AI capability maturity in Renly

AI Use Register Enterprise

Adds the organisation’s agents, active automations and configured connectors in Renly automatically (refreshed nightly and when an admin opens the register), and lets you add AI used outside Renly by hand. Scores inherent and residual risk across eight factors to a tier, names the accountable owner, tracks each use through a seven-stage lifecycle (intake, impact assessment, design, simulation, shadow, operating, retired), records the go or no-go decision at five gates with the date and the admin who recorded it, schedules reviews, and exports a disclosure statement for uses that need one.

Where it stops: Agents created inside a workspace and external (MCP) servers are added by hand, as is AI in other systems: Renly does not discover AI outside Renly. Sharing the register with a regulator or the DTA stays with you.

Evidence Passport All plans; mandatory evidence and organisation-wide export on Enterprise

Every response carries an evidence record: which model answered, which tools ran and who approved each write, the policy verdict, and hashes of the prompt and answer, sealed into the organisation’s own hash chain so tampering is detectable. The owner of a chat can export its evidence bundle on every plan. On Enterprise, marking a use ‘evidence mandatory’ refuses its agent’s write actions while signed evidence cannot be recorded, and admins can verify the chain and export recent evidence records.

Where it stops: Covers AI that runs in Renly. Capture is best effort unless evidence is marked mandatory, and the mandatory setting guards write actions, not reads. Export for a chosen period is available through the API.

Action rules and approvals All plans

Each of an agent’s actions has a rule, set per person: actions that send or change things are set to Ask first or Never, and other actions to Allowed or Never. Actions set to Never are hidden and refused. Sending, posting and publishing through Renly’s built-in tools always need a person. Approvals arrive as cards in chat, in an Approvals inbox or as signed email links; each is single-use, expires, is bound by a fingerprint to the consequential details the person reviewed (such as recipients, amounts, records or a command), and records who approved it. Automations can route approvals to a named approver.

Where it stops: Approval and override rates are recorded but not yet reported as a trend view. An admin can let an individual external (MCP) tool run without approval. Where a built-in action has no consequential detail to bind, the binding is logged rather than enforced.

Two-person approval Agents: all plans. AI Use Register setting: Enterprise

Mark an agent, or a use in the AI Use Register, as two-person: its approval-gated actions then need two different people, and both are recorded in the Evidence Passport.

Where it stops: Needs an organisation owner or admin other than the person who asked and the first approver.

Pause all agents (kill switch) All plans

An organisation admin can pause all agents in one click. While paused, automations, Tasks, triggers, hand-offs and sub-agents stop, and agents in chat, Microsoft Teams, Copilot Studio and connected clients can only reply: every tool call is refused, the pause is checked afresh before each action, and held approvals cannot run until agents are resumed. On Renly Desktop, a keyboard shortcut pauses computer control, and any paired device can be revoked remotely.

Where it stops: Halts AI running in Renly, not other systems. A command already sent to a desktop device finishes; the next one is refused. People acting for themselves, such as committing their own code, are not paused.

Rollback

Code sessions keep checkpoints of files an agent edits so they can be reverted; a chat can be rewound to an earlier message; agents keep versions that can be restored.

Where it stops: Changes an aborted run already made in other systems are listed for clean-up, not undone automatically.

Governed Runtime policy checks Enterprise

When switched on, an independent policy service checks each chat request before it reaches a model, and each automation step and agent hand-off before it runs. If the policy service cannot be reached, requests are blocked by default; admins can choose to warn or allow instead.

Where it stops: Off until an admin switches it on. Policies are written by the organisation.

Agent tests, promotion gate and shadow mode

Write test cases for an agent (expected terms, expected tools and a rubric) and run them against real data in a read-only sandbox with an AI judge. Mark an agent as production tier and, once it has active test cases, every change waits for a different admin’s approval after its tests pass; an override needs a written, audited reason. An admin can run a pending change in shadow mode: Renly replays a sample of real requests against it in the sandbox (up to 50 a day by default), compares its answers with the live ones and shows the agreement rate in the register.

Where it stops: Evaluates agents built in Renly. An admin can remove the production-tier mark and then edit directly.

Monitoring and export

Streams governed agent actions to an OpenTelemetry endpoint as GenAI traces carrying hashes rather than content, and audit and evidence events to a security information and event management system as signed webhooks. Owners are told when a run fails or an approval expires, and each agent carries a reliability score that flags bursts of failures.

