Everything you'd ask before you switch it on.

Sabiki AIRM, Agentic Identity Risk Management, is the independent Non-Human Identity Control Plane for Microsoft 365. It continuously discovers every non-human identity, AI agents included in a tenant, shows you what each one can access, scores the risk, and turns that into an independent posture you can govern, report on and improve.

It means three things: discover every non-human identity in the tenant, understand what each can reach and who owns it, and govern it, prioritise what matters, report it clearly, and keep humans in control as things change. It's the visibility layer that makes it safe to say "yes" to AI.

It's any account that isn't a person, an AI agent, a Copilot identity, a service principal, an automation account, an app integration. Each one signs in, holds permissions and can access data, but no person owns or watches it day to day. In a typical Microsoft 365 tenant they outnumber human users many times over.

Microsoft governs AI inside its own ecosystem. Sabiki adds an independent, tenant-wide layer that scores and governs every non-human identity, Microsoft-native and third-party, and benchmarks it against peers. It complements Microsoft; it doesn't replace it. (See the full comparison page.)

A managed service in plain language, confident, safe AI adoption. Most partners productise three lines on top of AIRM: NHI Onboarding Review (a one-off pre-deployment review), AI & Non-Human Identity Governance (a monthly retainer), and NHI Risk Posture (a quarterly evidence pack). Your brand, your pricing, your margin, all running off the same continuous observability.

Two ways. The AIRM licence carries a partner margin that grows with adoption (see Pricing). On top of that, the governance services you wrap around it are high-margin and almost entirely yours. The licence is the floor; the services are where the real value is.

Yes. AIRM is multi-tenant by default, one dashboard across your whole client portfolio, with risk scores and findings per tenant so you can see which customers are highest risk and prioritise remediation across the book.

Native integrations with ConnectWise, HaloPSA and Autotask turn findings into tickets automatically, with outbound webhooks for SIEM, Slack and Teams. No middleware required.

Yes, client reports are white-labelled with your logo and company name, so the executive summaries and posture reports look like they came from your team.

No. AIRM automates the discovery, scoring and reporting, and writes findings in plain English. It's built to run as a managed service without adding headcount.

Nothing is installed. AIRM is cloud-only and agentless, connecting to Microsoft 365 through the Microsoft Graph API. No endpoint software, no MX changes, nothing for the client to maintain.

It changes nothing without a person approving it, and reads identity and permission metadata, not your emails, files or documents. On day one it observes only; it doesn't change a permission, grant or identity unless you explicitly choose to act.

Yes. AIRM is designed to work within GDAP so partners can connect client tenants cleanly, without standing access where it isn't needed.

Yes, access can be revoked in one click from Entra ID at any time. You stay fully in control.

About five minutes to connect a tenant, and you can read the executive summary and full non-human identity inventory the same day. The first scan is usually the conversation-starter with a client.

It's a single, independent score for a tenant's non-human identity posture, think of it like a credit score for NHI risk. One number you can track, improve quarter on quarter, and benchmark. You get it from the free NHI Risk Score assessment, drawn from everything AIRM discovers across Microsoft-native and third-party identities. (AIRM produces the underlying risk rating.)

It's built from the signals AIRM continuously assesses across each identity, what it can access, how privileged it is, whether it's owned, whether it's dormant, and how its behaviour changes over time, rolled into one posture figure that updates every scan.

Within your own book, yes. The Sabiki NHI Risk Score^™ is calculated the same way in every tenant, so you can put one client next to the rest of your clients and the comparison means something.

We do not yet claim a market-wide peer benchmark. That needs a dataset large enough to be statistically meaningful, and we would rather say so than imply one exists. When it does, this page will say what it is built from.

No, we complement it. Microsoft secures the platform and governs AI within its own ecosystem. Sabiki adds the independent, cross-vendor non-human identity governance and a Sabiki NHI Risk Score^™ on top, so customers get more value from their Microsoft investment.

Keep them, Sabiki works alongside whatever Microsoft tier and tools a client has. Each Microsoft product solves part of the picture; Sabiki adds the tenant-wide non-human identity posture and an objective score Microsoft can't grade on itself. (See the full comparison page for where it fits across E3/E5/E7, Copilot, Agent 365 and Entra.)

Findings map automatically to 11 frameworks, including the AI-specific ones, EU AI Act, ISO 42001 and NIST AI RMF, plus DORA, ISO 27001, UK CAF, Essential Eight and more. You can generate a per-framework report in one click.

Yes. The quarterly NHI Risk Posture pack is written for leadership and is the evidence to hand an auditor, or a cyber-insurer asking about AI governance at renewal, where premiums rose 27% in 2025 (Marsh).

Per tenant, per month, across five bundles, from $149 (Basic) to $1,599 (Enterprise) MSRP, on an annual commitment. Pricing is per tenant, not per user, which is what makes the MSP economics work.

Partner margin grows with adoption, across three tiers: Registered −25% (1-24 tenants), Gold −33% (25-99), Elite −40% (100+). The higher tier applies across your whole adopted book.

Registered MSPs can run one free assessment per Microsoft 365 tenant and receive a fully featured report, including the recommended actions to remediate what it finds. It's one assessment per tenant. Executing those actions, plus continuous monitoring, re-assessment, triage and ongoing reporting, requires a paid AIRM subscription, per tenant. Why: non-human identities multiply continuously as AI adoption grows, so seeing the issues once isn't enough, fixing and maintaining them at best practice (aligned to the 11 governance frameworks) needs the subscription. You only pay once a client subscribes.

The free entry point is a free assessment per Microsoft 365 tenant, no credit card, which produces the fully featured report, including recommended actions. You pay the AIRM subscription, per tenant, only when a client decides to act on the report and keep the tenant monitored. (See "What's free, and what needs a subscription?" above.)