organisations exceeded their AI cost projections in 2026. Five governance patterns for controlling runaway agentic spend, any industry, any lifecycle stage: fleet budget ceilings, single-action pre-approval, cost ownership, forecast reconciliation, and per-agent cost baselining.

• Confirms a fleet-level spend ceiling is actually enforced, not just an alert nobody reads in time
• Confirms a specific expensive action gets sign-off before it runs, not after the money is spent
• Confirms a named, active individual is accountable for an agent's spend, and can actually see it
• A full worked example tying all five patterns to one runaway-spend recovery scenario
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

94% of organisations name agent sprawl their top AI governance concern. Five governance patterns for finding what's actually running, any industry, any lifecycle stage: shadow agent discovery, inventory reconciliation, no-code agent registration, duplicate consolidation, and decommissioned-agent cleanup.

• Confirms a discovery scan actually runs and an unregistered agent it finds gets contained
• Confirms registry count is reconciled against reality on a schedule, and a gap gets closed
• Confirms a no-code agent completes the same intake as an engineered agent before going live
• A full worked example tying all five patterns to one agent fleet's sprawl story
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

OWASP's #1 LLM risk for the third year running, and the detection problem is genuinely unsolved. Five governance patterns for the architectural discipline that actually narrows the blast radius: trust boundary declaration, tool scope minimization, high-impact action confirmation, decision provenance logging, and baseline behavior monitoring.

• Confirms untrusted content is explicitly classified and handled as data, never as instructions
• Confirms an agent's tool access is scoped to what its actual task requires
• Confirms a high-impact action traced to untrusted content is held for human confirmation
• A full worked example tying all five patterns to one support-triage agent's incident story
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund

Incidents per pull request are rising alongside AI-assisted PR volume. Seven governance patterns tracing a real CI/CD pipeline: commit test coverage, merge sign-off, dependency provenance, pipeline credential scoping, security testing, production promotion authorization, and rollback boundaries.

• Confirms a coverage threshold exists and actually blocks merge for agent-authored changes
• Confirms a named human approver signs off, with review depth scaled to risk
• Confirms production promotion requires explicit, pipeline-enforced human authorization
• A full worked example tying all seven patterns to one release moving through the pipeline
• 30-day money-back guarantee: if this pack doesn't meet the standard shown in the free sample, tell us why and get a full refund