Where it stops: Monitors AI running in Renly.

AI incident register Enterprise

Keeps each AI incident and near miss with severity, owner and timeline, linked to the agent, chat, automation run and evidence involved. Renly raises an incident itself when an agent’s reliability drops or someone reports a response; severity 1 and 2 incidents notify owners and admins. Records whether the privacy regulator or the DTA must be told, closes incidents with resolution and lessons, and exports to CSV.

Where it stops: Deciding whether to notify, and notifying, stays with the organisation.

Personal identifier masking

Before a tool result reaches the model, Renly masks tax file numbers, Medicare numbers and individual healthcare identifiers (checked against their check digits), passport, driver licence and Centrelink numbers when labelled, card numbers, credentials and some United States formats.

Where it stops: Applies to content passing through Renly; it is not an enterprise data loss prevention system.

Retention and legal hold Retention and legal hold: Pro Plus and Enterprise. Chat PDF and a chat’s evidence bundle: all plans. Export for a period: Enterprise

Set how long audit logs and chat transcripts are kept, export evidence bundles (records, a hash manifest and an attestation) for a chat or a period, export chats to PDF, and place a legal hold on one chat or the whole organisation so held content cannot be deleted until an admin lifts the hold.

Where it stops: Mapping AI content to your records authorities or retention classes stays with you.

Third-party tool containment MCP connections: Pro Plus and Enterprise

Tools on external MCP servers ask for approval by default, read-only hints count only from servers an admin marks as trusted, and each agent can be limited to named connectors, read only or read and write.

Where it stops: Applies to tools reached through Renly.

Single sign-on and provisioning SAML: Enterprise

Sign-in with Microsoft, Google or the organisation’s own identity provider over SAML 2.0 for verified domains, with single sign-on able to be required, and SCIM 2.0 provisioning and deprovisioning.

Where it stops: Sign-in inside the Teams tab and the Azure DevOps extension does not yet apply a required sign-in provider.

Data residency and bring-your-own model BYOLLM and BYOK: Enterprise

Renly’s databases and file storage run on Microsoft Azure in Australia East. Model calls go to the provider of the model in use, which may process them outside Australia. On Enterprise, model calls can be routed to the organisation’s own provider and stored credentials wrapped with the organisation’s own Azure Key Vault key.

Where it stops: Set each tool’s sensitivity ceiling with this processing location in mind.

Renly Academy Free

A free course on using AI and agents responsibly at work, with a short assessment, a certificate anyone can verify online, and a team completion register that a manager can export.

Where it stops: A completion record, not an accredited qualification; use alongside training the organisation mandates.

Objectives and metrics Beta

Objectives in Automations carry metrics with a baseline, a target and a target date. Each measurement records whether an agent, a person or the system recorded it and from which run, and progress is shown against target.

Where it stops: Measures what is recorded against the objective; finance attribution stays with you.

Automations

Tasks, workflows and objectives run on one engine with per-step approvals, named approvers, pre-flight checks before launch, run history and retry from a failed step.

Space

Shared pages and spaces with roles and version history. An AI edit made on a page is a proposal that a person accepts, saved as an attributed version; in spaces that require approval, agent writes go to review.

Where it stops: Outside approval spaces, an agent with edit rights saves a new version directly. Curating which sources AI may use stays with content owners.

Renly agents in Microsoft Teams and Copilot Teams and Copilot Chat: Pro Plus and Enterprise New

Renly agents answer in Microsoft Teams and Microsoft 365 Copilot Chat with approval cards, and Microsoft Copilot Studio agents can hand work to Renly agents through MCP and A2A under the same approvals, autonomy settings and evidence.

Where it stops: Built in October 2026 and being rolled out; ask Renly to confirm it is enabled for your tenant before relying on it.

Records Pro Plus and Enterprise Early access

Structured business data such as customers and cases that agents read and update under the same approvals and evidence as other actions.

Where it stops: Beta, enabled per organisation on request.

AI governance assessment Free

A free assessment on renly.ai against the Governing Intelligence framework that produces a draft AI governance strategy and roadmap in Word, kept in the browser.

Where it stops: A starting draft; the strategy is a leadership decision.