Simple Council
Review AI governance decisions with an evidence trail.
For CISOs, Heads of AI Governance, Compliance Officers, and Internal Audit leads at regulated mid-market organizations adopting Microsoft 365 E7 + Agent 365.
An agent inventory is a starting point for governance. Teams also need reviewed decisions, control mappings, and evidence they can inspect. Simple Council provides a policy engine and a hash-chained evidence ledger for actions routed through its own surfaces. Live Agent 365 interception and Microsoft Purview integration are not available, and no customer tenant has been validated yet. Framework blueprints are starting points for review, not a certification or a guarantee of compliance.
How to get Simple Council
Controlled early accessLive and in production, with new workspaces opened one at a time. You apply, a person reads it, and we set you up.
Where this stands
- Availability
- Controlled early access
What it costs
No published price yet. Pricing is set before this product opens.
Routes that are not open
- Direct trial with Simple Intelligence Not open yet
- Direct subscription from Simple Intelligence Not open yet
- Microsoft Marketplace Not published yet
- Sponsored subscription for member companies Not open yet
Simple Council on the same scale as everything else here
Every Simple Intelligence product is placed on one five-stage scale, so a stage means the same thing on this page as it does on any other. Nothing here is a schedule.
-
Done
Planned
Scoped and specified. Not being built yet.
-
Done
Building
Under active development.
-
Done
Validation
Built, and being tested against real work before anyone outside is let in.
-
Now
Early access
Open to a reviewed group, with onboarding run by a person.
-
Not started
Available
Open to anyone who qualifies, without waiting for a review.
This is where the product sits, not what you can do with it today. What you can do is the button at the top of this page, and it is worked out separately from evidence about the product rather than from this scale.
Your CISO has a new RFP question this quarter.
It reads something like: "How do you govern, audit, and enforce policy on every AI agent operating inside your Microsoft tenant, across every model provider, with evidence your regulator will accept?" It used to be a question your competitors asked your customers. Now your customers are asking it of you.
Agent 365 answers who is running. Your auditor asks what they did.
Agent 365 gave you one surface in the M365 admin center showing every agent operating across Copilot, Foundry, and the non-Microsoft providers your teams brought in, with telemetry tied into Defender and Purview. That is the hard half of the problem solved, and Council starts from it rather than replacing it. What an audit needs next is different in kind: activity mapped to the specific control families you file against, retained continuously, and provable after the fact.
Simple Council is the layer that turns registry into evidence.
It sits above Agent 365 as the compliance layer: an hourly evidence sweep that snapshots each agent against each adopted control, controls mapped against the six frameworks it ships with, a hash-chained ledger that makes tampering detectable, and a policy engine with four dispositions: block, route to a human, escalate to the agent owner, or allow with a richer audit record. Each disposition writes a row naming the rule that fired, in the same transaction as the decision itself. Phase 1 and 2 of that design are built and reviewed, and the production deployment is live. What the engine does not yet have is a connector that feeds it live agent calls, so it governs the actions that reach it rather than every action an agent takes. Customer-tenant validation and early-access onboarding are still pending, so what follows describes the product as built and deployed rather than one proven against a customer's own tenant.
Where this product actually is today.
Not yet
- Nothing intercepts live agent calls yet. The policy engine governs the actions that reach it through the product's own surfaces, and the connector that would carry an intercepted Agent 365 or Entra Agent ID call into it is designed and not built. Until it ships, treat this as governance over what you route through it, not a net under everything your agents do.
- There is no Microsoft Purview integration. Until today this site said evidence collection was bound to your Purview tenant, and that was wrong: no such connection exists in the product.
- No customer tenant has been validated yet. The product is deployed and running, and early-access organisations are onboarded one at a time, which is the stage this is at rather than a queue-management preference.
- The auditor pack is JSON, HTML, and ZIP. There is no signed PDF attestation.
- The Azure Marketplace offer is planned rather than published, so this cannot be bought through a Microsoft agreement yet.
- MISA and Microsoft Security Store listing is a target, not a status. Neither has been granted.
- The six blueprints are our control mappings, reviewed by us. They are not a certification, an audit opinion, or a regulator's endorsement, and adopting one does not make an organisation compliant with anything.
Verified against the product on 2026-08-20.
The four steps, and how far each one is built.
Planned Agent 365 connection
The design is a single read-only Entra grant covering the Agent 365 registry, agent telemetry from Defender, and audit signals from Purview, with the evidence Council produces written back into your own tenant. None of those three reads exists in the product yet: the registry sync is a stub and the Defender and Purview pulls are on the roadmap. What is deployed and running today is the service itself and its Entra sign-in, so the agents Council governs are the ones you record in it. It has not yet been validated against a live customer tenant.
Pick blueprints
Council ships six framework blueprints: ISO 42001, EU AI Act, NIST AI RMF, HIPAA, FFIEC, and CMMC L2. You pick the ones your firm files against, and Council maps each agent in your registry to the relevant controls. The control set in each blueprint is a curated starter set covering the most-cited controls, not a reproduction of the full standard, and you extend it with your own.
Author policy
Council ships default policies per blueprint. You override per agent, per tenant, per risk threshold. The policy decision tree lives in plain language, not in a code editor.
Enforce + evidence
What reaches the policy engine is evaluated against your rules and comes back as one of four dispositions: block, route to a reviewer, escalate to the agent owner, or allow with audit. Overlapping rules resolve strictest-wins, and the PolicyDecision row and its hash-chained Evidence row are written in one transaction, so a decision cannot exist without its record. Your auditor receives an export that recomputes the chain. Two limits are worth stating plainly. The disposition is a decision the calling system has to honour, because the dispatch that would chase a named reviewer or hold a call until an owner acknowledges it is designed and not built. And what reaches the engine today is what you route through Council, because the connector that would carry an intercepted Agent 365 or Entra Agent ID call into it is designed and not built.
What the policy engine and the evidence ledger do.
Compliance plane above Agent 365
Six blueprints covering the regulated industries SIG sells into: ISO 42001 (AI management), EU AI Act, NIST AI RMF, HIPAA (healthcare), FFIEC (financial services), and CMMC L2 (defense). Each blueprint maps onto the agents recorded in Council and shows which of them need which controls. Populating that list from the Agent 365 registry is the planned path and the sync behind it is not built, so today the agents are the ones you register.
Continuous evidence collection
Council writes evidence as decisions happen rather than in the week before an audit deadline, into an append-only hash-chained ledger, so when the auditor calls the records are already there. Identity context comes from Entra, which is the connection that is live. Enriching those records with agent telemetry from Defender, audit signals from Purview, and registry state from Agent 365 is the design and those three reads are planned rather than available.
Policy enforcement on the actions routed through Council
Four dispositions, decided per call and recorded. Block, meaning the call must not proceed. Route, naming a reviewer as the human seam. Escalate, marking it for the agent owner. Allow-with-audit, letting it proceed with a richer trail. Every disposition is a PolicyDecision row carrying the rule that fired, the agent, the tenant, the user, and the input fingerprint. Council decides and records; the system that made the call is what carries the decision out, and the notification and acknowledgement path that would sit between the two is designed rather than built.
Hash-chained Evidence ledger
Every Evidence record carries a hash that includes the previous record's hash. Tampering is detectable by recomputing the chain. Your auditor receives an export with the chain verifier; they can confirm no record was inserted, modified, or removed after the fact. The hash chain is the difference between "audit trail" and "tamper-evident attestation."
Where your data lives, and who can sign in.
Where your data lives
Council is designed to run on Azure Container Apps in your customer-tenant region. The control plane holds agent IDs, policy decisions, and evidence hashes; the underlying agent call payload stays where it was. For zero-data-retention requirements, Council pairs with SimpleForge to enforce on-device inference.
Where the audit trail lives
Evidence rows are written to Council's own append-only ledger in the deployment serving your tenant. Writing them onward into your Purview tenant is the design and there is no Purview connection in the product yet. The auditor pack is a JSON envelope, a readable HTML report, and a ZIP holding both. It is not a signed attestation. The chain verifier ships with the export and reports every break it finds with the agent, the control, and the timestamp, so a third party can recompute the chain rather than take a clean bill on trust.
Identity model
Entra ID OIDC throughout. Council enforces RBAC against the same groups your CISO already manages. Multi-tenant per the entity-level ruleset, and one deployment is meant to cover a firm's whole agent estate whatever surface registered each agent. That breadth depends on the registry sync, which is not built, so the coverage today is the agents recorded in Council.
Council on Council
Our intent is to be the first tenant. Simple Intelligence runs its own agent footprint through the same Council build we would sell, and the blueprints go through our own workload before they go to a customer. That is the plan for launch, not a description of today: the build is deployed, and no tenant has been validated against it yet, ours included.
Microsoft-first, and honest about the rest.
A name here is a system this product is built to work with, not a promise that the connection is live in your tenant. Where a connector carries a status label, that label is the current state. Anything that needs its own vendor registration, its own consent, or its own approval says so, and nothing on this page can be taken as available until it is confirmed for your tenant.
Microsoft surface
Cross-vendor agents in scope of the planned registry sync
Partner factory integrations
SIG portfolio
Reading data from an external system is not a way to sign in. Customer identity stays with Microsoft Entra ID, and no data source on this list authenticates a person.
What backs the claims on this page.
How this compares to what you already evaluated.
| Capability | Simple Council | Vanta (AI module) | Drata (AI module) | OneTrust AI Governance | Manual GRC |
|---|---|---|---|---|---|
| Cross-vendor agent registry (via Agent 365) | Planned | Partial | Partial | Limited | No |
| Continuous evidence to Purview | Planned | No | No | Partial | No |
| Hash-chained tamper-evident audit | ✓ | No | No | No | No |
| ISO 42001 blueprint | ✓ | Partial | Partial | ✓ | Spreadsheet |
| EU AI Act blueprint | ✓ | Partial | No | ✓ | Spreadsheet |
| HIPAA blueprint | ✓ | ✓ | ✓ | ✓ | Spreadsheet |
| CMMC L2 blueprint | ✓ | Partial | No | Partial | Spreadsheet |
| Azure-native deployment | ✓ | No | No | No | No |
| White-label for Solutions Partners (via Atlas) | ✓ | No | No | No | No |
Comparisons reflect publicly available product documentation as of 2026-06-18. Vendors evolve; if a row is out of date, tell us at /contact and we'll update.
How an evaluation starts.
- How access starts
- A conversation first. We scope the evaluation with you and provision the tenant ourselves.
- Evaluation length
- Agreed when the evaluation is scoped, up to 45 days for a standard evaluation. A longer guided pilot is approved case by case.
- Payment details
- None collected to begin an evaluation.
How ICM informs Simple Council
Simple Council is the product where the evidence is the point. It sits in front of other agents, decides what they may do against a compliance blueprint, and keeps a record an auditor can test rather than trust.
Built, and verified against the product code
-
Context in
The caller, the input, the requested action
A decision record captures the full request context: who asked, what they sent, which model was involved, and what they wanted done.
Gated
-
Retained or discarded
Evidence is hashed into a chain
Each piece of evidence hashes its own content, its capture time, and the hash before it. Replacing a record points at what superseded it rather than overwriting it, so an auditor can test whether the chain was tampered with instead of taking the platform's word for it.
Gated
-
What the agent proposes
A policy mapped from a blueprint control
Rules can be suggested from a blueprint mapping, and when they are, the policy record says so. A suggested policy is a proposal about the customer's own governance, not a change to it.
Gated
-
Human review a person decides
A policy is off until somebody enables it
New policies are created disabled. The escalate enforcement mode is itself a human seam: it holds the calling agent until a person approves the action rather than letting it proceed and reporting afterwards.
Gated
-
Permitted effect
Four outcomes, and no fifth
A decision is allowed, blocked, routed to a different handler or a human queue, or escalated for approval. There is no mode that lets an action through without one of those four being recorded.
Gated
-
Evidence recorded
An append-only decision log
The decision table is the audit trail and is exempt from the ordinary provenance columns because it is the thing they would point at. Each row carries the reason next to the outcome. The general audit log records the actor type, what changed before and after, and why.
Gated
-
Failure and refusal
Refusal is a mode, not an error
Block refuses and returns an error to the caller. Route sends the work somewhere a person can pick it up. Allow with audit lets a low-risk action through while raising a high-priority event about it. Each is a designed outcome with a record, rather than an exception somebody has to notice in a log.
Gated
These labels describe each step of this workflow on its own. They are not the availability of the product, which is the status shown at the top of this page.
This describes how the workflow is built. It is not a statement about availability, which is the status shown at the top of this page. ICM informs how context, review and evidence are designed here. It does not replace this product's database, its tenant isolation, its Entra identity, its audit tables, or its authorisation checks.
Methodology note
ICM was developed by Jake Van Clief and David McDermott. Simple Intelligence Group did not originate the methodology.
Simple Intelligence Group applies relevant ICM principles to context, provenance, traceability, human review, and workflow design.
Participation does not imply certification, endorsement, partnership, or an official ICM Architect designation.
The questions buyers ask in the first call.
We already use Microsoft Agent 365. Do we still need Simple Council?
Yes if your regulator expects framework-mapped evidence (ISO 42001, HIPAA, FFIEC, etc.) or your CISO needs policy enforcement at the agent-call layer. Agent 365 is the registry. Council is the compliance plane above it.
How does Council handle agents from Bedrock or Vertex AI?
Agent 365 registers those in your Microsoft tenant, and Council is designed to govern them through the same policy engine that governs Copilot Studio agents, so there is no separate control plane per provider. The sync that would pull those registrations into Council is not built yet, so today you register the agents you want governed.
Where does the evidence go?
Into Council's own append-only ledger, in the deployment that serves your tenant. The ledger is hash-chained and the export ships with the verifier, so your auditor can recompute the chain rather than trust a summary. Writing evidence onward into your own Purview tenant is the design and that connection is not built, so if Purview is a hard requirement for you, treat it as planned.
What is the implementation timeline?
We do not have an honest number yet, because no customer has been onboarded onto Council. What we can tell you is what the work consists of: granting the read connection, choosing the frameworks you file against, reviewing the default policies, and running an export past your auditor. Early access customers help us find out what that actually takes.
Is Council certified?
No. Certifications are targets, not attestations we hold today, and we would rather say so here than have you find out during a security review. Council targets MISA membership and a Microsoft Security Store listing. Council deployments run on Azure commercial cloud and inherit the Azure platform certifications that apply there.
How does pricing work for partners?
Council is designed to be white-labeled through Simple Atlas, our partner platform. The commercial terms for partners are being finalized and the referral program is not open yet. See /partners/ to register interest.
We have an existing GRC platform (OneTrust, Vanta, Drata, Hyperproof). Does Council replace it?
Council does not replace your enterprise GRC platform. Council is the AI-agent governance layer specifically. Most customers run Council alongside their existing GRC for everything-not-AI, with Council producing the AI-specific evidence packs the existing GRC cannot.
The rest of the portfolio.
How to start
-
Controlled early access Open
Live and in production, with new workspaces opened one at a time. You apply, a person reads it, and we set you up.
Apply for early access -
Direct trial with Simple Intelligence Not open yet
This product has no self-serve trial yet. Ask us and a person will set access up.
-
Direct subscription from Simple Intelligence Not open yet
There is no direct checkout for this product yet. Where a trial exists it converts to a paid subscription inside the product.
-
Microsoft Marketplace Not published yet
There is no transactable listing for this product yet, so it cannot be bought against a Microsoft agreement today.
-
Sponsored subscription for member companies Not open yet
The member sponsorship program is not open for this product yet.
Questions about evaluation length, extensions, or sponsored access are answered on the get started page.
Council is running. We are not onboarding customers yet.
The evidence engine, the hash-chained ledger, and the auditor export are built and tested, and the production deployment is live. Early-access onboarding has not started. If agent governance is on your roadmap for this year, we would rather show you the work now and hear what your auditor will ask for.